103514 Non-Farm Household Enterprises in Vietnam A Research Project using Data from VHLSS 2004, VHLSS 2002 and AHBS 2003. Wim Vijverberg with Hoang Thi Thanh Huong Nguyen Chien Tang Nguyen Ngoc Que Nguyen The Quan Phung Duc Tung Vu Thi Kim Mao July 2006 Table of Contents 1. Introduction and summary ...................................................................................................... 1 2. Contribution of Household Enterprises to Income and Employment..................................... 6 3. The Vietnam Household Living Standard Surveys in 2004 and 2002.................................... 9 4. Non-farm enterprises: characteristics and performance ....................................................... 11 5. Non-farm enterprises: evidence of survival and growth from panel data............................. 18 5.1 Building the enterprise panel database ......................................................................... 18 5.2 Enterprise survival, growth, start-up, and death ........................................................... 22 6. Assets and asset growth ........................................................................................................ 26 6.1. Household assets and its components. .......................................................................... 26 6.2. Assets over time............................................................................................................ 28 6.3. Household income and changes in assets over time ..................................................... 31 7. Investment climate ................................................................................................................ 34 7.1. MSEFI: An index of micro/small entrepreneur opinions about the business climate .. 34 7.2. Impact on Economic Performance................................................................................ 38 7.3. Community information................................................................................................ 41 8. Household enterprise networking strategies ......................................................................... 46 8.1. Participation in business associations or clubs ............................................................. 46 8.2. Contact with functional agencies .................................................................................. 49 8.3. Sources of information.................................................................................................. 52 8.4. Operation of the business.............................................................................................. 54 9. The impact of the investment climate................................................................................... 56 9.1 Modeling the impact of the investment climate............................................................ 57 9.2 Impact of community variables .................................................................................... 61 9.3 Impact of PCI and MSEFI ............................................................................................ 73 9.4 Summary ....................................................................................................................... 80 10. Counting household enterprises: AHBS and VHLSS........................................................... 81 References..................................................................................................................................... 84 Appendix..................................................................................................................................... A-1 A.1 MSEFI and its components for each province................................................................ A-1 A.2 Community variables ...................................................................................................... A-9 List of Tables Table 2.1: Income and sectoral contribution to income growth, by quintile, 2004 ........................ 6 Table 2.2: Labor force participation of households, by residence and household head's gender, 2004 (%).................................................................................................................................. 8 Table 2.3: Labor force participation and educational attainment of household members aged 16 and over, 2004 (column %)..................................................................................................... 8 Table 3.1: Content of the household questionnaire ........................................................................ 9 Table 3.2: Content of the community questionnaire..................................................................... 10 Table 4.1: Characteristics of non-farm enterprises, VHLSS 2004. .............................................. 12 Table 4.2: Employment and capital assets in non-farm household enterprises, VHLSS 2004 .... 14 Table 4.3: Variation of labor productivity by enterprise characteristics, VHLSS 2004a.............. 15 ii Table 4.3: Performance of household enterprises in 2002 and 2004 at 2004 pricesa ................... 17 Table 5.1. Difference among enterprises in panel households ..................................................... 21 Table 5.2: Growth in enterprise performance, VHLSS 2002 – 2004 ........................................... 24 Table 5.3: Percentage growth rate in enterprise profit for several enterprise characteristics ....... 24 Table 6.1: Assets per household ................................................................................................... 26 Table 6.2: Household income and household assets..................................................................... 28 Table 6.3: Total household assets: transitions within the distribution.......................................... 28 Table 6.4: Transitions in the distribution of assets, by asset type................................................. 29 Table 6.5: The Household Wealth Distribution in 2002-2004 ..................................................... 30 Table 6.6: Total household income and asset changes by type .................................................... 31 Table 6.7: Household income by component and asset change by type....................................... 32 Table 6.8: Change in household assets over 2002-2004 by region & income quintiles............... 33 Table 7.1: Level of provincial performance on the basis of MSEFI ............................................ 36 Table 7.2: MSEFI and component scores across regions, VHLSS 2004...................................... 37 Table 7.3: Association between MSEFI and farm and non-farm economic performance............ 39 Table 7.4: Community size ........................................................................................................... 42 Table 7.5: Conditions in the agricultural and non-farm sector, VHLSS 2002 - VHLSS 2004 .... 43 Table 7.6: Investment climate, VHLSS 2002-VHLSS 2004 ........................................................ 44 Table 8.1: Membership in business associations and clubs.......................................................... 47 Table 8.2: Enterprise scale by BAC membership status............................................................... 47 Table 8.3: Services received from business associations/clubs .................................................... 48 Table 8.4: Contact with agencies per year .................................................................................... 49 Table 8.5: Reasons for meeting with functional agencies ............................................................ 51 Table 8.6: Sources of information for pricing products................................................................ 53 Table 8.7: Information sources and enterprise operational scale.................................................. 54 Table 8.8: Delivery process of raw materials and finished outputs.............................................. 55 Table 8.9: Location of output markets.......................................................................................... 56 Table 9.1: Impact of economic conditions.................................................................................... 63 Table 9.2: Impact of agriculture.................................................................................................... 65 Table 9.3: Impact of infrastructure ............................................................................................... 66 Table 9.4: Impact of human resources.......................................................................................... 67 Table 9.5: Impact of natural disasters, fire and epidemics ........................................................... 69 Table 9.6: Impact of transport conditions ..................................................................................... 70 Table 9.7: Impact of communication facilities ............................................................................. 72 Table 9.8: Impact of financial conditions ..................................................................................... 73 Table 9.9: Impact of PCI............................................................................................................... 75 Table 9.10: Impact of MSEFI. ...................................................................................................... 78 Table 9.11: Important determinants of non-farm enterprise performance and entrepreneurship . 80 Table A.1: MSEFI aggregate and component indices for each province ................................... A-5 Table A.2: Association between MSEFI and farm and non-farm economic performance......... A-7 Table A.3: Definition and descriptive statistics of community (investment climate) variables ........................................................................................................................ A-10 List of Figures iii Figure 2.1: Average household income, by source and area of residence, 2004.............................6 Figure 4.1: Performance in household enterprises, VHLSS 2004 .................................................13 Figure 4.2: Productivity in household enterprises, VHLSS 2004..................................................15 Figure 5.1: Evolving performance: profits in 2002 and 2004........................................................23 Figure 5.2: Distribution of revenue and profit in panel and start-up enterprises, 2004.................25 Figure 6.1: Assets varying by household income ..........................................................................27 Figure 7.1: Hurdles to the operation and development of household enterprises..........................34 Figure 7.2: Urban/rural differences in hurdles to the operation and development of household enterprises ......................................................................................................................................35 Figure 7.3: MSEFI in three provinces............................................................................................37 Figure 7.4: Correlating MSEFI and PCI as alternative business climate summaries ....................38 Figure A.1: MSEFI and its components, by province .................................................................A.1 List of Text Boxes Box 5.1: Details on the Construction of the Panel Enterprise Sample ..........................................19 iv 1. Introduction and summary In a country such as Vietnam, non-farm household enterprises play a significant role in the economy. In urban areas, people make seek wage jobs in the public sector or in large private companies, but at this stage of Vietnam’s development there are not enough jobs for all who live in urban areas and seek employment. Non-farm self-employment, measured by the existence of household enterprises, provide an alternative: it draws on the creative entrepreneurial abilities of workers, it provides employment opportunities for many who are not able to find wage jobs, and it generates incomes for households who otherwise would live in poverty. In rural areas where agriculture is typically the dominant sector, non-farm enterprises are essential not just for the prosperity of the agricultural sector but also for the economy as a whole (e.g., Reardon et al., 2000; Mwabu and Thorbecke, 2004). Local agriculture cannot survive by selling to the local market: the local market is not large enough. Agricultural products must be sold on regional, national and international markets. Local producers must somehow tie into markets elsewhere, requiring communication, transportation and financial services. They need various inputs such as machinery, repair services, and fertilizer that are supplied by the non-farm sector. Successful farmers need working capital and investment funds to manage their farming operations, thus requiring access to credit. Moreover, rising farm incomes lead to increased levels of consumption, satisfied with commodities brought in by, and services produced by, non- farm entrepreneurs. Thus, the health of agriculture is intertwined with the vibrancy of the local non-farm sector. 1 There are still other reasons why the rural non-farm entrepreneurship is important for Vietnam. First, a significant part of the rural work force finds employment in non-farm production. The health of rural non-farm enterprises impacts the employment opportunities and the livelihood of large numbers of rural residents. Second, the combination of increasing agricultural productivity, an inelastic demand for food, and growing population pressure leads to a transfer of labor out of the agricultural sector. People will seek employment in the local non- farm sector, or they will migrate to urban areas. There are good reasons why it is in the interest of urban constituents to ensure that the rural economy flourishes: rural migrants increase crowding, pollution, unemployment, and the cost of public goods. Creating a healthy rural economy is one of the ways to discourage rural-to-urban migration and, more directly, smooth the transition of labor out of agriculture.2 Third, rural incomes are lower than urban incomes. This in itself is reason enough to pay careful attention to the rural “sector”, and not merely so from an agricultural perspective (World Bank, 2002). But rural-urban income differentials also provide a powerful migration incentive. Thus, in order to help the poor and at the same time prevent large-scale rural-urban migration and urban congestion, it is an urgent necessity to improve the rural business (investment) climate. This offers an opportunity to achieve primary 1 For example, Mwabu and Thorbecke (2004) argue that rural development in Sub-Saharan Africa has been a failure not merely because public policy was unfavorable to the farm sector, but rather because, in addition to exploitation and inimical treatment of agriculture, policy makers failed to create an investment climate that could support non- farm entrepreneurship. Neither part of the rural economy was given a chance to succeed. 2 E.g., Zhao (1999, 2002), de Janvry and Sadoulet (2000), Hare (2002), Johnson (2002). However, using 1990 Population Census data of China, Liang, Chen and Gu (2002) found that increased rural employment opportunities failed to deter both intraprovincial and interprovincial migration. 1 development objectives simultaneously: employment growth, poverty reduction, and desirable spatial population patterns. 3 This report examines non-farm enterprises in Vietnam as surveyed through the 2004 Vietnam Household Living Standards Survey (VHLSS), supplemented with information from the 2002 round of the same survey. The objective is to paint a full picture of household enterprises and of the entrepreneurship decision and to understand how the “investment climate” changes this picture. As defined in World Bank (2006), the investment climate consists of the political, administrative, economic and infrastructural conditions for getting a reasonable return on investment as perceived by potential private investors. As such, it includes a subjective appraisal by entrepreneurs of the enabling environment. Households are typically the smallest- scale (potential) entrepreneurs in the economy, easily overlooked in discussions of the behavior of and returns in the private sector, but in this report they are the center of attention. The VHLSS is a multipurpose survey that gathers information on topics ranging from education and health to farming operations and non-farm enterprises. The questionnaire for the 2004 round was enhanced with additional queries about household enterprises, which makes this database imminently suitable for the study at hand. The report starts off in Section 2 with a quick examination of the contribution of household enterprises to income and employment. Urban households derive 28.1% of their income from non-farm enterprise operations. This share is even higher for poor-to-median income households. Between 2002 and 2004, non-farm enterprise income grew by 11.4%, with all of this growth going to poor-to-median income households, as the same for high-income household declined. Among rural households, non-farm self-employment income constituted 17.2% of total income. This share is actually lower among low-income households and higher among household at the high end of the rural income scale. Households in the median income range (third and fourth quintile) saw the largest gains in non-farm self-employment income. Combined with the fact that more than half of the urban household and slightly over one third of the rural households have at least one non-farm self-employment member, it is clear that non- farm household enterprises are important for the standard of living of many households in Vietnam’s society. Section 3 describes the VHLSS in more detail, paying attention to the sampling design and the questionnaire instruments. There are small but significant differences between the 2002 and 2004 rounds. In particular, the investment climate variables are more accurately defined in 2004, both through questions directly addressed to the operators of household enterprises and through an extensive community questionnaire administered in rural communes. Thus, the investment climate in 2004 is described better than the change in the investment climate from 2002 to 2004. This places limits on our ability to understand the performance of enterprises and the behavior of entrepreneurial households—and the change in it—in the light of the investment climate. Section 4 turns the focus on the non-farm enterprises. The VHLSS captures 67.8 enterprises per 100 urban households and 44.3 enterprises per 100 rural households. Manufacturing (in particular, food and beverage production) is more popular in rural areas; service enterprises in the service and the hotel and restaurant sectors are more prevalent in urban 3 For further discussion on the concept and importance of the investment climate, see Vijverberg (2003) and World Bank (2006). 2 centers. Urban enterprises are larger: a typical urban enterprise generates roughly 125 percent more revenue, value added or profit than a rural enterprise; it uses more labor (1.82 workers on average) than its average rural counterpart (1.61 workers), and it is also more likely to hire employees for pay; and the VHLSS database records capital assets worth 20.8 million VND for an average urban enterprise as compared to 2.6 million VND for an average rural enterprise. Enterprises experienced substantial growth: in real terms, revenue rose by 27 percent, value added by 21 percent and enterprise profit by 15 percent. Growth was a bit faster in urban areas than in rural areas, but it was widespread and not merely concentrated among the largest enterprises. Section 5 goes a step further, asking the question how much turnover and growth exists among non-farm enterprises. This first requires the construction of a link between enterprises that are covered in the 2002 and 2004 rounds of the survey, because the design of the database does not incorporate this information. The best estimate of the annual enterprise survival rate is 83.2%. In other words, 16.8% of the non-farm enterprises cease to exist each year. On the other hand, the best estimate of the start-up rate (the percentage of enterprises in their first year of existence) is 18.1%. Stated in other terms, an average household has an 8.8% chance of starting a new enterprise in 2004. These statistics imply that, between 2002 and 2004, the population of non-farm enterprises expanded. Surviving enterprises generated more revenue, value added, expenditures, and profit in 2002 than abandoned enterprises. Comparing surviving enterprises with start-ups in 2004 illustrates the financial challenges that new enterprises face: among start-up enterprises, every financial performance measure (revenue, value added, or profit) is lower than among surviving (or panel) enterprises. Panel enterprises also provide more employment than start-ups. Revenue and profit in the median panel enterprise rose by 20 and 14%, respectively, but there is a large spread around these median values. For example, 39% of the enterprises saw revenue decline, and 43% faced a decrease in profit, whereas one fourth of the enterprises at least doubled their performance compared to 2002. Section 6 examines an important ingredient of enterprise growth: accumulation of business assets. The approach is actually broader, as the analysis considers agricultural assets, business assets, and consumer durables. Some assets may fulfill multiple purposes. They may be used on the farm or in the non-farm business (e.g., transport equipment). They may be shared between the household and the enterprise (refrigerators, TVs, furniture, transport, telephones). They may also be sold and exchanged for each other. Thus, assets are fungible, which the evidence in this report indeed indicates. An average household owns about 43.05 million VND worth of assets (which compares with an average annual income of 8.89 million VND), of which 23% are assets of a type that may be used for business purposes. The share of business assets is larger among higher-income households; 66% of the households do not report ownership of any asset that could be used for business purposes. The rate of growth in total assets among the poor kept up with that among the middle class and the wealthy—even if in absolute terms the gap widened. However, the rate of growth in business assets was the faster among middle-class (48%) and wealthy (129%) households than among the poor (17%). Of particular interest is the accumulation of business assets by source of income: (i) low-income agricultural households invest in non-farm household enterprises in an attempt to escape poverty; high-income agricultural households are 3 drifting somewhat into non-farm entrepreneurship but mostly improve living conditions for the household; (ii) small-scale business owners invest more in agricultural assets than in business assets or consumer durables, whereas households with high incomes from non-farm self- employment invest heavily into business assets while reducing holdings of agricultural assets; and (iii) households with low and median wage incomes build more agricultural than business assets, but among those with high wage incomes investments in business assets again dominate, indicating that some high wage earners may be considering transitions into private entrepreneurship. Overall, households with the greatest increase in assets have invested in business assets. Section 7 looks outside the household and examine the investment climate. This is done in two ways. First, the VHLSS questionnaire includes a series of questions about various hurdles that entrepreneurs may encounter. These are summarized into the so-called Micro and Small Enterprise Friendliness Index (MSEFI), which measures how business-friendly the environment is toward micro and small enterprises. The MSEFI may be decomposed into six subindices, pertaining to infrastructure, labor, finance, policy, registration, and corruption and security. Since the VHLSS sample is representative at the provincial level, the responses to each index may be aggregated to a province and generate input for a ranking of provinces or regions. Thus, the province with the fewest complaints is Binh Duong, with Hoa Binh a close second. Ho Chi Minh City ranks 28th, Da Nang comes in at 36th, Ha Noi ranks 58th, and the last place is taken by Can Tho. Judged by region, the Northwest scores the highest in the aggregate and, interestingly, also in every component, and the Northeast and Red River Delta regions achieve the lowest MSEFI values, again both in the aggregate and in every component. A comparison with the recently published Provincial Competitiveness Index (PCI; see VNCI, 2005) indicates that these two indices are complementary, measuring different aspects of the business climate facing the private sector. Second, many community variables are extracted from the wealth of information gathered through the VHLSS commune questionnaires. Communes reported nearly universal access to electricity, improved water quality, improved access to viable roads, and greater proximity to schools. Agricultural wages rose. However, proximity to post offices, to daily or periodic markets, and to health facilities decreased. Extension services diminished, and average land quality dropped as well. Section 8 returns to the enterprises and examines their networking strategies, which is described in four ways. First, very few enterprises participate in business associations or clubs. Second, contacts with functional (governmental) agencies are rather limited: only 14.5% of the enterprises had any contact with a government agency, and the average number of contacts for any purpose is 0.43 times per year. Urban enterprise are somewhat more likely (20%) than rural enterprises (12%) to be contacted, and the rate also varies substantially by industry and enterprise size. Third, the VHLSS questionnaire inquires about sources from which entrepreneurs gather information for pricing their products. Traders and firms operating in the same sector are most important. Urban/rural differences are slight; sectoral differences are more important. Fourth, the survey data indicate that entrepreneurs are able to provide most of the transport of raw materials that the enterprise needs and the products and services they sell to customers. One in ten enterprises sell to customers in other provinces, cities, or countries. Sections 4 through 8, and also section 2, are primarily descriptive in nature. The advantage of this approach is that the evidence is more accessible; the weakness lies in the fact that the evidence may represent biased or, worse, spurious associations. Section 9 therefore 4 considers the effect of the investment climate in a multivariate context. The analysis is necessarily broad, since the investment climate has many facets. To give a few examples, it includes quality of labor in the local economy, the formal and informal registration requirements that may hinder the entrepreneur-to-be, inspection guidelines, the availability of credit, the access to markets, the quality of the local infrastructure, natural disasters that have disrupted the economy, proximity to markets, and so forth. In section 9, the investment climate is measured through (i) a set of 50 variables measured at the (rural) commune level; (ii) the Provincial Competitiveness Index and its components that covers 42 provinces; and (iii) the MSEFI and its components that describes all 64 provinces of Vietnam. Enterprise outcomes are equally multifaceted. The analysis considers enterprise performance as measured by (i) revenue, value added and profit (both in total and per employee); (ii) enterprise growth; (iii) the household decision to operate an enterprise; (iv) enterprise survival; (v) new enterprise start-up; (vi) the household’s involvement in wage employment as an alternative way to earn a living; (vii) household assets, distinguished into agricultural assets, business assets, consumer durables, as well as the sum of them all; and (viii) sector choice for the operation of the enterprise, covering manufacturing and construction, trading, hotel and restaurant, services, and mining. The investment climate factors that appear to be most important are: population size of the community; proximity to major or large towns; proximity to markets; infrastructure; registration; and policy implementation or economic policy. Within Vietnam, the number of non-farm household enterprises is a subject of great interest, as the society explores ways to achieve higher levels of income, growth and employment. The private sector is undoubtedly important in this process, but the question arises how extensive the private sector is. People are entrepreneurial; as rational economic agents, they seek out opportunities to improve their standard of living. This creates a challenge for any government agency charged with the task of measuring the amount of production and employment in the economy. For this purpose, the Government of Vietnam collects information on household entrepreneurship through the Annual Household Business Survey (AHBS), which is a census of all business households in Vietnam. This allows a unique opportunity to examine whether the VHLSS survey covers the non-farm household enterprise population adequately. Thus, section 10 compares the VHLSS data with the AHBS. The VHLSS estimates that there are more business households (7.5 million) in Vietnam than the AHBS counts (2.9 million). However, the two data sources use different methodologies, which explains between one half and two thirds of the gap in the number of business households captured with the two survey instruments. 4 To turn this into an estimate of the understatement of production and employment, one must make assumptions about the type of business households that are not counted in the AHBS. Depending on the assumptions, the value added in the business household sector is understated by between 5.4 and 80.7% and yearround employment by between 16.0 and 80.7%. More importantly for the purpose of the present study on non-farm household enterprises and entrepreneurship, the VHLSS appears to be a suitable data source. 4 Technically, adjustment for the difference in methodology could have accounted for more than the gap between the VHLSS and the AHBS, which would have implied an undercount in VHLSS. 5 Figure 2.1: Average household income, by source and area of residence, 2004 10000 8000 Other VND 000s 6000 Nonfarm Wage 4000 Agriculture 2000 0 Total Urban Rural Source: authors’ calculation from VHLSS 2004 2. Contribution of Household Enterprises to Income and Employment One of the objectives of the Vietnam Household Living Standards Survey is to accurately measure income and expenditures. Figure 2.1 illustrates the importance of non-farm enterprises to household incomes. The average annual household income in 2004 was 5424 thousand VND, of which 1180 thousand VND (21.8%) derived from non-farm enterprise operations. In urban areas, the average household gained 2543 thousand VND from non-farm self-employment, or 28.1% of the household’s total income; in rural areas, 740 thousand VND or 17.2 % of total household income came from its non-farm enterprises. Table 2.1 differentiates the sectoral contribution to average urban and rural household income by quintile and also examines the contribution of each sectoral source (agriculture, wage, non-farm self-employment, and other sources) to the growth in household income between 2002 and 2004. Caution is advised when comparing information from two different surveys: the sampling frame is nearly identical but the 2004 questionnaire is somewhat modified, which may impact the inferences drawn from these data. In urban areas, non-farm enterprises contributed roughly 28 percent of the income in each quintile. They also contributed 28.6 to 36.4 percent of the growth in urban household incomes in the lower three quintiles. In the upper two quintiles growth in wage earnings and other income dominated the change in total household income; non-farm enterprises did not fare as well in 2004 as they did in 2002. Thus, in urban areas, non- farm self-employment income growth benefited the poor-to-medium income households. Table 2.1: Income and sectoral contribution to income growth, by quintile, 2004 Income quintile 1 2 3 4 5 Total A. Urban 6 Household income, 2004, VND 000s 2916 5042 7014 9866 19633 8892 Share of household income Agriculture 18.9 11.3 6.8 3.9 3.8 6.1 Wage 36.8 39.6 43.4 50.0 35.2 40.4 Non-farm self-employment 29.5 32.0 28.9 24.4 29.6 28.6 Other 14.8 17.1 20.9 21.8 31.4 24.9 Total 100.0 100.0 100.0 100.0 100.0 100.0 Contribution to total household income growth, 2002-2004 (%) Agriculture 9.7 14.2 6.1 -1.8 6.2 5.7 Wage 33.7 33.8 24.4 60.0 4.7 29.9 Non-farm self-employment 34.7 28.6 36.4 -2.7 -10.0 11.4 Other 22.0 23.4 33.1 44.4 99.1 53.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 B. Rural Household income, 2004, VND 000s 1565 2528 3445 4767 9209 4303 Share of household income Agriculture 59.7 51.0 44.0 40.0 33.8 40.7 Wage 23.2 27.4 26.8 27.7 22.2 24.8 Non-farm self-employment 6.1 9.6 16.2 17.4 21.4 17.2 Other 11.0 12.0 13.0 14.9 22.6 17.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 Contribution to total household income growth, 2002-2004 (%) Agriculture 36.8 33.9 22.6 34.0 23.9 28.2 Wage 43.8 45.8 28.8 21.3 23.9 27.8 Non-farm self-employment 4.6 3.5 30.4 22.7 7.6 14.5 Other 14.9 16.9 18.2 22.0 44.6 29.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 In contrast, rural households in the third to fifth quintile depend for roughly 20 percent of their income on the non-farm enterprise, but for the 40% poorest households, non-farm entrepreneurship is quite unimportant. Growth in non-farm enterprise income was important for households in the third and fourth quintiles. Thus, it was helpful for the rural middle class but it did not benefit the poor or the very rich households. Table 2.2 illustrates the importance of non-farm entrepreneurship to employment, considered from the household perspective. Thus, we find that 61 percent of all households in Vietnam have at least one member who holds at least one wage job; 73 percent engage in some sort of farm production; and 40.3 percent have at least one member who is engaged in non-farm self-employment. Many households are diversified: they draw earnings from several sectoral sources. For example, only 9.6 of the 61 percent of households engaged in wage jobs receive all of their household earnings from wage earnings. Since non-farm enterprises are the primary focus of this research project, the table highlights the diversity of household employment for the 40.3 percent of households engaged in some form in non-farm self-employment. 6.4 of the 40.3 percent do so exclusively; 12.7 percent combine it with farming operations only; 8.3 percent with wage jobs only; and 12.9 of the 40.3 percent have members involved in agriculture, wage, and non-farm self-employment all within the same year. 7 Once again, in urban areas, non-farm self-employment is more prevalent: more than half of the urban household have at least one non-farm self-employed member, as opposed to slightly over one third of the rural households. Specialization is more common in urban areas as well: 16.5 percent of all urban households are exclusively engaged in a non-farm enterprise, compared to only 3.0 percent of rural households. Table 2.2: Labor force participation of households, by residence and household head's gender, 2004 (%) Total Urban Rural Wage employment 61.1 69.2 58.3 Only wage employment 9.6 27.4 3.6 Farming 73.2 28.7 88.2 Only farming 17.4 4.2 21.8 Non-farm self employment 40.3 53.2 35.9 Only non-farm activity 6.4 16.5 3.0 With farming only 12.7 6.1 14.9 With wage employment only 8.3 23.4 3.2 With farming and wage employment 12.9 7.2 14.8 Not employed 2.4 4.0 1.9 Number of households 9,188 2,313 6,875 Similar to other countries around the world (Van der Sluis et al., 2005), educational attainment strongly influences the type of activity that people are engaged in. Table 2.3 considers the distribution across activities of household members (rather than households, since education is an individualized matter). Overall, 20.2 percent of the individuals aged 16 or older are active in non-farm self-employment. Compared with the total population, (i) adults without any education are more likely not employed or are farming; (ii) adults with a primary education are more likely found in farming and less likely not employed; (iii) adults with a secondary education are slightly more likely in non-farm self-employment and slightly less likely in farming; and (iv) those with post-secondary education are much more likely to hold a wage job. Viewed from another perspective and stated in a black-and-white manner, the least educated tend more towards farming; the most educated are usually employed for a wage or salary; and many of those in between find work in household enterprises. Table 2.3: Labor force participation and educational attainment of household members aged 16 and over, 2004 (column %) Post- No degree Primary Secondary Secondary Other Total Wage employment 23.9 33.2 33.5 76.7 43.7 33.0 Farming 54.7 61.2 49.0 13.9 40.4 51.8 Non-farm self employment 14.5 23.3 22.4 10.9 16.7 20.2 Only non-farm activity 7.4 11.2 10.1 6.2 8.4 9.5 With farming only 0.7 1.1 1.2 3.2 8.3 1.2 With wage employment only 5.6 9.8 9.8 0.9 0.0 8.4 With farming and wage employment 0.8 1.2 1.3 0.6 0.0 1.1 Not employed 28.7 12.7 23.3 12.5 26.3 21.6 Number of individuals 7436 7046 12648 1125 19 28274 8 Having illustrated the importance of the non-farm enterprise (or self-employment) sector in Vietnam, this report now proceeds with a more detailed description of the 2004 and 2002 VHLSS surveys, which is the primary data source of this study. 3. The Vietnam Household Living Standard Surveys in 2004 and 2002 The VHLSS survey approach is built around two questionnaires. The first is the household questionnaire, which collects various types of information related to the household and its members. Topics range from education and health to agriculture and non-farm entrepreneurship. The questionnaire underwent a significant change from 2002 to 2004, augmenting the range of questions asked about agricultural and non-farm enterprise operations and polishing many other topics at the same time. The 2004 version consists of 10 modules (Table 3.1). Information on non-farm enterprises appears primarily in modules 4 (income), 6 (fixed assets and durable goods) and 10 (non-farm enterprises: extended questions). Table 3.1: Content of the household questionnaire 2004 module number and title VHLSS 2004 VHLSS 2002 1: Roster Basic demographic information for all Basic demographic information for all household members household members 2: Education and training Years of schooling, training, diploma, Years of schooling, training, diploma, current study, school expenditures current study, school expenditures 3: Health and health care Illness and injuries, care (Module 4) Care 4: Income Employment and hours of work, wage (Module 3) Employment and hours of income agricultural production, non- work farm enterprise receipts and (Module 5) Wage income, agricultural expenditures, other income production, non-farm enterprise receipts and expenditures, other income, non-income money flows 5: Expendituresa Daily expenditures on food and drink, (Module 6) Daily expenditures on food non-food, other; annual consumption and drink, non-food, other; annual expenditures; non-consumption consumption expenditures; non- expenditures consumption expenditures 6: Fixed assets and durable Assets, durable appliances (separately) (Module 7) Assets and durable items goods (mixed together) 7: Accommodation Housing owned or rented, land plots, (Module 8) Housing owned or rented, water sources, electricity, Internet water sources, electricity, sanitation, TV, radio, newspapers 8: Participation in hunger Participation in poverty alleviation and (Module 9) Participation in poverty eradication and poverty hunger eradication programmes alleviation and hunger eradication reduction programmes 9: Agriculture, forestry and Further questions on agricultural topics N/A aquaculture (extended) 9 10: Non agricultural, Further questions on non-farm N/A forestry and aquacultural enterprise topics industries (extended) Note: a The Expenditures module was administered only to a subsample of households. The second questionnaire utilized by the VHLSS is the community questionnaire, which is aimed at the commune or ward in which the household resides. As such, it supplements the household questionnaire with information that is common to all residents in a community. Table 3.2 describes the sections of the community questionnaire, which cover everything from population size and infrastructure to agricultural wages and non-farm employment opportunities. Between 2002 and 2004, the community questionnaire was somewhat revised. In particular, the agricultural module is substantially enhanced, and a module on saving and lending institutions was added. One other change is important, though: the 2004 community questionnaire was administered only in rural areas, whereas the 2002 community questionnaire was filled out in both rural and urban areas. Thus, since the item changes are small enough, a comparison of commune conditions between 2002 and 2004 is warranted for rural areas. Table 3.2: Content of the community questionnaire Module number and title VHLSS 2002 VHLSS 2004a 0: Survey information Description of survey respondents Description of survey respondents 1: Basic features of Area type, numbers of households, Area type, numbers of household members and general religion registered (temporary, permanent), situation of commune migration, religion 2: General economic Main income source, general Main income source, general situation and support assessment on the changes of living assessment on the changes of living programs standard, support programs/projects standard, support programs/projects, disasters 3: Non-farm Types of enterprises and factories in and Types of enterprises and factories in and employment near the commune near the commune opportunities 4: Agriculture and Commune land fund, land areas, areas Commune land fund, land areas, areas land of main crops in commune/ward, of main crops in commune/ward, agricultural wage, difficulties in agricultural wage, agricultural agricultural production, agricultural calendar, difficulties in agricultural extension activities production, land right transfer, agricultural extension activities 5: Infrastructure Road, waterway, public transport, water Road, waterway, public transport, water source, electricity, distance to various source, electricity, travel time, distance places such as daily/periodical market, and cost to various places such as banks, post office daily/periodical market, banks, post office, infrastructure investment projects 10 6: Education Schools, difficulties in education, Schools, difficulties in education, programmes to eliminate illiteracy, programmes to eliminate illiteracy, kindergarten kindergarten 7: Health General situation of popular diseases, General situation of popular diseases, difficulties for health stations, difficulties for health stations, availability of health care services, availability of health care services, difficulties for people in accessing difficulties for people in accessing medical services medical services, transport means used 8: Public security and Social problems in the commune, drug Social problems in the commune, drug other social problems addiction, prostitution addiction, prostitution 9: Credit and saving N/A Institutions for saving and lending Note: a New items are highlighted in italics. The VHLSS 2002 sample contains 75000 households, of which 30000 filled out the expenditure module (Table 3.1). The VHLSS 2004 sample consists of 9300 households that filled out the expenditure module; 4500 of these 9300 were repeaters from the 2002 round. Thus, the two rounds provide a panel sample of 4500 households. This panel sample will be described in greater detail in Section 5. In 2002, the 30000 households were divided in four equal groups, with each group surveyed in the first month of each of the four quarters, respectively. The remaining 45000 that did not fill out the expenditure module were divided into two groups and surveyed in the first month of the first two quarters. In 2004, all households were visited in the month of March. The samples are representative of the nation of Vietnam overall and also by province. In this context, it should be noted that between 2002 and 2004 the number of provinces grew from 61 to 64: a few provinces were subdivided. This complicates the comparison of provinces over this period somewhat. 4. Non-farm enterprises: characteristics and performance The VHLSS 2004 survey captured information about 4544 household enterprises (Table 4.1). This means that there are 50.2 enterprises per 100 households. 5 As one should expect, the density of enterprises is higher in urban areas (67.8 per 100) than in rural areas (44.3). The industrial composition varies accordingly: manufacturing is more popular in rural areas, with food and beverage production providing all of the difference; service enterprises are found more frequently in urban centers. Hotel and restaurant businesses are also more prevalent in urban areas. About one half of the household enterprises operate from the household’s residence. Fifteen percent work from a place at a market; nine percent use another fixed location, which may be at an industrial zone, a trading center, a store location or some other place; nine percent use various fixed locations (e.g., selling from the home and from a market stand); and sixteen percent of the enterprises are mobile. Although spatial distances are larger in rural areas and 5 There are 9188 households in this database. Without the use of sampling weights, this suggests 49.5 enterprises per 100 households, but sampling weights must be applied, yielding the value of 50.2 stated in the text. 11 therefore customers are more dispersed, the differences between urban and rural enterprises are surprisingly small. The average age of a household enterprise is 7.7 years. Urban enterprises tend to be a little older: 8.2 versus 7.1 years. More than a third of NFHEs are fairly young. 34.4% and 41.3% of NFHEs in urban and rural areas, respectively, have been in operation less than 5 years. About 30% of all NFHEs have been in operation for 5 to 10 years. Only a few NFHEs have existed for a long time, more than 15 years. This reflects the history of economic policy towards private entrepreneurship in Vietnam. The vast majority of the enterprises are owned by Vietnamese (Kinh) or Chinese (Hoa) households. Other ethnic groups make up 3.0 and 13.5 of the households in urban and rural areas, respectively, but they operate only 1.7 and 8.2 percent of the enterprises. Table 4.1: Characteristics of non-farm enterprises, VHLSS 2004. Variables Total Urban Rural Number of enterprisesa 4,544 1,532 3,012 Percentage (row) 100.00 34.01 65.99 Number of enterprises per 100 households 50.2 67.8 44.3 Industrial Composition Mining 1.0 0.4 1.3 Manufacturing and construction 27.3 17.1 32.6 Food and beverage 11.3 4.7 14.8 Fur processing and products 4.0 4.9 3.5 Wood processing and products 5.2 0.7 7.6 Other manufacturing 6.1 6.2 6.1 Construction 0.6 0.7 0.6 Trading 45.1 45.6 44.8 Hotel and restaurant 11.0 15.6 8.6 Services 15.7 21.3 12.8 Transport 8.8 11.5 7.5 Business services 1.3 2.6 0.6 Personal services 3.0 3.7 2.6 Other services 2.6 3.4 2.1 Total 100.0 100.0 100.0 Operation Locationb Home 50.9 47.6 52.5 Markets 14.7 15.5 14.3 Another fixed location 9.1 13.8 6.7 Various locations 8.9 7.0 9.9 Mobile location 16.3 16.0 16.5 Total 100.0 100.0 100.0 Age of enterprise Average age (years) 7.7 8.2 7.5 0 ≤ Age of NFHE < 5 year 38.9 34.4 41.3 12 5 ≤ Age of NFHE < 10 year 29.9 31.2 29.2 10 ≤ Age of NFHE < 15 year 18.7 21.2 17.4 15 ≤ Age of NFHE < 20 year 5.4 6.1 5.1 Age of NFHE ≥ 20 year 7.1 7.1 7.1 Total 100.0 100.0 100.0 Ethnicity Kinh/Hoa 94.0 98.3 91.8 Other Ethnicities 6.0 1.7 8.2 Total 100.0 100.0 100.0 Notes: a Unweighted statistic b There are also 0.15% of enterprises that did not provide information about their location. Of great interest is of course the performance of the household enterprises and the amount of employment that they generate. Figure 4.1 shows the amount of revenue, value added and profit that a typical enterprise generates. The postscript “-U” indicate an urban statistic, and “-R” refers to a rural value. In these descriptions, the few really large enterprises tend to distort the mean values, and for this reason the figure also shows the median value. In each case, the median value is substantially smaller. For example, for the overall sample, enterprise revenue has a mean of 28.7 million VND and a median of 9.9; the same for value added is 16.5 and 7.2, and for profit 12.3 and 6.6, respectively. The distribution of these variables is therefore strongly Figure 4.1: Performance in household enterprises, VHLSS 2004 Revenue 28.7 9.9 Value added 16.5 7.2 Profit 12.3 6.6 Revenue-U 42.3 16.8 26.9 Mean Value added-U 12.4 Median Profit-U 19.1 11.2 Revenue-R 21.8 7.4 Value added-R 11.2 5.3 Profit-R 8.7 4.9 0 10 20 30 40 50 million VND Source: “-U” denotes an urban statistic; “-R” denotes a rural statistic. Calculations based on VHLSS 2004 13 right-skewed: it has a long right tail. The values presented in the figure also indicate that, regardless of whether one considers revenue, value added or profit, a typical urban enterprise generates roughly 125 percent more than a rural enterprise, for both the mean and the median. It is a bit surprising to see that the ratio of the urban to rural mean values is not much different than the ratio of the urban to rural median: apparently, the “typical small” and “typical large” enterprise differs by location to the same degree. Why is this so? There are many possible answers, but the first one is likely to be the quantity of labor and capital that these enterprises use. Indeed, urban enterprises use more labor than their rural counterparts: 1.82 workers on average as compared to 1.61 workers. They are also more likely to hire employees for pay. More importantly, an average urban enterprise utilizes capital assets worth 20.8 million VND, whereas an average rural enterprise has assets valued at 2.6 million VND; moreover, 55.2 percent of urban enterprises and 72.3 percent of rural enterprises reported no capital. (See Section 6 for more information on capital assets.) Table 4.2: Employment and capital assets in non-farm household enterprises, VHLSS 2004 Variables Total Urban Rural Employment: Number of workers Percentage distribution 1 67.10 64.47 68.46 2 22.10 20.68 22.86 3 5.24 7.11 4.26 4 2.63 4.07 1.87 5 0.88 1.17 0.73 6-10 1.15 1.48 0.98 11-20 0.62 0.64 0.61 > 20 0.38 0.15 0.23 Average number of workers 1.68 1.82 1.61 Enterprises with paid workers 10.43 15.76 7.62 Capital assets Average value (000VND) 7789 20797 2635 Enterprises with zero capital assets (%) 67.5 55.2 72.3 Source: Calculations based on VHLSS 2004 In view of the variation in size of the enterprises, a good measure of enterprise performance is labor productivity: it standardizes for the amount of labor used to generate the output. As Figure 4.2 shows, the average revenue per employee equals 14.1 million VND, leading to a value added per employee of 9 million VND and a profit for the enterprise of 7.6 million VND. The number of employees here counts both household members working with or without compensation in the enterprise and other workers hired by the enterprise. Unfortunately, we are not able to express this labor productivity per day or per hour of work, because the survey information is not adequate. Once again, there is a sizable gap between urban and rural enterprises: the labor productivity in urban enterprises is almost double that of rural enterprises. 14 Many factors contribute to the variation in labor productivity. Table 4.3 explores a number of them in a descriptive manner. Sectors with high value added per employee include all services except personal services, trading, hotel and restaurant, construction, and other manufacturing. Manufacturing of food, beverage, fur products and wood products are less rewarding. Labor productivity varies by enterprise location: the highest levels are found among enterprises operating from “other fixed locations”, in particular in trading zones, industrial zones and independent stores (not shown in the table). As one might expect, the enterprise age is significantly correlated with labor productivity. Young enterprises experience lower productivity; those that survive are more prosperous, as do enterprises operated under Kinh or Hoa ownership, and those that employ more workers. Figure 4.2: Productivity in household enterprises, VHLSS 2004 14.1 Revenue / 19.9 Employee 11.0 9.0 Total Value Added / 13.5 Urban Employee 6.6 Rural 7.6 Profit / Employee 11.2 5.7 0 5 10 15 20 25 million VND Source: Calculations based on VHLSS 2004 Table 4.3: Variation of labor productivity by enterprise characteristics, VHLSS 2004a Revenue / Value Added / Profit / Variable Employee Employee Employee Total Mean 14.1 9.0 7.6 Industrial Composition Mining 11.1 8.6 6.7 Manufacturing and construction 14.6 6.1 5.0 Food and beverage 16.2 5.3 4.6 Fur processing and products 9.1 6.8 6.1 Wood processing and products 6.1 2.8 2.6 Other manufacturing 22.9 9.4 7.1 Construction 16.3 9.6 4.7 Trading 11.3 9.4 8.2 Hotel and restaurant 22.3 9.3 8.3 Services 15.8 12.1 9.9 15 Transport 16.9 12.2 9.3 Business services 18.3 15.8 13.3 Personal services 9.8 8.1 7.1 Other services 17.4 14.4 13.0 Operation Location Home 13.9 8.0 6.9 Market 12.8 9.9 8.7 Other fixed location 18.9 12.7 9.6 Various locations 14.5 9.4 8.0 Mobile location 13.0 8.9 7.6 Age of enterprise 0 ≤ Age of NFHE < 5 year 12.5 7.2 6.6 5 ≤ Age of NFHE < 10 year 14.8 9.8 8.3 10 ≤ Age of NFHE < 15 year 15.7 9.9 8.5 15 ≤ Age of NFHE < 20 year 17.9 10.2 9.6 Age of NFHE ≥ 20 year 12.8 8.6 7.5 Ethnicity Kinh/Hoa 14.4 9.2 7.8 Other ethnicities 7.5 5.1 4.6 Employee scale Number of employees = 1 12.0 8.3 7.5 1< Number of employees ≤ 5 17.8 9.9 7.8 Number of employees > 5 29.2 16.1 7.3 Notes: a Values given in million VND Source: Authors’ computation based on VHLSS 2004 We conclude this descriptive exploration of the household enterprise sector in Vietnam with a comparison with 2002. The measures of enterprise performance in focus here are revenue, value added and profit as a total—rather than per employee, since the 2002 survey did not collect information on employment in the enterprise. To account for inflation, the values for 2002 are expressed at 2004 prices. 6 Thus, in real terms, revenue rose by 27 percent, value added by 21 percent and enterprise profit by 15 percent (Table 4.3). In fact, not just the average but almost the whole distribution of each enterprise performance measure shifted up: only the smallest and lowest-income enterprises around the fifth percentile failed to participate in this growth. Of course, the performance of individual enterprises fluctuates. Enterprises do not stay put at their position in the performance distribution. But for every enterprise that saw a diminishing performance, another saw its performance rise that much more, and a business that had closed its doors was replaced with a more productive start-up. (We shall examine panel enterprises in Section 5.) The message of Table 4.3 speaks about the aggregate: there was a widespread rise in the performance of household enterprises, at least when measured by the three variables of revenue, value added, and enterprise profits. 6 Allowing for 11.113 percent of inflation over the length of this period. 16 Table 4.3: Performance of household enterprises in 2002 and 2004 at 2004 pricesa VHLSS 2004 VHLSS 2002b Value Value Revenue added Profit Revenue added Profit Mean 28.7 16.5 12.3 22.6 13.6 10.7 Distribution (percentiles) 5 1.14 0.75 0.69 1.13 0.82 0.75 25 4.72 3.32 3.03 4.00 3.07 2.87 50 (median) 9.90 7.20 6.60 8.09 6.33 5.89 75 22.87 15.25 13.48 18.00 13.01 11.56 95 96.35 55.59 39.60 68.01 45.35 33.89 Residential location Urban 42.3 26.9 19.1 37.0 21.6 16.5 Rural 21.8 11.2 8.7 15.2 9.5 7.7 Industrial Composition Mining 29.6 23.6 15.2 17.3 12.3 7.6 Manufacturing and construction 40.2 14.8 9.8 33.7 13.7 9.7 Food and beverage 32.5 9.4 7.3 40.1 9.9 8.2 Fur processing and products 14.0 10.7 8.8 15.5 10.1 7.8 Wood processing and products 15.2 6.7 5.6 11.2 5.7 4.6 Other manufacturing 83.4 28.7 16.8 45.1 22.7 15.9 Construction 139.9 73.3 26.5 111.9 67.0 22.1 Trading 19.7 16.2 12.7 14.3 12.8 10.5 Hotel and restaurant 40.2 16.7 14.1 33.3 16.1 14.1 Services 26.7 19.9 13.8 19.6 14.5 10.9 Transport 29.6 20.4 13.5 22.3 16.1 11.6 Business services 46.3 39.2 22.3 26.2 16.9 11.8 Personal services 13.3 11.1 9.2 13.0 10.4 8.9 Other services 22.2 18.2 15.9 17.2 14.0 11.0 Ethnicity Kinh/Hoa 30.0 17.2 12.7 23.7 14.2 11.1 Other Ethnicities 8.9 6.0 5.2 5.8 3.9 3.5 Employee scale Number of employees = 1 12.0 8.3 7.5 16.4 11.1 8.5 1 < Number of employees ≤ 5 46.5 25.3 19.2 38.1 20.1 15.9 Number of employees > 5 348.7 173.0 73.2 40.4 24.1 22.4 Notes: a Values are in million VND b 2002 sample includes 13,532 NFHEs Source: Calculations based on VHLSS 2002 and VHLSS 2004 The growth occurred in both urban and rural areas. In urban areas, value added rose by 25 percent, which was a bit faster than that in rural areas (18 percent). Across the different sectors, the fastest growth occurred in business services, mining, other services, trading, other manufacturing, and transport, in this order. The remaining sectors lagged behind the overall average. Enterprises operated by other ethnicities experienced a 51 percent rise in value added. The last panel of Table 4.3 is especially revealing: enterprises with only one self-employed 17 worker stagnated and actually lost ground, but enterprises with more than five workers saw rates of growth of several hundred percent. Detailed inspection of the survey data shows that this is not an isolated incident but reflect a large shift in the whole distribution of all performance indicators in this subsample. It should be noted again, however, that growth is better studied with panel data than with repeated cross-sections. The inferences drawn in the previous paragraph are accurate if enterprises stay within the subgroups and do not move endogenously from one group to another. But this is precisely what may have taken place. For example, a successful one-person enterprise in 2002 may well expand and hire several workers. This growth behavior removes a high- performance enterprise from the lowest category and adds it to a higher category, thus destabilizing the descriptive statistics of both categories. For this reason, it is so fortunate that the VHLSS 2004 incorporated a panel aspect by revisiting a portion of the 2002 households. The next section delves into the construction and analysis of the panel enterprise sample. 5. Non-farm enterprises: evidence of survival and growth from panel data 5.1 Building the enterprise panel database Before proceeding with a description of the survival and growth experience of enterprises from 2002 to 20004, it is necessary to discuss the construction of the enterprise panel itself. The VHLSS 2004 was administered to 9188 households. About one half of these were designated to be repeat visits to 2002 households, but since households move or cease existence, it is to be expected that the actual number of panel households is somewhat lower. Indeed, the records of the 2004 database indicate that 4476 households were supposed to be panel households for which information had been collected in 2002, complete with information on the demographic composition of these households in 2002. Checks on these households revealed that errors had crept into the data. In particular: • 42 households had to be removed because they were linked to a 2002 household that could be linked more accurately to another member of the 2004 sample. • 3709 households were straightforwardly linked between 2002 and 2004, occasionally with a minor correction to the demographic information about household members. • 177 households listed clusters in 2002 where the 2002 survey actually had not drawn any observations. • 548 households in 2004 did not appear to match up with the designated household in 2002. However, a comparison with all other households in the same 2002 cluster on the basis of demographics (gender and age) and educational attainment identified 346 additional panel households. In all, therefore, we will work with 4055 (= 3709 + 346) panel households and discard the remaining 421 (= 42 + 177 + 202) households. 18 Box 1: Details on the Construction of the Panel Enterprise Sample. The 4055 households operated 1851 enterprises in 2002 and 1969 enterprises in 2004. Of these, 840 match uniquely and unambiguously. This leaves 2139 records in the combined databases (1011 in 2002a, and 1129 in 2004). 854 may be removed from this because there is no enterprise in the other year: 384 in 2002 and 470 in 2004. The remainder (1285 records) must be inspected manually, on the basis of industry code, enterprise age in 2004, the identity of the entrepreneur in 2002, his activity in 2002 and 2004, and the length of his experience as recorded in 2004, revenue in 2002 and 2004, and licensing in 2002 and 2004. Matches are determined through all of these factors, which frequently work together to determine an unambiguous match. Of the 1285 records, 798 are linked, implying 399 matched enterprises. Overall, there are 609 enterprises in 2002 that fail to be matched. Among these: 19 enterprises have been clearly abandoned; 549 had no possible match in 2004 because no other enterprise was reported in 2004; 39 enterprises could not be matched even though there were unmatched enterprises in 2004 as well; and for 2 the link is possible to establish but too uncertain to call. Also, 730 of the 2004 enterprises fail to be matched. Among these: 286 are probably new start-ups, as they cannot be linked and started in 2002 or later; 366 had an enterprise age of 2001 or earlier; and 37 enterprises could not be matched even though there were unmatched enterprises in 2002 as well; and for 2 the link is possible to establish but too uncertain to call. This leaves 1239 panel enterprises. Not all of these always have the same industry code in 2002 and 2004. The coding of industry is subject to substantial variation. It happens frequently that the 2002 enterprise is listed under one code and the entrepreneur's economic activity under another, and that the latter code is the one under which the 2004 enterprise is listed, whereas the entrepreneur in 2004 reports many years of experience in that work. The entrepreneur experience frequently corresponds with the enterprise age, which is another suggestion about the identity of the panel enterprise. There are indications in the experience information that suggest that some self-employment activities in 2002 were not recorded at all. __________________ Note: a One enterprise was dropped because of crucial missing information on the industry affiliation. In regard to enterprises, the 2004 questionnaire makes no attempt to identify a link with 2002. The 4055 households operated 1851 enterprises in 2002 and 1969 enterprises in 2004. A tedious procedure that is outline in detail in Box 1 yields an enterprise panel consisting of 1239 enterprises; 568 enterprises that were in operation in 2002 but appear to have been abandoned by 2004; 286 enterprises in 2004 that started up between the two surveys; and another 366 enterprises in 2004 that reportedly had started in 2001 or earlier but had not been reported in 2002. Furthermore, there are 41 enterprises in both years that cannot be matched with each other, because the information in the two years shows too much discrepancy (or, in the case of one pair in one household, too much similarity!). There are different conclusions one may draw from this. First, if over a period of two years only 1239 out of a total of 1851 enterprises survive, the annual enterprise death rate equals 1 − (1239 1851) = 18.4% . This compares quite well with the annual enterprise death rate of 0.5 17% computed by Vijverberg and Haughton (2004) for the period of 1993-1998. However, the last-mentioned group of 41 enterprises might be considered survivors as well, even if they are 19 not identifiable across the two years. In that case, the enterprise death rate equals 16.8%, using the same type of calculation. 7 Second, if only 286 enterprises started operation in 2002 or later—which is actually a generous interpretation, given that the 2002 survey took place during 2002 and not merely at the beginning of it—, the start-up rate is equal to 7.9%. 8 If this number is accurate, enterprise death is much more common than enterprise start-up and the number of non-farm household enterprises should drop rapidly over time. Nevertheless, from 2002 to 2004, the VHLSS sample uncovered more, not fewer, household enterprises. This paradox derives from the 366 enterprises that appear in the VHLSS 2004 but reportedly started before 2002, yet were not recorded in the VHLSS 2002. In past surveys, it is known that entrepreneurs do not provide particularly accurate answers about the age of the enterprise. In the panel enterprises of VLSS 1998, the average age was only 3.8 years more than their reported age in 1993, five years prior (Vijverberg and Haughton 2004). For 53 percent, the reported increase in enterprise age deviated from 5 by more than 2 year. Unfortunately, the VHLSS 2002 did not record the age of the enterprise; thus, we cannot test whether a quicker repeat visit generates more accurate responses. Suppose we follow the strategy of Vijverberg and Haughton (2004), counting all of the 366 enterprises as start-ups as well. Then, the start-up rate rises to 18.1%, which is higher than the estimated death rate of 16.8% recorded above. Thus, it is reasonable that the sample of household enterprises expanded between 2002 and 2004. However, as 35 percent of the 366 started in 1995 or before, this strategy would completely disregard the responses about the starting year of these 366 enterprises. A more reasonable strategy might be to count those that reportedly started operation in 2000 or 2001 (116, i.e., 32 percent of the 366) as start-ups: with this allowance for response error, the start-up rate would still rise to 11.1%. The remaining 250 enterprises would have already existed in 2002, together with an estimated additional 95 enterprises 9 that had not been reported but had ceased operation between 2002 and 2004, and the total 2002 sample should have been 1851 + 250 + 95 = 2196 household enterprises. In this way, the household enterprise sector appears to be in decline, from 2196 in 2002 to 1969 in 2004, which is then consistent with the finding that the death rate exceeds the start-up rate. Which of these two strategies is more acceptable? Table 5.1 sheds more light on the various subsamples and helps resolve the issue. For VHLSS 2004, the table contains five columns: the first describes the panel enterprises; the remaining four describe the other enterprises in the panel households. The first of these four shows the 286 enterprises that started up after the 2002 round of the survey. The other three contain the 366 enterprises with starting 7 One might argue for a third estimate: it is possible that the group of 366 enterprises in 2004 with a starting date in 2001 or earlier actually did exist in 2002. Thus, the number of enterprises in 2002 should have been 1851 + 366 = 2217, of which 1239 + 41 + 366 = 1646 survived, yielding an annual enterprise death rate of 13.8%. However, consider that these 366 in 2004 were originally more than 366 in 2002. In fact, if d is the enterprise death rate, the original group was equal to 366 / (1-d)2. Building this into the calculation gets us back to the previous estimate of 16.8%. 8 This is computed as follows. Of the 2004 sample of 1969 enterprises, 286 are new. Let s denote the start-up rate and d the annual enterprise death rate, which is set at 16.8%. Enterprises starting up in the first year after the 2002 are at risk of death; those in the second year are assumed to survive to the survey date. Thus, 286 = 1969 (s(1-d)+s). Solving for s yields s = 0.0793. 9 The 250 enterprises in 2004 come from a pool of 250 / (1-0.168)2 = 345 enterprises in 2002. Thus, between the surveys, 95 enterprises ceased existence. 20 dates in 2001 or before as a group, and subdivided in subsamples with starting dates in 2000- 2001 and in 1999 and before. Statistical tests show no statistically significant differences between the 2002-2004, 2000-2001 and 1999-and-before start-up subsamples in any enterprise performance value or in residential location, and there are only two differences in industrial composition and enterprise location: (i) the industrial composition of the 1999-and-before subsample is more concentrated in manufacturing and construction and less in services and trading, and (ii) enterprise location of the 2000-2001 subsample is oriented less towards home and more towards a market location. Thus, the 366 enterprises in the 2004 sample that have a reported starting data before 2002 but yet were not observed in the 2002 sample are mostly statistically indistinguishable from the start-up enterprises that reportedly started between the surveys. For this reason, we will conclude that the 366 enterprises are indeed enterprises that truly started up between the 2002 and 2004 rounds of the VHLSS survey, and we will disregard the stated enterprise age information. In further defense of this (difficult) conclusion, it should be noted that any other conclusion would imply that the number of non-farm household enterprises in 2002 is understated, i.e., that a significant number of non-farm income-earning activities had been unreported. But the same argument might apply to the 2004 sample. Yet, there is consistency in the 2004 database in that whenever any household member mentioned a non-farm self- employment activity, there is a report on a non-farm household enterprise, 10 and the same consistency exists in the 2002 database. 11 Thus, it is hard to explain why there are 366 enterprises in 2004 that reportedly started before 2002 but did not show up in the 2002 sample. Table 5.1. Difference among enterprises in panel households 2002 2004 Enterprises reportedly starting up in 2004 2001 1999 Abandoned Panel Panel and 2002- and 2000- and Variable Statistic enterprises enterprises enterprises before 2004 before 2001 before Number of enterprises count 568 1239 1239 652 286 366 166 198 Residential location Urban % 23.8 32.3 32.9 25.2 26.1 24.4 22.2 26.6 Rural % 76.3 67.7 67.1 74.8 73.9 75.6 77.9 73.4 Industrial composition Mining % 2.9 1.1 0.9 1.2 1.8 0.7 1.2 0.4 Manuf & construction % 31.3 25.5 27.7 30.3 26.7 33.1 26.6 39.0 Trading % 36.2 47.3 44.8 40.0 37.0 42.3 42.4 41.6 Hotel & restaurant % 5.9 11.0 13.0 9.8 14.1 6.5 7.0 6.1 Services % 23.8 14.6 13.5 18.7 20.4 17.4 22.8 13.0 a a a a a a Agric/forestry/aquac. % 0.0 0.4 10 The structure of the questionnaire allows a respondent to indicate in three places whether he/she is self-employed in a non-farm activity. The statement in the text applies to the first, global response. The second and third response are given in detailed descriptions of the activity. Among 9188 households, there are only 33 inconsistencies between the type of work that the respondent undertook and the sector of activity of the enterprise (if any). 11 With the exception of three household enterprises being recorded when no member reported any non-farm self- employment, and one household that failed to report on a household enterprise when according to member reports it should have. When there are 29530 households, that still indicates quite clean data. 21 Enterprise location a a Home % 52.6 50.3 51.4 49.5 43.9 54.2 a a Market % 16.6 11.3 8.5 13.6 15.0 12.0 a a Other fixed location % 8.1 10.9 12.5 9.6 13.2 6.6 a a Mixed locations % 9.4 7.1 8.0 6.4 5.7 7.0 a a Mobile % 13.4 20.4 19.6 21.0 22.3 20.1 Enterprise performanceb Revenue mean 12.4 21.9c 32.2 15.8 17.8 14.2 14.7 13.7 median 6.1 9.6 11.2 7.2 7.2 6.9 6.9 6.8 st.dev. 28.9 43.8 92.5 43.3 59.6 23.3 25.9 20.9 Value added mean 8.6 14.6c 17.9 9.5 9.8 9.2 9.9 8.7 median 4.6 7.0 8.1 4.8 4.9 4.6 4.7 4.6 st.dev. 15.7 28.1 45.2 16.6 19.2 14.1 14.9 13.5 Expenditures mean 5.1 10.6c 19.0 8.4 10.3 6.9 7.0 6.9 median 1.2 2.2 2.9 1.8 2.2 1.5 1.8 1.4 st.dev. 17.4 31.4 75.5 37.2 52.4 17.4 19.9 15.2 Profit mean 7.3 11.4c 13.2 7.4 7.5 7.2 7.7 6.8 median 4.2 6.3 7.4 4.6 4.6 4.6 4.6 4.3 st.dev. 12.7 16.3 22.7 9.6 10.7 8.7 9.3 8.2 a a Number of workers mean 1.8 1.6 1.9 1.4 1.3 1.5 a a st.dev. 3.3 4.2 6.2 1.2 0.9 1.4 a a Age of enterprise mean 9.2 5.0 1.2 8.0 4.0 11.4 a a st.dev. 7.3 5.7 0.8 6.2 0.8 6.7 Notes: a Statistic is not available from the questionnaire b Monetary values are in million VND, in 2004 prices. c The mean and standard deviation is seriously distorted by one outlying enterprise in urban manufacturing and construction. With this enterprise, the mean values are: revenue 37.2, value added 15.9, expenditures 24.9, and profit 12.3. Source: authors’ computation from VHLSS 2002 and 2004. 5.2 Enterprise survival, growth, start-up, and death On the basis of the discussion in Section 5.1, the enterprise death rate is estimated at 16.8%; in other words, an average household enterprise has a 16.8% chance of not being in operation one year later. The start-up rate is 18.1%. This says that there is an 18.1% chance that any typical enterprise is in its first year of operation. Sometimes a start-up rate is expressed relative to the number of households rather than the number of enterprises. In this way, the start- up rate is the ratio of 356 (new enterprises in 2004, equal to 0.18075 × 1969) and 4055 (panel households), or 8.8%. In other words, an average household has an 8.8% chance of starting a new enterprise in 2004. Table 5.1 facilitates the comparison between enterprises that were abandoned and those that continued, and between surviving enterprises and start-ups. Let us start with the 2002 enterprises. Those that continued on into 2004 generated more revenue, value added, expenditures, and profit than abandoned enterprises. For example, the average profit of panel enterprises in 2002 is 11.4 million VND (omitting one outlier), whereas abandoned enterprises brought in 7.3 million VND (values given in 2004 prices). Rates of enterprise survival were higher in trading and hotel and restaurant; those in manufacturing and construction, services, and mining were more at risk of termination. 22 Figure 5.1: Evolving performance: profits in 2002 and 2004 14 12 10 ln(Profit 2004) 8 6 4 2 2 4 6 8 10 12 14 ln(Profit 2002) Table 5.1 also invites a comparison of the circumstances of panel enterprises over time. The difference in urban/rural residential location is caused by a change in designation of the cluster where the household resides. The aggregate industrial composition does not change much over time, but there are significant portions in each industry that switch: 7 percent out of manufacturing and construction (mostly into trading); 15 percent out of trading (into hotel and restaurant, as well as manufacturing); 24 percent out of hotel and restaurant (almost all into trading); and 19 percent out of services (into trading and manufacturing). Statistics on enterprise performance in 2002 and 2004 suggest that a robust amount of growth occurred among non-farm household enterprises. The median enterprise saw its revenue rise from 9.6 to 11.2 million VND; the average rose from 21.9 to 32.2. Similar increases are found with respect to value added, expenditures and profit. Table 5.2 elaborates on this by describing the growth rate that each enterprise experienced. The mean growth rate in revenue was 78% and in profit 57%. The standard deviation is large, indicating that the average growth performance may not reflect the typical (median) enterprise. Indeed, revenue and profit in the median enterprise rose by 20 and 14 percent, respectively. Another aspect that is buried in aggregate statistics (such as Table 5.1) is the fact that performance worsened in some enterprises: 39 percent of the enterprises saw revenue decline, and 43 percent faced a decrease in profit. On the other hand, one fourth of the enterprises did well, at least doubling their performance 23 compared to 2002. Figure 5.1 shows variation in profit growth performance in full detail. Log scales are used to minimize high-end outliers. The 45-degree line indicates equal performance: above the line, profits grew; below the line, they shrank. One unit vertically indicates 171 (= 100(e1-1) ) percent increase in performance. As much as profits varied between the two rounds of the survey, the correlation between these two log-profit variables is still 0.73. Undoubtedly, some of this variation is driven by measurement error, but even if this exists, the aggregate average performance measures improved significantly over this two-year period. Table 5.2: Growth in enterprise performance, VHLSS 2002 – 2004 Percentage Percentile % with positive growth in Mean St.dev 5 25 50 75 95 growth rate Revenue 78 234 -73 -23 20 100 383 61 Value added 57 169 -70 -26 14 82 304 58 Expenditures 347 1605 -88 -36 31 170 1271 60 Profit 57 174 -70 -27 14 82 307 57 An interesting question of course is whether this growth performance was widespread or else only concentrated in parts of the economy. Table 5.3 sheds light on this question, focusing for brevity’s sake only on enterprise profit. Rural growth rates in profits were higher than urban growth rates. Enterprises operating from the residence experienced lower growth rates at the median but performed average overall. Judged by the age of the enterprise, the divergence in the mean and median among the youngest enterprises is striking: many young enterprises struggle, but a few grow enormously fast. Growth rates in services and hotel/restaurant lag behind; mining runs ahead, but that is a small sector. In order to see whether changes in the industrial orientation of the enterprise were driven by profit motives, Table 5.3 also considers subsamples. Those that move out of manufacturing and construction fared better, but enterprises moving in had a harder time. On the other hand, those moving into the trading and hotel and restaurant sectors increased their profit at a higher rate. Enterprises moving out of trading, moving out of hotel and restaurant, and staying in hotel and restaurant were generally worse off than the overall average. In the service sector, the median stayer increased profit only by 5 percent. Table 5.3: Percentage growth rate in enterprise profit for several enterprise characteristics Mean (overall = 57%) Median (overall = 14%) Overall Stay in Move out Move in Overall Stay in Move out Move in Industrial composition Mining 165 177 3 66 72 72 23 66 Manuf & construction 55 58 138 30 12 15 63 -3 Trading 57 53 48 89 19 17 8 36 Hotel & restaurant 67 67 67 67 3 -2 5 21 Services 44 44 65 42 5 5 36 13 Residential location Urban 41 5 Rural 65 17 24 Enterprise location Home 59 5 Market 47 19 Other fixed location 66 23 Mixed locations 86 44 Mobile 43 17 Age of enterprise Age 0-4 68 7 Age 5-9 54 18 Age > 10 53 14 Comparing panel enterprises with start-ups illustrates the financial challenges that new enterprises face. Among start-up enterprises, every financial performance measure is lower than among panel enterprises. Graphically, this is also pointed out in Figure 5.2, showing the cumulative distribution of revenue and profit for both panel and start-up enterprises. The horizontal displacement of the curves between panel and start-up enterprises is roughly 0.5 from top to bottom, implying that the whole distribution of both revenue and profit among start-up enterprises lies 40% (= e-0.5 – 1) below that of panel enterprises. Viewed otherwise, start-up enterprises with revenue or profit in the upper middle quartile (50th to 75th percentile) would rank in the lower middle quartile (25th to 75th percentile) among panel enterprises. The more established enterprises also employ more workers: 1.8 against 1.6. They more likely operate from a market location, from the home, or from mixed fixed locations; newer enterprises are more likely mobile or operate from other fixed locations (shops, industrial zones, trading zones). Correspondingly, panel enterprises are more often involved in trading or hotel and restaurant; start-ups are more likely to operate in services or in manufacturing and construction. And whereas two out of three panel enterprises are rural, three out of four start-ups are rural, as a reflection of the higher turnover in rural areas. Figure 5.2: Distribution of revenue and profit in panel and start-up enterprises, 2004 100 Cumulative percentage 80 Panel: Revenue 60 Panel: Profit Start-up: Revenue 40 Start-up: Profit 20 0 4 6 8 10 12 14 Log value 25 6. Assets and asset growth 6.1. Household assets and its components. This section turns our attention to the household assets recorded in the VHLSS surveys of 2002 and 2004. In particular, it focuses on the panel sample of households that were surveyed in both years. This allows us to examine changes in the distribution of assets between these two years, the growth in each household, and the relationship with income and saving. The questionnaires list approximately 60 types of assets, which must be aggregated somehow to make any study of them manageable. We boil them down to agricultural assets, business assets, and consumer durables that are used in the household only. The background paper to this section (Nguyen and Vijverberg, 2006) describes the aggregation process, considers also a four-part disaggregation (adding a category of assets for use both in non-farm enterprise production and agriculture), and provides additional tables. The list of assets is virtually identical between the two rounds of the survey, and although the questionnaire design differs slightly, this is unlikely to cause errors of inference. 12 Assets are valued on the basis of the original purchase price and the reported age relative to the expected asset lifecycle, adjusted for inflation. How much assets do households own? Table 6.1 offers insights. An average household owns about 43.05 million VND worth of assets. This compares with an average annual income of 8.89 million VND. As is often the case with financial household outcomes, the asset distribution is very skewed. The median household actually owns assets worth 19.2 million VND. In any case, productive assets that may be used for business purposes constitute 9.74 million VND, and agricultural assets 14.23 million VND. This average includes large numbers of households that do not own business or agricultural assets; the averages are taken over the whole sample. Urban households understandably hold more business assets and fewer agricultural assets than rural ones. Consumer durables are also much greater in urban areas. Across regions, one observes significant differences. Generally, southern households hold more assets than those in the north, especially when it comes to business and agricultural assets; consumer durables do not differ by that much. Table 6.1: Assets per household Agriculture Business Consumer durables Total assets Total 14.23 9.74 19.08 43.05 Urban/rural residence Urban 8.48 26.87 36.80 72.15 Rural 16.18 3.95 13.09 33.22 12 While the respondents indicated in the 2002 round the proportion of time that a given asset was used for production or for daily living, in 2004 the enumerator composed one list for assets in household use and another for productive assets. Thus, it seems that the 2002 strategy would elicit more reports of shared assets. Actually, the 2004 strategy indicated more overlap (6%) between the two lists than the 2002 strategy would have suggested, when 2.6% of the assets were shared between household and productive uses. 26 Region Red River Delta 8.03 8.16 20.27 36.46 Northeast 10.00 9.15 19.14 38.29 Northwest 10.62 1.60 10.39 22.60 North Central Coast 7.46 3.71 12.45 23.61 South Central Coast 8.43 13.71 18.25 40.38 Central Highlands 34.10 9.51 17.21 60.82 Southeast 20.18 15.85 31.95 67.98 Mekong River Delta 22.28 10.71 14.11 47.10 Note: Values in million VND Figure 6.1: Assets varying by household income A: Cumulative distribution B: Percentage distribution 100% 160 80% Million VND 120 60% Percent 80 40% 40 20% 0 0% 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Income deciles Income decile Agriculture Business Consumer durables Agriculture Business Consumer durables Source: VHLSS 2004 Assets and income are of course related. Figure 6.1 shows value and the percentage distribution of assets by income decile. Total assets rise slowly but steadily from the first to the ninth decile, with a slight quickening at the fifth. While agricultural assets rise in value, its share falls from 63 to 29%, after briefly rising from 35 to 43% between decile 5 and 7. In the lower five deciles, the drop in the agricultural share is accompanied by an increase in consumer durables; the share of business assets rises only from 3.0 to 8.6%. Business assets rise faster between deciles 5 and 9, rising from a share of 8.6% to 21%. Then in the tenth decile, the average total asset holding doubles relative to the ninth decile. Each component rises, but panel B indicates that business assets rise the fastest: the share rises from 21 to 39% of the total. Figure 6.1 might leave the impression that there is a strong correlation between the amount of income that a household receives and the amount of assets that it owns. This is generally true, as Table 6.2 illustrates, but the correlation is not as tight as one might have expected. There are households with large incomes and little assets, as well as households with large asset holdings and little income. In the middle quintiles, where differences between 27 quintiles are not that large, position by asset and income quintile varies much. Overall, the correlation between assets and income equals 0.47 in 2002 and 0.41 in 2004. 13 Table 6.2: Household income and household assets Income quintile Asset quintile 1 2 3 4 5 Total 1 10.77 5.09 2.50 1.25 0.40 20.00 2 5.33 6.13 4.76 2.77 1.02 20.00 3 2.16 5.13 5.51 5.02 2.18 20.00 4 0.90 2.70 4.89 6.38 5.13 20.00 5 0.41 1.02 2.46 4.69 11.41 20.00 Total 19.56 20.07 20.12 20.11 20.14 100.00 Note: cells show percentage of households relative to the total sample. Columns do not add to 20%, because asset values are missing for a small number of households. 6.2. Assets over time Assets are the foundation of the economy. Without assets, little production takes place. Without assets, it is difficult to smooth consumption from one period to another and thus to overcome shocks to the household or to the economy. Dynamic changes in assets, even over a short period of two years such as from 2002 to 2004, determine the future of the household and indeed of the economy. Thus, we turn our attention now to the change in assets over time, using the information from VHLSS 2002 and 2004. We desire to track the changes within a household, and thus we are restricted to the sample of 4055 panel households that are surveyed in both rounds. To compare the two rounds of the survey and avoid confounding effects of inflation, values are expressed in prices of 2002. 14 Table 6.3 shows where households find themselves in 2004, given their asset position in 2002. Thus, 63.6% of the poorest households in 2002 still find themselves in the lowest asset quintile in 2004, and 66.6% of the richest households are still among the richest in 2004. Nevertheless, the number of transitions to other cells is significant, especially in the middle quintiles. Table 6.3: Total household assets: transitions within the distribution 2004 Total asset quintile 1 2 3 4 5 1 63.59 22.61 10.05 2.02 1.74 2 23.57 40.39 21.74 10.20 4.10 2002 Total asset quintile 3 6.76 27.35 34.99 22.66 8.24 4 3.86 7.98 27.12 41.34 19.71 5 1.02 1.78 6.76 23.82 66.61 Note: cells report row percentages 13 Rank correlation coefficients are less sensitive to outliers. They indicate the correlation of the rank order in the distribution of income and assets. For 2002, the Spearman rank correlation coefficient equals 0.62, and for 2004 it equals 0.65. This does indicate a higher level of correlation in the income and wealth distribution of households. 14 An inflation rate factor of 1.11125 is used, based on the annual inflation rate of 6.3% for 2002 (counted as a half year), 3% for 2003 (counted as a full year), and 9.5% for 2004 (counted as a half year). 28 This type of analysis is also instructive for each type of assets separately. However, the concept of quintiles is not effective if there are too many households with zero-valued assets. For example, suppose that 50% of the households do not own a particular class of assets. An automatic division into quintiles would fill the lowest two quintiles and half of the middle quintile with those households, giving the suggestion that they are somehow different, whereas they are really all the same. Thus, if there are more than 20% of the households with a zero valued asset, we combine all of them in the bottom group and divide the remaining households that have positive values into the other four quintiles. This is emphasized in the Table 6.4 by renumbering the quintiles, starting them off with 0. Quintile 1 contains households with the smallest positive amounts of assets, and households in quintile 4 own the most. In terms of agricultural assets, on average, over period of 2002-2004 there are 35.2% of household that became relatively better off, 19.0% became worse off, and 46.8% remained in the same quintile group. However, it is quite different between groups. Around 45% households of 2002’s richest group (in terms of agricultural assets) became relatively worse-off, of which 24% lost one rank, i.e., went down from quintile 4 to quintile 3. Of those households without recorded agricultural assets in 2002 (quintile 0, which contains 30.5% of the households), 25.4% did record some in 2004, 3.09% of which ended up in the highest quintile (4). Among the households that were in the poorest quintile with agricultural assets (quintile 1), more than half did not record agricultural assets in 2004 and about a fourth reported relatively more assets. Ownership of business assets is not as widespread: quintile 0 contains 65.5% of the households. Accordingly, transitions out of the quintile 0 are less frequent (Table 6.4b). Surprisingly, a large number of households reported business assets in 2002 but not in 2004, and this is true in each quintile (1-4). According to Panel C of Table 6.4, transitions in the household consumer durables distribution are fairly substantial as well. There are therefore indications that measurement error in asset values may be substantial, which should not be unexpected. Nevertheless, the general trends are convincing. Table 6.4: Transitions in the distribution of assets, by asset type. A. Agricultural assets 2004 agro asset quintile 0 1 2 3 4 0 74.58 13.14 5.14 4.04 3.09 1 54.07 21.82 11.80 7.53 4.78 2002 agro asset quintile 2 16.62 28.61 28.87 17.36 8.55 3 9.74 11.09 31.14 32.98 15.04 4 8.44 4.44 8.06 24.05 55.01 B. Business assets 2004 business assets quintile 0 1 2 3 4 0 88.26 4.27 3.3 2.6 1.57 1 68.46 11.24 7.7 7.21 5.4 2002 business asset quintile 2 68.64 9.35 9.18 7.55 5.28 3 56.53 8.36 12.82 10.48 11.81 4 37.89 3.02 9.42 19.5 30.16 29 C. Household consumer durable goods 2004 consumer durables quintile 1 2 3 4 5 1 65.3 19.1 9.1 5.7 0.9 2 26.6 37.9 20.9 11.8 2.7 2002 consumer durables quintile 3 5.5 27.0 35.6 22.6 9.4 4 3.2 9.5 25.9 39.5 21.8 5 1.1 2.9 8.8 21.8 65.5 Note: Row percentages Tables 6.3 and 6.4 illustrate the relative position in the income distribution, but they are not able to illustrate the rise in assets in each household. The question arises whether the poor are able to catch up or whether the rich are now farther ahead. Table 6.5 sheds light on this question. Households are divided into terciles to reduce the detail of the information and yet show relevant patterns. Each major cell in the 3x3 grid contains a first row, showing the transition percentage, and a matrix of three columns of four rows, showing average asset values of four asset categories in 2002 and in 2004 as well as their ratio and thus the growth factor. Thus, 72.1% of the households that were ranked as poor in both 2002 were still poor in 2004. They owned 2824 thousand VND in total assets, 889 thousand VND of which were agricultural assets, 63 thousand VND business assets and 1871 consumer durables. These values rose to a total of 3574 thousand VND, for an increase of 27%. Of the separate components, agricultural assets rose 40%, the most of the three categories. Table 6.5: The Household Wealth Distribution in 2002-2004 Asset tercile 2004 1 2 3 72.13 22.91 4.96 % HH (Poor02-Poor04) (Poor02-Middle04) (Poor02-Wealthy04) 2002 2004 2004/02 2002 2004 2004/02 2002 2004 2004/02 1 Total assets 2824 3573 1.27 3868 14462 3.74 4290 63461 14.79 Agro assets 889 1248 1.40 1420 5306 3.74 1111 38884 35.00 Business assets 63 74 1.17 109 877 8.05 148 11120 75.14 Consumer durables 1871 2252 1.20 2339 8279 3.54 3031 13457 4.44 Asset tercile 2002 23.44 54.22 22.34 % HH (Middle02-Poor04) (Middle02-Middle04) (Middle02-Wealthy04) 2002 2004 2004/02 2002 2004 2004/02 2002 2004 2004/02 2 Total assets 11851 5615 0.47 14022 16063 1.15 15451 64129 4.15 Agro assets 4868 1811 0.37 4747 5527 1.16 5367 25756 4.80 Business assets 579 132 0.23 735 1086 1.48 800 4414 5.52 Consumer durables 6404 3673 0.57 8540 9450 1.11 9284 33959 3.66 3 3.58 23.40 73.02 % HH (Wealthy02-Poor04) (Wealthy02-Middle04) (Wealthy02-Wealthy04) 2002 2004 2004/02 2002 2004 2004/02 2002 2004 2004/02 Total assets 52050 4991 0.10 38593 19540 0.51 76598 94958 1.24 30 Agro assets 30282 884 0.03 14835 5213 0.35 28195 33992 1.21 Business assets 3547 221 0.06 3590 1158 0.32 10956 25088 2.29 Consumer durables 18221 3887 0.21 20168 13169 0.65 37446 35879 0.96 Notes: 2004 assets are valued in prices of 2002 (VND 1000). “2004/02” means ratio of 2004 to 2002 asset values. The middle class experience a rise in assets of 15%, and the wealthiest class saw its assets go up by 24%. Thus, in percentage terms, the poorest had the greatest increase. However, it is very notable that in the middle and wealthiest terciles business assets rose by far the fastest; consumer durables take a back seat, and agricultural assets do not rise particularly fast either. These trends also appear in the cells above the diagonal, which represent households that are relatively better off in 2004. On the other hand, below the diagonal each asset is reduced, but household consumer durables are reduced the least. As mentioned before, measurement error may be blamed for some of the transitions indicated in Table 6.3, but the trend is once again unmistakable: households that have made the greatest advances have invested in business assets. 6.3. Household income and changes in assets over time We now turn to the link between income and asset investment. This of course relates to the saving behavior of households: agricultural assets, business assets and consumer durables are part of the portfolio from which households choose in order to hold their savings. Table 6.6 starts this off by relating the real change in assets between 2002 and 2004 to total household income quintile position in 2002. With the exception of quintile 5, there is an increasing relationship between income and asset growth. From quintile 1 to 4, every income group accumulated more assets. For quintile 1 to 3, more than half of the investment goes into agricultural assets, but the share is dropping in quintile 4, and the average household in quintile 5 actually reduced its agricultural assets. Business assets are of little consequence in the lower quintiles, though a slight rise is evident, but households in income quintile 5 invest heavily in business assets, even to the point of reducing their consumer durables and, as mentioned, agricultural assets. Table 6.6: Total household income and asset changes by type Changes in household assets by income quintiles over 2002-2004 2002 income quintiles Total assets Agro assets Business assets Consumer durables 1 3403 2023 275 1105 2 5316 2953 1057 1305 3 7182 5361 968 853 4 13510 6292 1831 5387 5 9904 -2941 14486 -1642 Total 7904 2737 3764 1404 Note: Values are in thousands VND at 2002 prices. The relationship that is evident in Table 6.6 may be further refined by separately studying each income component. How do households invest out of their agricultural income? Do they use it to finance non-farm business operations, or do they plough savings back into agricultural 31 assets? Table 6.7a provides the evidence on this issue. Notice again the appearance of quintile 0, which consists of all households that received no agricultural income in 2002—but may have derived high incomes from other sources. The asset allocation of this quintile should therefore be only loosely compared with the higher quintiles that did receive farm income; our main emphasis should be on quintiles 1 to 4. Those with the smallest positive amount of agricultural income (quintile 1) actually invest more into business assets than those in quintiles 2 and 3 and less into agricultural assets. In quintile 4, we notice a dramatic increase in consumer durables and also a higher rate of investment in business assets. Investments in agricultural assets are roughly similar across quintiles 2, 3 and 4. Thus, one may infer that poor agricultural households invest in non-farm household enterprises in an attempt to escape poverty; rich agricultural households are drifting somewhat into non-farm entrepreneurship but mostly improve living conditions for the household. Table 6.7b performs the same analysis with respect to non-farm business income. The question of interest here is whether households might be using this income to finance investments in agricultural assets. If so, the business is of secondary interest to the household and may be abandoned as soon as farm objectives are achieved. For small-scale business owners (quintiles 1 and 2), this may indeed be the objective: investments in agricultural assets are larger than those in business assets or consumer durables. In quintile 3, there is a greater emphasis on business assets, and households in quintile 4 invest heavily into business assets, reducing holdings of agricultural assets. The emphasis of entrepreneurship is at the cost of consumer durables, in strong contrast with the allocation of agricultural income. Finally, Table 6.7c considers the impact of wage income on asset investment. Total assets do not rise much by wage income quintile. Higher wage incomes are apparently mostly spent on day-to-day consumption. Households in quintiles 1, 2 and 3 build more agricultural than business assets, but in quintile 4 investments in business assets once again dominate, indicating that some high wage earners are considering transitions into private entrepreneurship. Table 6.7: Household income by component and asset change by type A: Agriculture income Changes in household assets by agricultural income quintiles over 2002-2004 2002 agricultural income quintiles Total assets Agro assets Business assets Consumer durables 0 11089 1197 10906 -1014 1 4605 1518 2727 360 2 3908 3336 -203 775 3 5440 3950 635 856 4 13933 3760 3921 6252 Total 7889 2719 3772 1398 B: Business income Changes in household assets by business income quintiles over 2002-2004 2002 business income quintiles Total assets Agro assets Business assets Consumer durables 0 7462 4084 758 2619 1 4464 2762 678 1023 2 4551 2871 1752 -72 32 3 5398 2756 3637 -995 4 20060 -6359 28836 -2416 Total 7904 2737 3764 1404 C: Wage income Changes in household assets by wage income quintiles over 2002-2004 2002 wage income quintiles Total assets Agro assets Business assets Consumer durables 0 10291 3039 4257 2995 1 6320 3427 2195 698 2 6543 3160 1874 1509 3 5436 3780 1289 366 4 6144 -273 8129 -1713 Total 7904 2737 3764 1404 In the final tabulation of this description of asset change, Table 6.8 compares asset investment by region across income quintiles. As regional/quintile subsamples become smaller, patterns become a little more difficult to discern. For example, the amount of asset growth in the North East (quintile 4) and South East (quintile 3), and the lack of it in the South Central Coast (quintile 2) and North Central Coast (quintile 4 and 5), violate the general trend that is evident in the total. Even so, investments in the Central Highland and South East regions are higher and more widespread than elsewhere. Another way to view patterns of investment is gained from inspecting the average and median of the rate of change in each household. In Vietnam, the average rate of change in household assets is 150% (Table 6.8b), but the median is a much more modest 11%. With measurement error possibly creating outliers in the average rate of change, the median indicates the more typical experience of households in Vietnam. Here, it appears that the typical household in quintiles 1 and 2 accumulates assets more rapidly than those in higher income quintiles, though of course generally from a lower base. In higher income quintiles, with the exception of the Central Highlands, median households saw a slight decrease in their assets. Thus, the increase in the general average asset holdings among higher quintile households is driven by large increases among a relatively small number of households. Table 6.8: Change in household assets over 2002-2004 by region & income quintiles A. Average change in value 2002 income quintiles Region Total 1 2 3 4 5 Red River Delta 1754 5583 7182 8916 19256 7826 North East 2680 4331 2575 40871 4295 11827 North West 3045 1686 3153 1687 10261 3601 North Central Coast 2359 1686 1555 -1858 -177 1160 South Central Coast 3446 -573 3339 2489 -9486 512 Central Highland 4668 14564 16657 15591 41281 16814 South East 6701 15048 24218 16144 10860 14454 Mekong River Delta 7086 6547 4826 11855 7819 7921 Total 3403 5316 7182 13510 9904 7904 33 B. Average and median rate of change Average rate of change in assets Median rate of change in assets Region 2002 income quintile 2002 income quintile Total Total 1 2 3 4 5 1 2 3 4 5 RRD 147 70 109 77 79 99 7 18 28 3 -10 9 NE 206 181 97 239 26 158 39 34 14 -8 -11 11 NW 434 50 105 48 214 212 60 24 9 -8 25 24 NCC 125 100 47 13 20 77 28 10 -12 -7 -11 6 SCC 273 139 95 50 64 131 29 18 22 -7 -13 2 CH 133 1185 222 81 77 391 24 90 40 23 30 39 SE 439 516 673 152 85 269 6 90 60 2 -5 6 MRD 309 162 95 71 74 124 55 29 15 0 3 13 Total 212 210 161 98 72 150 27 23 18 -1 -6 11 Note: Values in thousands VND in 2002 prices. 7. Investment climate 7.1. MSEFI: An index of micro/small entrepreneur opinions about the business climate The 2004 questionnaire elicits opinions of household enterprise operators about various aspects of the business climate. The questionnaire lists 17 different areas that might impact the management and operation of non-farm enterprises: electricity, communication and postal services, transportation, land uses, waste water and solid disposal treatment, financial access, finance cost, taxes, business registration and operation license, labor regulations, skill and Figure 7.1: Hurdles to the operation and development of household enterprises Electricity Unfair competition 0.6 Communication, postal service Crimes, theft, lack of security Transportation, roads 0.4 Corruption Land for production 0.2 Unstable macro economy Waste water/solid disposal 0 Inconsistent economic policy Financial access Commercial/customs regulation Financial expenditures Skill and education Taxes Labor regulation Registration/license Source: VHLSS 2004. Scores are averages across enterprises on responses ranging from 0 (no hurdle) to 4 (serious hurdle). 34 education level of workers, commercial and custom regulations, inconstant economic policies, instability of the macro economy, corruption, crimes, thefts and lack of security, and unfair, unhealthy competition. Entrepreneurs evaluated each component on a scale of 0 (no hurdle) to 4 (serious hurdle). They actually had the option to list “irrelevant” of “do not know” as an option. If they responded “irrelevant”, this is treated as equivalent to a scale score of 0 (“no hurdle”), but a response of “do not know” is in essence a missing response. Figure 7.1 indicates the average scores on each hurdle. Note that the scores run from 0 to 4, with higher scores indicating higher hurdles. The most complaints were addressed at the topic of unfair and unhealthy competition. Transportation and roads, and crimes, theft and lack of security were other hurdles that were frequently mentioned. On the other hand, labor, commercial and customs regulation received few complaints, and corruption was not mentioned often either. There are some differences between urban and rural communities (Figure 7.2). The three major complaints in both urban and rural communities are unfair and unhealthy competition; crime, theft and lack of security; and transportation and roads. But rural entrepreneurs have more complaints than urban colleagues with regard to transportation and roads, skill and education of the workforce, and electricity. Urban entrepreneurs worry more about unfair competition, crimes, land, financial access, taxes and an unstable macro economy. In a background paper to this report, Vu (2006) documents differences in reported hurdles between industrial sectors. As one might expect, transport enterprises complain frequently (score=1.23) about transportation and road hurdles, but they also mention theft and corruption and unfair competition more often. Mining enterprises mention land and Figure 7.2: Urban/rural differences in hurdles to the operation and development of household enterprises Unfair competition Crimes, theft, lack of security Transportation, roads Land for production Financial access Taxes Unstable macro economy Inconsistent economic policy Urban Skill and education Rural Electricity Registration/license Financial expenditures Corruption Waste water/solid disposal Communication, postal service Commercial/customs Labor regulation 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 35 transportation issues more frequently. Entrepreneurs in the construction business complain about theft, lack of labor skill, labor regulation, access to and the cost of finance, and transportation. It would be interesting to use these responses about hurdles to household enterprises as indicators of the local business climate. Unfortunately, the VHLSS sample is not representative at the local or district level, but it is representative at the provincial level. Thus we aggregate the responses to the province level by first computing the average hurdle across all enterprises and items, and then rescaling the average response from 0 to 100, where 0 represents the highest hurdle average and 100 represents the lowest average. The resulting score is referred to as the Micro/Small Enterprise Friendliness Index or MSEFI. 15 Thus, a high value reflects an environment with few hurdles. Table 7.1 divides the provinces into five groups according to the MSEFI; provinces are listed in alphabetical order. Detailed scores are tabled in the Appendix. The province with the fewest complaints is Binh Duong, with Hoa Binh a close second. Ho Chi Minh City ranks 28th, Da Nang comes in at 36th, Ha Noi ranks 58th, and the last place is taken by Can Tho. Table 7.1: Level of provincial performance on the basis of MSEFI Performance group and Number of Provinces (in alphabetical order) MSEFI range provinces An Giang, Ben Tre, Binh Duong, Dak Nong, Dien Bien, Hoa Binh, Lai High: 80 – 100 11 Chau, Ninh Thuan, Quang Nam, Quang Ninh, Tay Ninh Ba Ria-Vung Tau, Binh Dinh, Binh Phuoc, Binh Thuan, Ca Mau, Da Nang, Dak Lak, Dong Thap, Gia Lai, Ha Giang, Ha Tinh, Hai Phong, Mid-high: 60 – 80 27 Ho Chi Minh City, Khanh Hoa, Kon Tum, Lam Dong, Long An, Nam Dinh, Nghe An, Ninh Binh, Phu Yen, Soc Trang, Son La, Thua Thien – Hue, Tra Vinh, Tuyen Quang, Vinh Long Bac Giang, Bac Lieu, Bac Ninh, Dong Nai, Hai Duong, Hau Giang, Middle: 40 – 60 17 Lang Son, Lao Cai, Phu Tho, Quang Binh, Quang Ngai, Quang Tri, Thai Binh, Thai Nguyen, Thanh Hoa, Tien Giang, Yen Bai Mid-low: 20 – 40 6 Bac Kan, Cao Bang, Ha Nam, Ha Noi, Ha Tay, Kien Giang Low: 0 – 20 3 Can Tho, Hung Yen, Vinh Phuc In the way described above, the MSEFI is an aggregate measure of the business climate. It is of course feasible also to calculate a MSEFI sub-index by province for groups of influential factors that cover infrastructure, labor, finance, policy, registration, and corruption and security groups. 16 This allows one to more clearly pinpoint the reasons for a low ranking in the aggregate 15 The background paper to this section (Vu, 2006) designs a different Business Friendliness Index, for which the lowest scores are the most desirable. In that study, the aggregate score is a weighted sum of the sub-indices that are similar to those discussed below. 16 The MSEFI of infrastructure was created based on performance in regard to electricity, communication and postal services, transportation, land, and waste treatment. The MSEFI of finance was computed on the basis of financial access and expenditure. The MSEFI of labor represents labor regulation and the skill and education level of workers. The MSEFI of policy summarizes commercial and customs regulations, inconsistent economic policies, and instability of the macro economy. The MSEFI of registration refers to business registration and operation license. Finally, the MSEFI of corruption and security represented for corruption, crime, theft and lack of security, and 36 MSEFI, even if the aggregate index is not a weighted average of the sub-indices. Figure 7.3 shows the aggregate and component scores for three rather arbitrarily selected provinces. In Ha Noi province, the aggregate score of 38.3 comes from low scores on infrastructure (I), registration (R), policy (P) and corruption and security (C). Labor (L) and finance (F) are not the best but, according to the VHLSS entrepreneurs, are not the most urgent priorities. Scores are higher in every aspect in Ho Chi Minh City; finance, registration, and policy draw somewhat lower scores. Son La is a province where some business climate factors are nearly the best in Vietnam (registration and finance), and other factors could use substantial improvement (labor and policy). Table 7.2 summarizes MSEFI scores by region. The Northwest region scores the highest in the aggregate and, interestingly, also in every component. The Northeast and Red River Delta regions achieve the lowest MSEFI values, again both in the aggregate and in every component. Figure 7.3: MSEFI in three provinces MSEFI MSEFI MSEFI 100 100 100 C I C I C I 0 0 0 P F P F P F L R L R L R a. Ha Noi b. Ho Chi Minh City c. Son La Table 7.2: MSEFI and component scores across regions, VHLSS 2004 Region MSEFI Infrastructure Finance Registration Labor Policy Corruption Red River Delta 47.27 48.75 61.28 53.76 51.03 42.40 64.08 Northeast 52.00 59.34 55.11 67.19 58.31 45.90 65.56 Northwest 84.67 84.11 92.91 90.15 83.90 83.62 84.87 North Central Coast 61.13 66.02 73.37 74.75 64.69 45.18 74.01 South Central Coast 69.93 67.97 79.27 71.51 74.02 75.59 76.88 Central Highlands 70.41 83.43 69.15 68.08 72.45 61.51 75.07 Southeast 74.67 79.80 74.33 71.18 76.37 72.54 81.92 Mekong River Delta 62.76 71.68 68.75 72.57 71.25 70.66 68.22 Total 68.51 73.96 74.17 73.37 72.54 67.36 75.25 In 2005, the Vietnam Chamber of Commerce and Industry launched the Provincial Competitiveness Index (PCI), which aimed to rank provinces by competitiveness on the basis of information gathered in 2002-2004 in a study led by Dr. Edmund Molesky. This index is based on responses by larger-scale private sector entrepreneurs and covers 42 provinces. A scatterplot unfair and unhealthy competition. Because each component index is scaled in its own way, the aggregate is not a weighted average of the component indices. 37 (Figure 7.4) indicates a low level of correlation. Indeed, the correlation coefficient over the provinces for which both indices are available equals only 0.029. While the PCI index does incorporate information other than entrepreneur opinions of the investment climate, if both samples of entrepreneurs are representative of the private sector, one would have expected a larger positive correlation value. Apparently, micro- and small-scale entrepreneurs have a different view of the business climate than medium- and large-scale entrepreneurs. Private sector policy does not impact all of them equally. Thus, policymakers need to realize that their policies are implicitly targeted (and perhaps unintentionally so). Policies may need to be broader, encompassing the whole spectrum of private sector entrepreneurship, or they may need to focus more on subgroups. For example, different finance policies may be needed for medium/large-scale entrepreneurs than for micro/small-scale entrepreneurs. Figure 7.4: Correlating MSEFI and PCI as alternative business climate summaries 80 70 60 PCI 50 40 r = 0.029 30 0 20 40 60 80 100 MSEFI 7.2. Impact on Economic Performance Having developed these measures of business climate opinion, we must of course ask how they impact enterprise performance and household choices. Theoretically, it is actually not clear in which direction the influence might go. When micro and small-scale entrepreneurs experience fewer hurdles, one would expect that more households would start a business and that enterprises would be more successful. Certainly, when transaction costs decrease, productivity increases, leading to rising enterprise incomes. But in a competitive environment, economic (or excess) profits are eroded by firm entry, even if enterprise incomes rise because of increasing productivity. Thus, there may be more enterprises but the average performance may still not improve dramatically. At the same time, a better economic climate may stimulate growth in firms. Those with the best entrepreneurial talent will benefit the most and actually crowd out 38 other firms, both in the output market and by “buying out” the lagging entrepreneurs by offering them a job. Entrepreneurship becomes more concentrated among fewer households, and wage employment expands. In other words, improving economic conditions generate conflicting forces that prevent us from predicting how household choices will change. Quite likely, the net outcome of these forces depends on the stage of economic development of the country. The impact of MSEFI is therefore an empirical matter, one that describes Vietnam in 2004 but that may be different at a future time or in another country. Table 7.3 puts together the most important outcomes of household behavior. The first pair of columns concerns households engaging in farm and non-farm enterprise activity. The second triplet of columns looks at their income. The third and fourth pair of columns examine their assets in 2004 and, for panel households, the change in assets since 2002, respectively. The fifth pair of columns counts the number of workers in the farm and in the non-farm enterprise. The sixth pair of columns looks at capital intensity, defined as the ratio of assets over number of household workers. In Panel A, we find the average for all these variables in each MSEFI group. Thus, for example, 81% of the households living in a low-MSEFI province have members working on a farm; they receive 6.62 million VND worth of farm income per year, 6.19 million VND non-farm business income, and 6.31 million VND wage earnings. In this panel, we also find the correlation coefficient that measures the association between MSEFI and each of these variables. (With roughly 9000 households involved in the calculation of the correlation coefficient, any correlation coefficient greater than 0.021 or less than –0.021 is statistically significant.) Generally the associations are rather weak. In provinces with higher MSEFI scores, fewer households engage in farming, and the number of agricultural household workers seems to decrease. The number of households with an enterprise seems rather constant, while the average number of household workers in the enterprise appears to diminish. Farm incomes rise, but the trends with respect to enterprise income and wage earnings is unclear. Farm assets, farm asset investment, and farm capital intensity all seem to increase, but in provinces with a high MSEFI households own fewer business assets. But few of these descriptions represent monotonic trends, and the correlations, while conforming to these characterizations, are quite low. One reason why correlation coefficients are low is because the economic performance variable (such as business income) may be zero for many households. For that reason, the table also lists a correlation coefficient that is restricted to the subsample with nonzero values on the economic performance variable, designated as “correlation+”. The correlations are generally higher. Table 7.3: Association between MSEFI and farm and non-farm economic performance Number of Household Change in household Asset intensity activitya Incomeb Assetsb assetsb workers (assets/labor)b Farm NFHE Farm NFHE Wage Farm NFHE Farm NFHE Farm NFHE Farm NFHE A. MSEFI Low 0.81 0.43 6.62 6.19 6.31 9.70 5.75 4.61 2.48 1.75 0.73 6.72 7.67 Mid-low 0.66 0.44 5.41 5.65 11.32 7.55 8.65 0.54 0.82 1.44 0.69 6.78 13.19 Middle 0.85 0.39 7.91 4.97 5.79 11.95 8.53 3.82 2.85 1.88 0.62 7.03 12.51 Mid-high 0.67 0.41 7.14 6.69 9.65 16.73 12.75 1.69 7.58 1.46 0.62 14.98 19.38 39 Number of Household Change in household Asset intensity activitya Incomeb Assetsb assetsb workers (assets/labor)b Farm NFHE Farm NFHE Wage Farm NFHE Farm NFHE Farm NFHE Farm NFHE High 0.74 0.39 8.16 5.37 7.95 18.14 3.95 8.25 1.42 1.50 0.59 13.54 5.98 Total 0.73 0.40 7.27 5.90 8.39 14.23 9.74 3.25 4.51 1.59 0.63 10.99 14.65 Correlation -0.062 -0.016 0.029 0.010 0.023 0.046 0.006 0.003 0.014 -0.075 -0.047 0.037 0.028 Correlation+ 0.060 0.030 0.023 0.080 0.020 0.005 0.022 -0.040 -0.028 0.063 0.018 Without Ha Noi Mid-low 0.83 0.44 7.20 4.84 6.11 8.04 7.28 0.84 -0.39 1.87 0.74 4.56 10.46 Correlation -0.113 -0.014 0.008 0.014 0.073 0.042 0.007 0.001 0.015 -0.087 -0.057 0.048 0.031 B. Correlation with MSEFI components Full sample Infrastructure -0.077 -0.013 0.026 0.017 0.037 0.049 0.006 0.001 0.016 -0.070 -0.053 0.040 -0.002 Finance 0.026 -0.011 0.025 -0.039 -0.043 0.000 -0.012 -0.011 0.000 -0.023 -0.021 -0.014 0.004 Regulation 0.021 -0.028 0.035 -0.018 -0.056 -0.017 -0.001 -0.006 -0.004 -0.028 -0.047 -0.007 0.011 Labor -0.182 0.007 -0.015 0.062 0.115 0.028 0.021 0.009 0.008 -0.098 -0.063 0.049 0.040 Policy -0.115 0.004 0.029 0.037 0.039 0.034 0.018 -0.004 0.021 -0.102 -0.035 0.039 0.036 Corruption -0.024 -0.015 0.025 0.006 0.004 0.057 0.004 0.012 0.013 -0.062 -0.025 0.038 0.038 Without Ha Noi Infrastructure -0.118 -0.012 0.009 0.021 0.078 0.046 0.007 -0.001 0.016 -0.078 -0.060 0.049 0.001 Finance -0.005 -0.010 0.010 -0.037 -0.013 -0.005 -0.011 -0.012 0.000 -0.031 -0.029 -0.009 0.006 Regulation -0.044 -0.027 0.006 -0.014 0.006 -0.028 0.000 -0.009 -0.005 -0.044 -0.062 0.004 0.016 Labor -0.178 0.007 -0.010 0.062 0.111 0.031 0.021 0.010 0.008 -0.094 -0.060 0.048 0.039 Policy -0.162 0.006 0.011 0.042 0.084 0.030 0.019 -0.006 0.021 -0.112 -0.043 0.048 0.039 Corruption -0.088 -0.013 -0.002 0.012 0.069 0.053 0.005 0.010 0.013 -0.079 -0.040 0.053 0.043 C. PCI Low 0.83 0.43 7.12 4.28 5.79 15.07 5.27 7.73 4.39 1.77 0.70 10.66 7.16 Mid-low 0.81 0.39 7.38 3.84 5.18 8.38 3.45 1.72 0.04 1.67 0.59 5.68 6.21 Middle 0.77 0.40 8.61 6.16 6.99 15.80 14.87 6.65 2.49 1.64 0.65 11.20 23.07 Mid-high 0.58 0.45 4.67 8.22 12.88 9.34 14.36 -3.03 9.59 1.20 0.70 10.83 19.14 High 0.64 0.42 7.76 7.13 10.88 18.38 6.95 9.71 1.79 1.30 0.67 16.06 9.93 Missingc 0.87 0.32 10.09 4.05 5.06 21.72 8.03 4.61 3.56 2.12 0.48 12.49 14.74 Total 0.73 0.40 7.27 5.90 8.39 14.23 9.74 3.25 4.51 1.59 0.63 10.99 14.65 Correlation -0.189 0.014 -0.030 0.087 0.156 0.001 0.028 -0.015 0.013 -0.062 -0.026 0.028 0.018 Notes: a Proportion of household engaged in the stated activity b Values in millions VND c PCI was measured in 42 of the 64 provinces. The missing category represent the remaining 22 provinces. + Correlation among households with nonzero values of the stated variable Another reason for unsettled patterns and low correlation coefficients is the presence of outliers. The table illustrates the impact of one of them: at first blush, the province of Ha Noi does not seem typical among the other provinces in the mid-low MSEFI group, because it represents a higher-income, more urbanized environment. Omitting Ha Noi indeed creates a more transparent pattern and correlation: higher MSEFI scores tend to reduce the size of the farm sector, lower the number of workers involved in household enterprises, raise the capital intensity of both farm and non-farm enterprise production, and increase wage incomes. As hypothesized at the beginning of this section, these patterns verify the process of transition towards a more concentrated form of entrepreneurship with fewer but larger enterprises. 40 A study of the MSEFI components points out forces that bring this transition about. Or, more accurately, it points out hurdles that prevent this transition from happening, since MSEFI describes hurdles that entrepreneurs encounter. The most potent indicator appears to the labor component of MSEFI. Higher labor scores (associated with fewer labor regulations and better- skilled workers) reduce the size of the farm sector, raise wage earnings, reduce the number of non-farm self-employed workers, and increase capital intensity. Infrastructure impacts most of these same variables. MSEFI’s policy component (associated with commercial and customs regulation, inconsistent policies, and an unstable macro economy) also leads to a smaller farm sector and greater capital intensity, but at the same time is associated with higher incomes in every sector, including agriculture. Better scores on corruption and security (including unfair competitive practices) are associated with higher wage earnings, higher levels of farm assets, a smaller farm sector, and greater capital intensity. The MSEFI components of finance and regulation have lesser impacts; if anything, provinces with higher scores on these components have lower business incomes and wage earnings. Altogether, one might notice that several of these components impact the household economic performance indicators in the same way: since the correlation among the components ranges from 0.46 to 0.67, it is actually not clear precisely which type of hurdle prevents the patterns indicated by Panel B of Table 7.3. Multivariate regression analysis (in Section 10) is necessary, but may yet not yield an answer because of the correlation among the components. Panel C of Table 7.3 relates the household behavioral outcomes to the PCI index that was proposed by the VCCI (2005) study. The patterns among the grouped households (low to high, omitting those with a missing PCI score) and the correlations indicate that households in higher- PCI provinces are less likely to engage in farming, have fewer members working on the farm, and earn less income from farming. They earn higher incomes from non-farm enterprises and from wage earnings, but have fewer workers active in a household enterprise. Evidence on asset patterns and capital intensity is too weak to interpret. Given the lack of correlation between MSEFI and PCI, these results are almost surprisingly similar to the inferences drawn from the impact of MSEFI. Apparently, even though factors measured by PCI do not register among the entrepreneur responses about hurdles, they do determine the economic environment to which households respond. Yet, MSEFI has its own informational content, which appears to matter as well. 7.3. Community information The MSEFI measures aspects of the investment climate from the micro- and small-scale entrepreneur’s point of view. The VHLSS community questionnaire also contains a wealth of information on the business (or investment) climate in the communities where the household resides and the enterprise is active. This section looks at this information, comparing it between 2002 and 2004 in order to get a sense about the changes occurring in the communities. Recall in this comparison that the 2004 survey gathered information only in rural communes, whereas the 2002 community survey visited urban communities (wards) also. First of all, we examine population size and the number of household as a proxy for market size. As shown in Table 7.4, the population size of the surveyed communes in 2004 is smaller than that in rural communes in 2002, both in terms of total population size and the number of households. The 2002 urban communities are clearly larger than the rural communes. 41 This is one reason why we must separate urban from rural communities in this study of investment climate variables. Our discussion will focus primarily on the comparison of 2002 and 2004 rural communes, however; whenever 2002 urban statistics are available, they are provided merely for comparison with 2002 rural communes. The VHLSS 2004 collects information on migration into or out of the community. As Table 7.4 shows, in slightly more than half of the rural communes, more people are moving out than moving in, but at the same time there is net in-migration in more than one third of the communes as well. Overall, the average net migration is about –3 persons per rural commune per year. All of this implies that people are more likely to move out of the rural community but the difference is very small. Table 7.4: Community size 2004 rural 2002 rural 2002 urban Commune population (number of persons) 8599** 8995 14159+++ (4784) (5992) (8320) Number of households 1889** 1961 3102+++ (989) (1092) (1633) Migration (%) More people moving out than moving in 56.4 Zero net migration 5.9 More people moving in than moving out 37.7 Note: Standard deviations are in parentheses ** Different from the 2002 rural sample at the 5 percent significance level +++ Different from the 2002 rural sample at the 1 percent significance level Source: Calculation based on VHLSS 2002 and 2004 The percentage of communes where respondents thought that the living standards in their community were better than 5 years ago increases slightly from 97.69% in 2002 to 98.60% in 2004 (Hoang, 2006). Similarly, a majority reported in 2004 that agricultural yields and productivity had improved during the 5 years prior. Even so, communes have to cope with difficult circumstances occasionally. The average commune experiences 1.84 natural disasters (fire, epidemic, flood, typhoon, drought, pest disease, or other), which affect large shares of their households (Hoang, 2006). These occurrences may have an impact on the prosperity of non- farm enterprises, either because of direct damages suffered or because of reduced demand for their products. Table 7.5 considers changes in the agricultural and non-farm sector as reported at the community level. Rural communes faced a slight decrease in land quality, a reduced agricultural extension service level (as measured by distance to the extension service center and the number of visits during the year) and higher wages. Improved agricultural conditions may reduce the attractiveness of non-farm business activities, but at the same time demand for non-farm products and services are positively impacted by income levels in the community and thus also by the prosperity of the farm sector. Community respondents reported a sharp increase in the number of enterprises between 2002 and 2004. There are also small increases in the reported presence of a traditional business 42 and the existence of a production or service base or a craft village. In other words, communes are moving in the direction of non-farm productive activities. Table 7.5: Conditions in the agricultural and non-farm sector, VHLSS 2002 - VHLSS 2004 2004 rural 2002 rural 2002 urban Agriculture Agricultural land quality index 3.60*** 3.79 3.63 (1: lowest, …, 6: highest quality) (1.45) (1.22) (1.29) Agricultural extension index 0.57*** 1.01 1.05+++ (0.81) (0.37) (0.30) *** Agricultural wage (000dong/person/day) 23.81 19.16 21.11+++ (5.79) (5.12) (5.46) Non-farm sector Enterprise density per 1000 persons 3.16*** 0.22 (11.97) (0.51) *** Existence of a traditional business 0.122 0.095 (0.327) (0.293) Existence of production/service base or craft village 0.663 0.643 (0.473) (0.479) Note: Standard deviations are in parentheses *** Different from the 2002 rural sample at the 1 percent significance level +++ Different from the 2002 rural sample at the 1 percent significance level Source: Calculation based on VHLSS 2002 and 2004 Table 7.6 summarizes a large number of investment climate variables that were measured at the community level. The table lists a number of ‘proximity indices’. These indices range from 0 to 1, with 1 indicating a presence right within the commune and 0 indicating that the object is a long distance (usually at least 10 kilometers) away from the commune. Thus, a proximity index of 0.70 implies that the object is 3 kilometers away from the commune. For more detail on the definition of these and other community variables, see Appendix A.2. The first group of variables addresses infrastructure. The provision of electricity is pretty much universal in 2004, improving six percentage points relative to 2002 to 98%. The quality of water sources also improved, but there is obviously still much room for improvement. Most households have access to a cultural post office or live near a radio relay station. The second group of variables concerns communication. Access to the post office and to telephone services is quite good in most communes. The telephone information was omitted from the 2004 questionnaire. The third set of variables looks at transportation, both through modes and to various destinations. The road index (measuring proximity and yearround accessibility) improved, as a reflection of infrastructure investment projects that are reportedly taking place in rural areas. Some communes without good road access do have access to waterways. Many communes are within a couple of kilometers of market places, daily or periodic, or are located within a reasonable distance to a small or even a larger town. Access to those locations obviously depends on the quality of the transportation network. Average travel time to a daily market was reportedly 18 minutes; to a periodic market 22 minutes; to a wholesale market 30 minutes; to the nearest small town 34 minutes (Hoang, 2006). Note that improved connection with far-away places opens up opportunities for entrepreneurship: the potential clientele expands. At the same 43 time, competition may also rise, as remote producers come to offer their commodities in the local community. Thus, the net effect of these transportation variables is generally uncertain. Another consideration, of relevance to road, water, and public transportation indices, is that improved access may open up employment opportunities elsewhere and thus reduce the desire for local entrepreneurship: thus, improved access may raise the standard of living but not the level of entrepreneurship. Access to education has also improved significantly, which is again a reflection of large investments in education in rural areas. (Note that the proximity index uses a cut-off distance of 25 kilometers.) Only upper secondary schools are not widely available yet within a distance of 25 kilometers. Community respondents do report a slight increase in the number of problems at schools, such as inadequate supplies, low pay, limited budget, and so forth. This could be an indication of a lower quality of education, but one may also wonder if this could be a reflection of an understanding on the part of community leaders that only those with complaints receive attention from bureaucratic agencies higher up in the budgeting line of command. The last set of variables concerns health care. Somewhat surprisingly, the proximity index of health care facilities went down between 2002 and 2004. There is a chance in the list of health care providers between the two questionnaires, such that the proximity indices may not be comparable between the two years. Health care quality remained the same. Table 7.6: Investment climate, VHLSS 2002-VHLSS 2004 Investment Climate Topic 2004 rural 2002 rural 2002 urban Infrastructure Availability of electricity 0.981*** 0.924 0.997+++ (0.138) (0.266) (0.054) Water source in dry seasona 2.75*** 1.89 3.61+++ (1.09) (0.82) (1.45) Water source in dry seasona 3.12*** 1.78 3.53+++ (1.00) (0.80) (1.48) Availability: cultural post office 0.82 (0.38) Availability: cultural house 0.30 (0.46) Availability: radio relay station 0.77 (0.42) Communication Proximity index: post office 0.79*** 0.86 (0.27) (0.26) Proximity index: telephone 0.89 (0.24) Transportation Road index 0.91*** 0.84 (0.22) (0.31) Waterway index 0.19 (0.39) Public transport index 5.46 5.09 (16.51) (9.57) Presence of communal / intercommunal market 0.62 (0.49) 44 Proximity index: wholesale market 0.26 (0.36) Proximity index: periodic market 0.40*** 0.71 (0.41) (0.35) Proximity index: daily market 0.74*** 0.77 (0.31) (0.31) Proximity index: district town N/A 0.40 (0.39) Proximity index: small town 0.69 (0.30) Proximity index: major townb 0.32 (0.32) Proximity index: large town 0.30 (0.33) Education Proximity index: primary school 0.97*** 0.91 (0.07) (0.14) Proximity index: lower secondary school 0.91*** 0.77 (0.18) (0.30) Proximity index: upper secondary school 0.55*** 0.30 (0.40) (0.35) # Problems at primary school 1.69*** 1.51 (0.55) (0.83) # Problems at lower secondary school 1.61*** 1.37 (0.63) (0.88) # Problems at upper secondary school 1.12*** 0.76 (0.88) (0.93) Health Proximity index: health facilities 0.71*** 0.81 (0.18) (0.16) # Problems at health facilities 2.21 2.20 (0.74) (0.76) Note: a 0: other source, 1: river/pond, 2: drilled well, 3: dug well, 4: rain water, 5: piped well b The questionnaire makes specific reference to Ha Noi, Haiphong, and Ho Chi Minh City as examples. Standard deviations are in parentheses *** Different from the 2002 rural sample at the 1 percent significance level +++ Different from the 2002 rural sample at the 1 percent significance level Source: Calculation based on VHLSS 2002 and 2004 It is worthwhile to examine how these investment climate variables impact the performance of the household enterprises and indeed the entrepreneurship decision itself. However, rather than attempting a descriptive analysis similar to that in other sections (e.g., Section 7.2 above or Section 8 below), we will hold off with this exercise until the multivariate regression analysis in Section 10, as different aspects of the investment climate are likely correlated, which might yield spurious correlations with indicators of enterprise performance and entrepreneurship selection. 45 8. Household enterprise networking strategies This section examines several topics that seem peripheral to a study such as this about enterprise performance and entrepreneurship selection but that may be crucial to the success of entrepreneurial ventures. Networking strategies are essential to the production process, just as the proper use of technology is. For example, it determines the transaction cost of input procurement and of marketing; it may reduce the cost of working capital; and it impacts the selection of technology. Driven by availability in the VHLSS 2004 data, we consider the following topics. (i) How deeply are enterprises involved in business association or clubs? (ii) How frequently are they contacted by functional (or government) agencies, and for what purpose is that contact established? (iii) How do entrepreneurs gather information for pricing their products? (iv) Do they rely on inputs supplies and customers for the transport of raw materials and final products, or do they arrange for transport themselves? In this regard, does the enterprise sell to customers in other communities? 8.1. Participation in business associations or clubs Business associations or clubs (BAC) for enterprise promotion play a more and more important role in production and business activities. However the number of enterprises that participated in business association or clubs in Vietnam is small. According to the VLSS 2004, only 0.5% of the enterprises participate in business associations or clubs in the whole country (Table 8.1). Among participating enterprises, there are only 8.3% involved in VCCI, 67.1% in another association or club, and 31.6% in unofficial groups of businesses or households. For an enterprise to be involved in more than one association or club is rare (7%). The percentage of urban enterprises (0.86%) participating in BAC is a little higher than that of rural ones (0.36%). All of the VCCI members are urban. Among economic regions, there is not any enterprise in the North West and the Central Highland regions involved in a business association or club. The Red River Delta region has highest percentage of enterprises participating in business associations with 0.74% of the surveyed enterprises, followed by the South Central Coast and the Mekong River Delta (0.69%). The type of membership varies by region. For example, the Mekong River Delta is the only region where there are any enterprises that are member of VCCI. In the North Central Coast region, participating enterprises are involved only in other associations or clubs, while in the South East, they are members only of unofficial group of businesses or households. Among industry sectors, the only enterprises that participate in BAC are in manufacturing and construction, in trading and in services. The percentage of service enterprises involving in BAC reach 1.7% of its enterprises, following by manufacture and construction at 0.57%, and trading at 0.18%. None in the service sector is a member of VCCI; more than 78% of the participating enterprises are members of other associations or clubs; and about 34% are members of unofficial group of businesses or households. In the manufacturing and construction sector, about 14% of BAC participating enterprises are involved in VCCI, 54% in other associations or clubs, and 32% in unofficial group of businesses or households; food and beverage enterprises are involved only in VCCI and other associations or clubs, wood processing and production enterprises are involved in other associations or clubs, and other manufacturing enterprises are involved in other associations or clubs and unofficial groups of businesses or 46 households. In the trading sector, members of VCCI are 25% of participating enterprises; members of other associations or clubs 53%; and members of unofficial group of businesses or households 21%. Table 8.1: Membership in business associations and clubs Among participating enterprises, participate with at least one BAC Vietnam Chamber Unofficial group of of Commerce and Other clubs or businesses or Member of several % Participating Industry associations households organizations Total 0.50 8.31 67.07 31.63 7.02 Urban/rural residence Urban 0.86 14.03 67.35 30.46 11.84 Rural 0.31 0.00 66.67 33.33 0.00 Region Red River Delta 0.74 0.00 75.72 24.28 0.00 North East 0.52 0.00 63.98 70.55 34.54 North West 0.00 North Central Coast 0.14 0.00 100.00 0.00 0.00 South Central Coast 0.69 0.00 65.47 62.20 27.67 Central Highlands 0.00 South East 0.20 0.00 0.00 100.00 0.0 Mekong River Delta 0.69 30.30 69.70 0.00 0.0 Industry sector Mining 0.00 Manufacturing and construction 0.57 13.94 54.08 31.98 0.00 Trading 0.18 25.17 53.39 21.44 0.00 Hotel and restaurant 0.00 Services 1.70 0.00 78.49 34.58 13.07 Source: Calculating from VLSS 2004; unweighted percentages. According to the survey data, BAC members have a larger operational scale than enterprises that are not associated with any organization. As Table 8.2 shows, the average revenue from member enterprises reaches more than 151 million VND while that of non-member enterprises is less than 29 million VND. The average expenditure from BAC members is more than 107 million VND while that from non-BAC members is about 16 million VND. The total and paid number of workers among BAC members are 5.6 and 3.9 persons per enterprise, respectively, while the same among non-BAC members are only 1.7 and 0.3. Needless to say, all of these differences are statistically highly significant. Table 8.2: Enterprise scale by BAC membership status Member Non-member a Enterprise revenue 151,448 28,762 Enterprise expendituresa 107,419 16,456 Enterprise employment 5.57 1.67 47 Enterprise paid employment 3.87 0.31 Note: a VND 000s Source: Calculating from VLSS 2004; unweighted statistics BAC members have more opportunities in their production and market development through the services that they received from these associations or clubs. Major activities from BAC include: training on management; training on production methods; assistance on product quality and diversification; marketing, advertisement and seeking for outlets for products; seeking for and contacting input suppliers; and distributing information on current policies and legal issues. According to the VLSS 2004, among BAC participating enterprises, a highest percentage is 44.9% of enterprises receiving training on production methods, followed by 33.4% enterprises receiving training on management, 33.1% getting benefit in seeking and contacting input suppliers from BAC, 26.7% getting information on current policies and legal issues, 21.5% improving marketing, advertisement and seeking for outlets for products, and 21.2% receiving assistance on product quality and diversification (see Table 8.3). The service package differed substantially by urban/rural residence. For example, urban enterprises received more training on management and assistance on product quality, while rural enterprises received more training on production methods, assistances for marketing, advertisement and seeking for outlets for products, seeking and contacting input suppliers. This may express the different needs of enterprises in each area, or it may be a consequence of the type of organizations that enterprises in each area associate with (see Table 8.1). Some evidence of the latter is found in the next panel of Table 8.3: enterprises in the North Central Coast area all reported training on production methods as the only service received, but they were members of only one type of organization. Across the various regions, different services receive different emphasis. According to industry sectors, there are large differences among industry sectors in accessing BAC’s services. Manufacture and construction participated most actively in BAC’s activities with higher percentage of participating enterprises: more than 60% received training on production methods and searching for and contacting input suppliers, and about 50% received assistance on product quality and diversification, and marketing, advertisement and seeking for outlets for products. In the trading sector, more than 50% of BAC's members received training on management and production methods. Quite naturally, assistance on product quality and diversification was not as important for them. Enterprises in the service sector did not enjoy as many services. The highest percentages were in the area of training on management and production methods, and on information on current policies and legal issues, but these percentages reached only 26%-29%. Only 6% received assistance on product quality and diversification, and marketing, advertisement and seeking for outlets for products. Table 8.3: Services received from business associations/clubs Assistance on Marketing, Training on product quality advertisement Searching for Information on Training on production and and searching and contacting current policies management methods development/ for outlets for input suppliers and legal issues diversification products Total 33.41 44.89 21.19 21.49 33.06 26.66 48 Urban/rural residence Urban 43.56 38.04 26.97 18.62 25.64 30.86 Rural 18.65 54.85 12.78 25.67 43.85 20.55 Region Red River Delta 46.13 57.03 24.28 24.28 24.28 26.97 North East 36.02 29.45 36.02 36.02 65.46 36.02 N. Central Coast 0.00 100.00 0.00 0.00 0.00 0.00 S. Central Coast 0.00 0.00 37.80 0.00 34.53 34.53 South East 54.91 45.09 45.09 45.09 45.09 54.91 Mekong River D. 28.80 47.66 0.00 19.15 34.30 15.15 Industry sector Manufacturing and construction 33.81 68.91 48.56 49.60 63.54 28.44 Trading 57.30 53.39 21.44 21.44 38.97 21.44 Services 25.84 28.94 5.90 5.90 14.32 27.27 Source: Calculating from VLSS 2004; unweighted percentages. 8.2. Contact with functional agencies Non-farm household enterprises have limited contacts with functional (or government) agencies. According to the survey data, only 10.6% of all enterprises had contact with functional agencies once or twice per year, 3.8% had been contacted 3 or more times per year, and 85.5% did not report having any contact with agencies (Table 8.4). The average number of contacts with functional agencies in enterprises is 0.43 times per year per enterprise nationwide; the highest numbers of contacts per year are 2.8 times in Danang city, and 2.1 times in computer related activities. Urban enterprises have more opportunities for contact with functional agencies than rural enterprises. The percentage of urban enterprises having contact with agencies at least 1 time per year reaches 20%, while that of rural ones is less than 12%, while the percentage being contacted over 3 times per year is 6.3% among urban enterprises and only 2.5% among their rural counterparts. There is also a significant difference in contact opportunities with agencies among regions. Higher percentages are found in the South East, Mekong River Delta and Red River Delta. In the North West, enterprises that have been visited by agencies at 1 time per year accounted for a highest percentage of 8.2% while none of them had been visited 3 time or more. Table 8.4: Contact with agencies per year Number of surveyed No contact 1 time 2 times Over 3 times enterprises Total 85.52 5.03 5.63 3.82 4371 Urban/rural residence Urban 80.13 6.11 7.48 6.29 1511 Rural 88.36 4.45 4.66 2.52 2860 Region Red River Delta 84.96 6.20 5.32 3.52 1041 49 North East 87.28 3.72 5.35 3.65 531 North West 88.41 8.22 3.37 0.00 97 North Central Coast 87.95 1.81 6.28 3.96 497 South Central Coast 89.48 4.22 3.25 3.04 463 Central Highlands 87.09 2.57 4.72 5.62 225 South East 84.24 4.44 6.51 4.81 634 Mekong River Delta 82.58 7.31 6.48 3.63 883 Industry sector Mining 81.54 6.23 6.72 5.51 48 Manufacturing and construction 87.70 5.13 4.63 2.53 1155 Trading 85.72 4.75 5.32 4.21 2005 Hotel and restaurant 82.02 5.03 7.54 5.42 481 Services 84.06 5.58 6.78 3.58 682 Enterprise scale Enterprise revenuea 23,745 52,498 44,314 103,312 4371 a Enterprise expenditure 12,901 33,083 25,222 73,541 4371 Enterprise employment 1.57 2.26 1.89 3.20 4368 Enterprise paid employment 0.23 0.81 0.48 1.65 4361 Note: a Values in thousands VND Source: Calculating from VLSS 2004; unweighted statistics Next, consider variations by industry sector. The highest percentages of enterprises working with functional agencies are found in mining (18.5%) and hotel and restaurant (18.0%) sectors. Lower percentages are found in manufacturing and construction (12.3%), trading (14.3%) and service (15.9%) sectors. In mining and hotel and restaurant enterprises, percentages of those that have contacts with agencies are 5-6% for one time, 6-7% for two times and more than 5% for three or more times. In trading sector, these figures are a little lower with 4.8% for one time, 5.3% for two times and 4.2% for three or more times. In the manufacturing and construction sector, percentages of enterprises contacting agencies account for 5.1% in one time, 4.6% in two times, and only 2.5% in three or more times. Among manufacturing enterprises, fur and wood processing and production, and construction enterprises have fewer opportunities to work with agencies than food and beverage, and other manufacturing ones. In the service sector, 5.6% of enterprises having contacts with agencies for one time, 6.8% for two times and 3.6% for three or more times and 84.1% having not any contact with agencies. The highest percentage of enterprises having contact with agencies in service sector is found in business service with 27%, of which 7.2% for one time, 9.8% for two times and about 10% for three or more times. In term of enterprise scale, there are significant differences in revenue, operation expenditure, employment and paid employment among enterprises based on number of times to contact with functional agencies. Enterprises that have higher revenues, expenditures, and number of employment face more contact with agencies. Thus, enterprises with the most contacts have average revenues of 103 million VND, expenditures of 73 million VND, employment of 3.2 workers and hired employment of 1.7 workers, whereas enterprises that were never contacted had revenues of 24 million VND, expenditures of 13 million VND, employment of 1.6 workers and hired employment of 0.2 workers. Curiously, enterprises with one contact tend to be larger than those with two contacts. 50 The survey questionnaire lists several reasons for meeting with functional agencies, including: working safety, labor, financial, checking business license with real activities, environmental, product quality, and other issues (Table 8.5). Among these, 3-4% of enterprises were contacted for the purpose of checking their business license with their actual business activities and for environmental issues. Lower percentages (1-3%) of enterprises faced contacts about labor, working safety, financial and product quality issues. 7.9% of the entrepreneurs listed other issues. These national trends are roughly representative of urban and rural areas separately, albeit that rural contact rates are generally lower. Across the various regions of Vietnam, checking business license is a frequent objective. Apart from this issue, environmental, product quality and other issues are main reasons for working with agencies in most regions: 2-4% of enterprises have contacts with agencies about environmental and product quality issues, and 5-10% meet about other issues. Work safety issues are given more emphasis only in the North West and Central Highlands, where the percentages of enterprises working with agencies for these reasons reach more than 4%. Labor and financial issues have not been paid attention in most regions, especially in the North West region, where has not any enterprise contacting agencies for this reason. In term of industry sectors, work safety issues generate more frequent visits (4-5%) in mining and transport enterprises, due to the hazardous work in these sectors. Labor issues focus on mining enterprises only (4.9%) and are more rare in other sectors (0.8-1.5%). Financial issues receive little attention in all sectors. Checking business licenses occurs more frequently in trading, hotel and restaurant, transport and other service enterprises (more than 5% each) and in other services (8%), but less frequently in mining, wood and fur processing and production, and food and beverage enterprises (less than 2%). This expresses the more disorderly situation of business registration and licensing in the trading, hotel and restaurant, and service enterprises (tertiary sector) than in mining, manufacture and construction enterprises (secondary sector). In reverse, visiting for environmental issues occur more frequent in the tertiary sector than in the secondary sector. Product quality issues are paid more attention in manufacturing and construction enterprises (especially in food and beverage ones), trading enterprises, and restaurants, where product quality has become a great topic interest of the whole society nowadays. Table 8.5: Reasons for meeting with functional agencies Checking business Product Work safety Labor Financial license with Environment quality issues issues issues real activities al issues issues Other issues Total, as a percentage of all enterprises 2.11 1.06 2.17 4.41 3.16 2.90 7.91 Urban/rural residence Urban 3.22 1.12 2.81 6.61 4.19 3.41 10.62 Rural 1.52 1.03 1.84 3.24 2.62 2.62 6.47 Region Red River Delta 1.68 1.09 2.47 4.13 2.95 2.51 9.21 North East 0.86 0.55 2.00 3.88 3.26 3.06 5.10 North West 4.12 0.00 0.00 4.12 1.37 1.37 4.73 51 North Central Coast 2.59 1.06 2.26 4.25 2.93 3.78 4.44 South Central Coast 2.59 0.89 2.24 3.92 3.62 2.07 5.52 Central Highlands 4.43 0.93 3.19 7.55 4.56 4.09 6.75 South East 2.17 1.00 1.79 5.38 3.05 3.11 8.76 Mekong River Delta 2.09 1.50 2.07 3.88 3.23 2.84 10.48 Industry sector Mining 5.18 4.93 0.00 0.00 6.46 0.00 10.36 Manufacturing and construction 1.90 0.95 1.96 2.42 3.08 2.51 7.30 Trading 1.68 0.96 2.37 5.05 2.78 3.35 6.85 Hotel and restaurant 2.09 0.83 2.71 5.58 4.72 4.32 9.89 Services 3.54 1.46 1.71 5.24 3.11 1.37 10.46 Specific industry Mining 5.18 4.93 0.00 0.00 6.46 0.00 10.36 Food and beverage 1.83 0.57 1.87 2.40 4.00 3.19 5.44 Fur proc and prod 0.00 0.65 1.94 1.93 0.13 0.13 6.79 Wood proc and prod 1.68 1.50 1.99 1.15 1.68 3.49 6.84 Construction 3.32 3.32 6.45 3.32 3.32 6.45 3.32 Other manufacturing 3.35 1.04 1.54 3.79 4.66 1.49 11.51 Trading 1.68 0.96 2.37 5.05 2.78 3.35 6.85 Hotel and restaurant 2.09 0.83 2.71 5.58 4.72 4.32 9.89 Transport 5.09 1.69 2.15 5.11 2.93 1.10 10.35 Business services 2.98 2.98 4.19 3.90 6.19 0.00 22.59 Personal services 1.50 0.89 0.89 3.43 1.55 0.89 5.66 Other services 1.18 0.61 0.00 8.36 3.90 3.44 9.94 Source: Calculating from VLSS 2004; unweighted percentages. 8.3. Sources of information Information plays a very important role in the business activities of household enterprises, especially in non-farm activities where product demand has a higher elasticity than that for farm products. The VHLSS 2004 questionnaire asks entrepreneurs from which sources they acquired information for pricing their products. As recorded in Table 8.6, the primary sources are other traders (44.7%) and firms operating in the same sector (40.4%). Mass media (newspaper, TV, radio) offer information to 5-10% of the enterprises, and the internet is rarely used (0.8%), but many enterprises (36.1%) do find other sources for their pricing practices. There are slight differences in these percentages between rural and urban area. The main difference is that urban enterprises rely more often on other firms in the same sector and their rural counterparts receive more information from traders. This pattern seems plausible: rural enterprises are more scattered. Among economic regions, the northern regions (Red River Delta, North East, North West, and North Central Coast) rely more heavily on mass media and traders; yet, the southern regions (South Central Coast, Central Highlands, South East, and Mekong River Delta) do not compensate for that by seeking other sources. In other words, enterprises in northern provinces are more diligent in seeking out more information from public sources. Differences among sectors are more substantial. The mining sector stands out in that none of the mining enterprises used information from mass media or the Internet. Manufacturing 52 enterprises follow pretty much the general trend. Construction businesses frequently consider the pricing practices of their competitors, and those in manufacturing other than food, fur and wood processing and production gather information more frequently from others in the business and from the television. More than half of trading enterprises receive information from other traders. Hotel and restaurant operators are less likely than the average enterprise to gather information. The same is true for transporters, in particular with respect to traders, as one might well expect. Business service providers rely more on the internet than any other group, more also on the signals received from others in the same business, and less on traders and mass media. Enterprises offering personal and other services have relatively little contact with traders; providers of other services compensate for that with a reliance on mass media. In all, the information sources are varied and depend on the nature of the product/service. Face-to-face contact is clearly important. Table 8.6: Sources of information for pricing products Firms operating in the same Newspaper Radio TV Internet Other traders activity type Other Percentage of all enterprises 4.82 5.74 9.46 0.83 44.66 40.38 36.07 Urban/rural residence Urban 6.00 4.62 9.40 1.24 41.68 45.52 34.54 Rural 4.18 6.33 9.48 0.62 46.24 37.65 36.88 Region Red River Delta 6.59 8.76 13.87 1.65 55.15 43.35 37.65 North East 5.42 7.10 13.54 0.19 47.39 42.67 37.93 North West 7.55 10.44 13.62 1.42 40.26 39.55 28.72 North Central Coast 4.03 4.65 8.38 0.43 52.45 33.86 32.29 South Central Coast 2.72 2.56 4.35 0.00 40.53 46.29 32.37 Central Highlands 3.18 2.47 8.79 0.51 45.52 42.56 37.38 South East 3.99 2.49 4.80 0.73 36.13 38.85 35.10 Mekong River Delta 4.60 6.54 8.77 0.86 33.87 37.57 38.30 Industry sector Mining 0.00 0.00 0.00 0.00 47.02 48.74 23.04 Manufacturing and construction 4.45 4.94 7.62 0.65 43.50 42.49 37.66 Trading 5.42 6.60 11.61 0.73 55.51 38.17 34.40 Hotel and restaurant 3.18 4.90 7.06 0.00 31.81 39.95 38.75 Services 5.14 5.51 8.51 2.08 23.93 43.11 37.28 Specific industry Mining 0.00 0.00 0.00 0.00 47.02 48.74 23.04 Food and beverage 2.83 4.70 5.75 0.50 39.57 40.12 41.09 Fur proc and prod 4.07 5.03 4.97 0.68 31.54 42.54 37.24 Wood proc and prod 4.39 5.34 7.80 0.94 58.34 30.82 32.94 Construction 1.75 3.15 4.91 0.00 34.63 68.80 11.48 Other manufacturing 7.69 5.13 12.58 0.71 45.94 53.52 38.96 53 Trading 5.42 6.60 11.61 0.73 55.51 38.17 34.40 Hotel and restaurant 3.18 4.90 7.06 0.00 31.81 39.95 38.75 Transport 4.46 5.93 8.11 1.30 24.82 40.97 37.25 Business services 4.90 0.00 6.56 5.31 32.86 56.33 28.53 Personal service 5.06 3.36 7.10 4.01 15.23 42.89 40.57 Other services 7.71 9.46 12.5 0.78 26.18 43.76 38.17 Source: Calculating from VLSS 2004; unweighted percentages. Sample size ranges from 4326 to 4337 enterprises. The operational scale of enterprises that have used various sources of information is found larger than that of enterprises that did not use it (Table 8.7). Gaps are larger when it comes to the use of the internet, of newspapers, and of firms operating in the same field of activity. Especially with respect to the internet, the gap is large: revenue and expenditures are more than three times as large as those among non-users. This illustrates the limitation in access to the Internet among small enterprises. Smaller gaps are found with respect to the use of radio and TV, and a smaller gap yet for those acquiring information from other traders. The only break in use/non-use pattern occurs among enterprises using information from other sources. Those that use other sources appear to have a larger operational scale in terms of revenue, expenditure, and employment than those, who have not used these sources. This expresses that relevant information sources may differ and may be specialized for particular enterprises. Table 8.7: Information sources and enterprise operational scale Firms operating in Other the same Sources Newspaper Radio TV Internet traders activities Other Not Not Not Not Not Not Not Using using Using using Using using Using using Using using Using using Using using Revenuea 75.6 27.2 44.9 28.5 40.6 28.2 97.2 28.9 34.5 25.2 45.2 18.8 27.4 30.5 Expenditurea 52.8 15.2 27.2 16.3 23.6 16.3 66.1 16.6 20.7 13.9 28.6 9.1 16.0 17.5 Employment 2.03 1.67 1.89 1.67 1.77 1.67 3.08 1.67 1.71 1.66 2.07 1.42 1.55 1.76 Paid employment 0.61 0.31 0.46 0.32 0.39 0.32 1.84 0.31 0.34 0.31 0.61 0.13 0.25 0.37 Note: a Values in million VND. Source: Calculating from VLSS 2004; unweighted statistics. All differences are statistically significant. 8.4. Operation of the business In the process of production, the enterprise interacts with sellers of raw materials and intermediate inputs and with customers in local and far-away markets. One aspect of this networking process is covered in the VHLSS 2004, which is transportation of commodities. Table 8.8 shows means of transport of raw materials if the enterprise indeed used raw materials (part A) and means of transport of products or services from the enterprise to the customers (part B). It is typical even for a small household enterprise to transport raw materials itself: 75% did so. The main exception is construction enterprises, which typically hire transport, probably due to the bulk and weight of the materials purchased. Yet also more than one third had the seller deliver the raw materials. 54 More than half of the entrepreneurs indicated that the output was transported by the buyer. Quite likely, many of the customers are households and small private enterprises, who purchase small quantities. Somewhat less than one half of the entrepreneurs delivered the output with their own means of transportation. One in nine used hired transport. For a few industries, these questions did not particularly well describe the production process. For example, construction and transport enterprises did not all respond to these questions, and service providers may have produced their output at the location of the customer, making the question difficult to answer. To have both raw materials delivered and output picked up might make the enterprise more vulnerable to outsiders. An entrepreneur with his/her own means of transport may be more at liberty to pursue other suppliers or other customers if the existing ones fail to live up to their promises. Of the 57% that relied on buyer transport, one fifth also relied on raw materials suppliers for delivery and more than half did not use raw materials. Thus, almost one half of all enterprises depend on others for essential transport. To the degree that these are retail customers, this is not a problem, but otherwise it may create a position of dependency. Table 8.8: Delivery process of raw materials and finished outputs A: Means by which B: Means by which acquired raw materials products or services arrived at the enterprisea Percent using were delivered to buyers Percent using Seller Own Hired more than one Buyer Own Hired more than one Industry delivered means transport means transported means transport means Mining 19 68 13 0 68 48 12 31 Food and beverage 35 79 3 18 59 58 3 26 Fur proc and prod 27 83 0 10 80 21 2 11 Wood proc and prod 26 75 7 8 59 54 4 17 Construction 39 23 58 11 29 16 25 23 Other manufacturing 48 62 23 26 63 49 21 34 Trading 35 78 6 17 64 44 18 30 Hotel and restaurant 52 71 3 26 61 22 4 13 Transport 21 87 0 9 9 77 0 4 Business services 18 82 4 5 43 27 9 12 Personal service 17 86 2 4 48 16 1 9 Other services 34 81 3 17 39 30 5 13 Total 38 75 6 18 57 44 11 23 Note: a Provided that the enterprise used raw materials. Row percentages within each category may not add to 100 as entrepreneurs may use several means. The second aspect of the enterprise’s operation network concerns the location of customers. The VHLSS2004 questionnaire lists three environments: “markets within this province/city”, “markets in other provinces/cities”, and international markets. Unfortunately, the first and second questions are phrased somewhat ambiguously. The first seems to refer to local markets, but a province is already a larger environment than the local community. Surprisingly, 24% of the enterprises did not sell anything in markets in this first category, and another 17% stated that the sales there were negligible. Most likely, many enterprises viewed this question as asking about sales in markets within the province other than the local community—but given the 55 percentages of hotel and restaurant operators that sold important quantities in this market category, one still has to wonder, since hotel and restaurant services are not transportable. Sales to markets in other provinces/cities ought to refer to, indeed, other provinces, but the phrasing of “other cities” makes it ambiguous once again. Other than wood processing and productions, other manufacturing, and transport enterprises, few have significant sales in this second category. Sales to international markets are rare, experienced only by some in mining, wood processing and production, and other manufacturing. Table 8.9: Location of output markets Percent Markets within Markets in other serving province/city provinces/cities International markets several Industry M I N O M I N O M I N O markets Mining 37 26 21 16 2 2 10 85 3 6 0 91 20 Food and beverage 40 20 14 26 1 1 1 97 0 1 0 99 5 Fur proc and prod 40 20 15 24 0 2 1 98 0 0 0 100 3 Wood proc and prod 35 26 17 22 2 7 3 87 3 1 1 95 15 Construction 40 22 4 35 0 0 7 93 0 0 0 100 11 Other manufacturing 38 26 17 19 4 10 5 81 1 1 1 97 22 Trading 36 26 16 23 3 5 2 91 0 1 0 99 8 Hotel and restaurant 38 20 20 22 0 1 1 99 0 0 0 99 2 Transport 33 15 21 31 3 8 6 84 0 0 0 100 18 Business services 50 10 20 20 0 4 3 93 0 0 0 100 9 Personal service 35 19 15 31 0 1 0 99 0 0 0 100 1 Other services 39 16 16 29 0 5 1 94 0 1 0 99 5 Total 37 23 17 24 2 4 2 91 0 1 0 99 9 Note: M=Most important, I=Important, N=Negligible, O=None Row percentages within each market category. 9. The impact of the investment climate In this section, we turn our attention to the quantitative impact of investment climate variables that have been the subject of Section 7 on the enterprise and household outcome variables that were discussed in various other parts of this paper. In this analysis, we consider enterprise performance in 2004, enterprise income growth between 2002 and 2004, the entrepreneurship status in 2004, enterprise survival, enterprise start-up between 2002 and 2004, wage employment in 2004, assets in 2004, and sector choice. Since the community variables are measured only in rural communes, the PCI index in 42 provinces, and the MSEFI in all 64 provinces, it is necessary to separate the analysis accordingly. Section 9.1 outlines the methodology; Section 9.2 considers community variables; Section 9.3 looks at the impact of the PCI and MSEFI indices. A summary in Section 9.4 brings the most important investment climate variables together. 56 9.1 Modeling the impact of the investment climate The investment climate has many facets. For example, business decisions are driven by the quality of labor, the many regulations that an entrepreneur must follow, the availability of credit, the access to markets, the quality of the local infrastructure, any natural disasters that have disrupted the economy, and so forth. It is impossible to know a priori which factors are most important. Thus, provided sufficient information is available, all must be investigated. We use the following strategy. The 50 community variables that enter into the analysis are divided into eight groups: economic conditions (9 variables), agriculture (5), infrastructure (7), human resources (8), disasters (8), transport (10), communication (1), and banks (2). The precise definition of these community variables is provided in Appendix A.2. Let us refer to these variables as Z1 through Z8, where Z1 indicates the nine economic conditions variables individually or together, Z2 indicates the agricultural variables, etc. Let Y be an enterprise variable of interest. It is naturally related to various characteristics of the enterprise or the household, denoted as X1, …, Xk or in shorthand (vector) notation as X. If investment climate variables matter, Y is also related to Z1 through Z8. We gather evidence on this issue in four steps. First, we compute correlation coefficients between Y and each Z. Second, we estimate eight so-called partial models that explain Y on the basis of X and a single group Zj with j=1,…,8. 17 Third, we estimate a so-called full model that explains Y with X and all Z variables. Fourth, we reduce the size of the full model by dropping investment climate variables that appear to be irrelevant on the basis of the first three steps. 18 The tables below report the results of the first, second and fourth step of this analysis. The motivation for this strategy is the following. A simple correlation between Y and a single Z variable or a simple regression of Y on a single Z is likely subject to bias. For example, Z1 may not have an impact on Y itself but both may be correlated with other Z’s or X’s, and as a result Z1 may still be correlated with Y. Thus, the correlation coefficients calculated in step 1 cannot provide conclusive evidence. Step 2 controls for some of the factors (in X) that may cause a spurious correlation between Y and a given Z variable, but the potential for bias may still exist since one group of Z variables is likely correlated with other groups. The full model in step 3 controls for all measured investment climate variables and therefore measures the impact of each variable individually. However, the number of investment climate variables is so large that the evidence of the impact of each of them is hampered by multicollinearity 19 and the inclusion of variables that actually have no impact on Y 20. Therefore, by removing in step 4 those variables that appear irrelevant on the basis of steps 1 to 3, the evidence about the relationship between the investment climate and the enterprise variable Y becomes clearer. In all, therefore, 17 In this step, we do actually divide the first group (economic conditions) further into two subgroups, namely investment projects and other variables. But this does actually not matter for the presentation of the results. 18 With respect to Z5 (disasters), in steps 3 and 4, only the total is entered. With respect to Z4 (human resources), the count of the number of problems in education or health care is sometimes omitted in order to reduce multicollinearity, even if these variables appeared statistically significant in steps 1 or 2. 19 Multicollinearity is caused by correlation among the explanatory variables and reduces the precision of the estimated impact of each one of them. In other words, the standard deviations of the estimated slopes rise substantially, the t-statistics become smaller, and the slope estimates themselves become less reliable (they fluctuate more from one sample to the next). 20 Irrelevant explanatory variables do not cause a bias—unlike omitted relevant explanatory variables such as in step 2—but they do raise the standard deviations of the estimated slopes of the other, presumably relevant variables. 57 the results obtained in step 4 are best. But it is still worthwhile to examine results from steps 1 and 2: important investment climate variables will show an impact at all stages of the analysis. In our analysis of the impact of PCI and MSEFI, the approach is similar, with one exception. We stop at the third step, as there is no need to estimate a reduced version of the full model. Note that we do not employ PCI and MSEFI in the same model, since MSEFI is measured in all 64 provinces, whereas PCI is measured only in 42 provinces. Let us now turn to the precise meaning of the dependent variable Y. As mentioned, we consider enterprise performance in 2004, enterprise income growth between 2002 and 2004, the entrepreneurship status in 2004, enterprise survival, enterprise start-up, wage employment in 2004, assets in 2004, and sector choice. Let us discuss these in turn. • Enterprise performance is measured by (i) enterprise total revenue, (ii) value added, (iii) profit, which is revenue minus expenditures, (iv) revenue per employee, (v) value added per employee, and (vi) profit per employee. To reduce the impact of outliers and (relatedly) deal with heteroskedasticity, each of these variables is specified in logarithmic fashion. This means that slope estimates indicate the proportional impact of a one-unit change in Z on Y. The regression models are estimated on the entire sample of non-farm household enterprises in 2004. • Enterprise income growth is indicated through the percentage change from 2002 to 2004 in (i) enterprise total revenue, (ii) value added, and (iii) profit, which is revenue minus expenditures. Growth in the per-employee variables cannot be computed since the questionnaire in 2002 did not record the number of employees in the enterprise. The income growth regression models are estimated on the sample of surviving panel enterprises, without consideration of sample selectivity effects that may exist for the reason that the condition for entry into this sample is survival between 2002 and 2004. • Entrepreneurship status is measured at the household level: does the household operate a non-farm enterprise? Since the entrepreneurship status variable is a 0/1 indicator, the regression model is estimated by logit methods. Furthermore, since the simple correlation coefficient is an inappropriate statistic when the dependent variable is dichotomous, we substitute it for a bivariate logit model, explaining entrepreneurship status with only a single Z variable. The sample consists of households in 2004. • Enterprise survival can be measured only among enterprises in operation by 2002 panel households. Section 5 discusses the procedure that was followed to determine which enterprises survived in 2004. In this analysis, we count both the 1239 panel enterprises and the 41 non-matched enterprises as survivors. The model is estimated with logit methods. • As argued in Section 5, the study of enterprise start-up focuses on each enterprise in 2004 panel households that was not in operation in 2002, regardless of the reported age of the enterprise. The model is estimated with logit methods. A test of whether the start-up process differs between households that already operated an enterprise in 2002 and those that did not revealed no difference, once prior entrepreneurship status in 2002 is controlled for. • The analysis also includes a study of the participation in wage employment, measured at the household level. Wage employment is an alternative way of earning a living (as is indeed farming). People offer their services to others, i.e., employers who themselves respond to the investment climate. Thus, the investment climate drives both non-farm household 58 entrepreneurship and wage employment. In the results reported below, there are many instances where an investment climate variable that would be expected to improve the business climate and encourage entrepreneurship actually (paradoxically) diminishes the rate of non-farm household entrepreneurship but also raises the likelihood that one or more household members work for a wage. Therefore, in order to obtain a more complete picture of the impact of investment climate variables, a study of wage employment is needed. The model is estimated by logit methods on the sample of 2004 households. • As in Section 6, three variables indicate the asset position of the household in 2004: (i) agricultural assets; (ii) business assets; and (iii) consumer durables. We are less comfortable about estimating a pure asset growth model in a way similar to the income growth model, since (a) the model needs to acknowledge the fact that assets owned in 2002 may be liquidated and converted into assets of another type by 2004, and (b) the two rounds of the questionnaire are not completely identical (though this is a minor consideration). Thus, the dependent variables are the asset values in 2004, and in each equation the set of explanatory variables includes all three asset variables in 2002. Even so, the emphasis of the discussion below is on the investment climate variables. The regression models are estimated on the sample of panel households in 2004. • In the study of sector choice, we distinguish five sectors: mining, manufacturing and construction, trading, hotel and restaurant, and services. In urban areas, few mining enterprises exist; thus, this choice option is dropped from the model. With five (or four) sectors, the best estimation method is a multinomial logit approach. The models are estimated on the sample of 2004 non-farm household enterprises. As in other topics above, the simple correlation coefficient could be substituted with a multinomial logit model where the explanatory variable is a single Z, but to save space this result has been omitted from the table. The analysis uses the sample of 2004 non-farm enterprises. This outlines the empirical methodology in broad lines. Before turning to the estimation results that concern the investment climate variables, two notes are in order. First, it is plausible that growth rates respond to changes in the business climate, more so than the conditions of the business climate itself. However, between 2002 and 2004, the community questionnaire changed substantially, adding and rephrasing many questions. Thus the questionnaires allow only for the measurement of only a few changes in the business climate. Therefore, we elect to specify parallel models based on (many) levels rather than (a few) changes in the business climate. Second, there are still a number of details we must present on the results obtained for each topic. • In the enterprise performance models, the set of characteristics that is denoted as X includes the enterprise age, industry, location of operation, the act of selling in markets other than locally, the ethnicity of the household, and urban residence. As it turns out, the investment climate variables proved to have a very similar effect on each enterprise performance measure. Thus, to simplify the discussion of the results, the parameter estimates have been averaged over the six regression models. R2 values of the fully specified regression models range from 0.208 to 0.265 when community variables are entered, from 0.274 to 0.335 when PCI components are used, and from 0.234 to 0.297 with MSEFI components. Note that this does not mean to imply that the provincial PCI components capture the investment climate 59 better than the communes’ community variables: part of this difference is simply caused by the variation in the samples. • Enterprise growth equations do not fit as well, as one might expect since the growth rate depends on two noisy measures (e.g., profits in 2004 and profits in 2002). Full models achieve R2 values between 0.075 and 0.090: the models explain only between 7.5 and 9 percent of the variation in the growth rates. R2 value are lower when aggregate indices are used: between 0.049 and 0.812 with PCI components and 0.031 and 0.056 with MSEFI components. The set of characteristics that is denoted as X includes the age, industry and location of operation of the enterprise, the ethnicity of the household, and urban residence. Enterprise characteristics pertain to 2004, since the 2002 questionnaire did not record these types of variables. • Household characteristics that enter the entrepreneurship status equation include the number of members of various age groups, their average education, gender and ethnicity, and urban residence. The fit of logit models is often indicated with the pseudo-R2 value, which in this case equals 0.060 when using the full list of community variables, 0.054 with the PCI components, and 0.058 with the MSEFI components. 21 • Enterprise survival is driven by characteristics of the enterprise, the entrepreneur, and the household, under the presumption that the continuation of the enterprise is a household decision in which the entrepreneur has a strong say. Thus, the list of explanatory variables includes the industry of the enterprise in 2002, the enterprise income (or profit) it generated, the asset composition of the household, the age, education and gender of the entrepreneur in 2002, and urban residence. The pseudo-R2 of the fully specified community variable model is 0.159; with PCI components, it reached 0.104; and with MSEFI components, it came out at 0.087. • Enterprise start-up between 2002 and 2004 is again a household decision and therefore determined by household factors: its entrepreneurship status in 2002, its asset position, the total income to its disposal (as an indicator to overcome the start-up capital barrier), ethnicity, the number of adults in various age categories, their average education and gender, and urban location. All variables pertain to 2002. The pseudo-R2 values are 0.068 with the full list of community variables, 0.037 with PCI components, and 0.029 with MSEFI components. Even while the models are statistically highly significant, much of the variation in household decisions to start a new enterprise time is left unexplained. • Household characteristics entered into the wage employment equation are the same as those for entrepreneurship status: the number of members of various age groups, their average education, gender and ethnicity, and urban residence. The pseudo-R2 values equals 0.107 when using the full list of community variables, 0.111 with the PCI components, and 0.083 with the MSEFI components. • The household’s asset position in 2004 (in prices of 2002) is explained with the asset position in 2002, household income, the ethnicity, age and education level of the household head, the 21 However, this value of, e.g., 0.060 should not be interpreted as saying that the model explains 6 percent of the variation in entrepreneurship status. Rather, it represents a ratio of the log-likelihood function where the slopes are free and one where the slopes are held to 0—which does not translate to anything easily grasped. The pseudo-R2 is therefore simply an index of the goodness of fit. 60 number of working members in the household and their age, schooling and gender. All of the latter pertain to 2002 as well: the idea is to capture factors that motivate the household to invest in assets of various kinds. For the fully specified model with community variables, the pseudo-R2 values of the tobit model equal 0.083 and 0.094 for agricultural and business assets and 0.218 for consumer durables. With PCI components, the pseudo-R2 values equal 0.124, 0.079 and 0.273 for the three asset types, and with MSEFI components they are 0.110, 0.073, and 0.243. • The characteristics that are assumed to determine sector choice are the number of adult workers in various age categories, their education level and gender, and urban residence. One would presume that the type of schooling or apprenticeship that members of the household received would be important in this choice, but such information is simply not available. The pseudo-R2 values of the multinomial logit model equal 0.087 for the fully specified community variables model, 0.026 (urban) and 0.064 (rural) for the model with PCI components, and 0.018 (urban) and 0.040 (rural) when the model contains MSEFI components. The urban PCI and MSEFI models fit poorly and are indeed statistically insignificant; i.e., the variables that are entered into the model of sector choice among urban enterprises fail to reveal a pattern with respect to age, education, gender and PCI or MSEFI business climate indicators. The general message is that the observable characteristics of the enterprise, the household, and the business climate help to explain a small portion of the variation in the many aspects of the non-farm household enterprise sector that are studied here—but also that much remains to be discovered. The degree to which the business climate variables contribute is of the greatest policy relevance and will be elaborated in the next two sections. The evidence is presented by group of investment climate variables. In this way, it is easier to see whether a particular variable or group of variables has an impact on anything. An alternative way would be to present the estimation results for each dependent variable separately; this shows better which variables impact a given dependent variable. 9.2 Impact of community variables In the following tables, columns indicated with “c” refer to correlation results; those with “p” summarize partial regression models where only one group of community variables is entered together with enterprise and/or household characteristics (“X”); and “b” indicate results from (multinomial) logistic models that include the given investment variable as the only explanatory variable. In these two columns, “p” and “n” stand for statistically significant positive and negative associations; “p?” and “n?” indicate associations that are statistically significant only at the 10 percent level; and “z” (for zero) means a statistically insignificant association. Columns with “Coeff” and “t” show estimation results from fully specified models from which some irrelevant variables have been removed. The interpretation is as follows. • In the case of enterprise performance variables, since the dependent variable was in logarithmic form, the parameter estimate indicates the proportional impact of a one-unit change in the explanatory variable. For example, in Table 9.1, an increase in the population 61 of 1000 residents (which is one unit of ‘population’) raises enterprise performance by 3.1 percent. • In income growth rate equations, the parameter estimate shows the percentage point change in the growth rate as a result of a one-unit change in the explanatory variable. Thus, an increase in population of 1000 residents lowers revenue growth by 2.35 percentage points, value added growth by 2.64 percentage points and profit growth by 2.79 percentage points. • Results of the logit models for entrepreneurship status, enterprise survival, enterprise start- up, and wage employment are also reported as marginal effects, in percentage terms. Thus, an addition of 1000 residents increases the chance that a given household is engaged in a non-farm enterprise by 0.66 percentage points; the chance that a given enterprise survives is 0.139 percentage point higher; the likelihood that a given household starts a new enterprise is reduced by 0.468 percentage points, and the probability that at least one household member holds a wage job decreases by 0.14 percentage points. • Assets are stated in logarithmic form. Thus, the parameter estimates show the proportional impact per unit change in the explanatory variable. When population grows by 1000 residents, a household’s agricultural assets drop by 16 percent, business assets drop by 4.8 percent, consumer durables rise by 1.0 percent, and total assets decrease by 0.7 percent. • The results of the multinomial logit sector choice model show the marginal impact in percentage terms on the likelihood of choosing a given sector. Statistical significance is indicated: “*” indicates a 10 percent statistical significance; “**” denotes a 5 percent statistical significance. In a community with 1000 more residents, the likelihood that an enterprise is in mining diminishes by 0.06 percentage point (and is not statistically significant); manufacturing/construction is also less likely (-0.86 percentage point). A given enterprise is more likely involved in trading (up 0.56 percentage point), services (up 0.22 percentage point) or hotel/restaurant (up 0.12 percentage point), though the latter two effects are statistically insignificant. Table 9.1 lists nine variables under the heading of (general) economic conditions. Population size is an important factor: in larger communities, non-farm entrepreneurship is more prevalent and enterprises tend to be larger (have more revenue, generate more profits). Yet growth and new start-up is lower. The nature of the sector shifts from manufacturing to trading. Enterprise density shows up positively in many partial analyses, but it often fails to be significant in the full model. It does have a significant positive effect on entrepreneurship status, and there it could be interpreted as an agglomeration effect: enterprises benefit from being part of a cluster. But one might wonder if the macro variable of enterprise density might not be a reflection of micro choices of entrepreneurship status. If so, the positive effect is definitional rather than a behavioral response of agglomeration. The negative impact on wage employment would not be inconsistent with that. Net migration impacts entrepreneurship status, enterprise start-up and wage employment positively. It indicates a favorable market potential, in an environment where households tend to be wealthier (at least in terms of consumer durables). Enterprises are more likely involved in the hotel and restaurant business. When community respondents indicate that the local economy has improved over the previous five years, assets other than those used for agriculture tend to be lower, and enterprises 62 are more involved in manufacturing and less in trading. It appears that these communities were not prosperous to start with. A factory in the area lowers household enterprise performance (after controlling for other factors, as the correlation coefficient is positive!), but it encourages entrepreneurship and enterprise start-up. The latter is a significant finding: households appear to learn from productive activity at the factory, or else factory activity creates niches for household enterprises. Wage employment also rises when there is a factory in the area, which is a plausible result. The last four variables represent investments in community projects. Investment in infrastructure has little impact on non-farm enterprises, though it might have an effect on larger enterprises, as wage employment rises with it. Agricultural irrigation projects are associated with lower enterprise performance but yet a greater likelihood of new enterprise start-up. More enterprises are active in manufacturing and construction, fewer in trading: a community with more agricultural irrigation projects is strengthening its agricultural roots, which manufacturing enterprises supplement better than trading enterprises. Investment in utility projects (electric power and water) stimulates investment in business assets. Projects that involve school and health facilities—and thus raise the quality of human resources—reduce enterprise survival, enterprise start-up, and agricultural assets. This reflects a general pattern that the more educated workers leave farming for non-farm self-employment and eventually, with even more schooling completed, enter wage employment (Van der Sluis, Van Praag, and Vijverberg 2005). The fact that investments in school and health facilities do not have an impact on the likelihood of wage employment would still not be inconsistent with this pattern of transition: these investments are recent, but the transition takes longer. Table 9.1: Impact of economic conditions A. Enterprise performance and growth Average of six enterprise Growth rate of performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Population (000s) p p 0.031 5.80 z z -2.35 -1.23 z n? -2.64 -2.04 z n? -2.79 -2.17 Enterprise density p p 0.002 1.14 z z 0.39 0.8 z z 0.53 0.93 z z 0.68 1.02 Net migration p? z zz z z z z Improving economy z z zz z z z z Factory in area p n -0.100 -1.81 z z z z z z Invest: infrastructure z z z n -12.34 -1.63 z z -1.90 -0.32 z z -2.85 -0.47 Invest: agric irrigation n n -0.067 -1.95 z z 3.27 0.21 z z -5.51 -0.62 z z -6.91 -0.73 Invest: utilities z z zz z z z z Invest: health/school z z zz z z z z B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Population (000s) p p 0.660 3.78 z z 0.139 0.31 z n? -0.468 -2.20 p z -0.137 -0.71 Enterprise density p p 0.190 3.53 z z z z z n -0.144 -2.25 Net migration (00s) p p? 0.757 1.51 z z 1.024 0.76 p p 0.016 2.07 p p? 0.008 1.67 Improving economy z z z z -7.170 -0.42 z z z z Factory in area p p 4.547 2.86 z z 1.598 0.35 p? p? 3.787 1.85 p p 4.853 2.82 63 Invest: infrastructure z z z z z n? -0.940 -1.25 p? p? 1.283 1.83 Invest: agric irrigation z z n? z -3.070 -0.99 z z z z 0.984 0.90 Invest: utilities z p? 1.517 1.51 z z z p 2.699 2.24 z z Invest: health/school z z n n -8.198 -3.34 n n -2.539 -2.63 z z C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Population (000s) n n -0.160 -6.55 p z -0.048 -0.70 p p 0.010 2.02 p n? -0.007 -1.11 Enterprise density n z -0.006 -0.74 p z 0.024 1.16 p z 0.003 1.55 z z 0.002 1.29 Net migration z n? -0.001 -1.87 p p? 0.003 1.68 p p? 0.000 1.71 z z 0.000 -0.35 Improving economy z z 0.711 0.82 z z -3.043 -1.31 z n? -0.313 -1.65 n n -0.393 -2.5 Factory in area n n? -0.201 -0.92 p z -0.511 -0.75 p z -0.057 -1.19 z z -0.074 -1.37 Invest: infrastructure n z 0.001 0.01 z z -0.140 -0.50 z z -0.037 -1.92 z z 0.004 0.18 Invest: agric irrigation z p 0.214 1.42 z z -0.222 -0.48 n z -0.021 -0.65 n n -0.043 -1.21 Invest: utilities z z 0.047 0.34 z z 0.768 1.87 n z 0.032 1.08 z z 0.003 0.09 Invest: health/school z n? -0.303 -2.65 z z -0.341 -0.97 z z -0.036 -1.45 z n? -0.068 -2.33 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Population (000s) n n p p p -0.063 -0.836** 0.561** 0.115 0.223 Enterprise density Net migration (00s) n n? p -0.139** 0.868 -0.772 0.578** -0.535 Improving economy n n -3.177** 17.231 -16.926* 0.992 1.881 Factory in area Invest: infrastructure p? -0.280 0.406 -0.137 0.192 -0.181 Invest: agric irrigation p n? n? n? -0.032 5.778** -3.283* -1.195 -1.269 Invest: utilities n p p -0.105 -4.440** 2.490 1.210 0.845 Invest: health/school n? -0.237 0.503 1.149 -0.784 -0.630 Table 9.2 focuses on the second group of business climate variables, those describing the agricultural sector. Access to extension services reduces the rate of growth in the enterprise, but it also tends to reduce agricultural assets. One might have expected that it would reduce the likelihood of non-farm entrepreneurship as Vijverberg and Haughton (2004) found, but no such effect is evident here. This could be because the VHLSS 2002/2004 sample used in these regressions is rural, whereas the Vijverberg/Haughton VHLSS 1993/1998 sample combined urban and rural households. Higher agricultural wages benefit non-farm enterprise performance. More enterprises offer hotel/restaurant services; fewer are engaged in trading. Yet, agricultural wages generate no impact on entrepreneurship status, survival or start-up. 22 The wage effect stands in contrast with the estimated impact of agricultural productivity improvement, which benefits farmers more than laborers: enterprise performance is lower, and total assets are reduced. Land quality should raise 22 In the Vijverberg and Haughton (2004) study of VLSS 1993/1998 data, the agricultural wage raised the likelihood of non-farm entrepreneurship in 1998 and start-up between 1993 and 1998, but had no impact on enterprise survival. 64 agricultural incomes also. In partial analyses, it has an effect on some facets of non-farm entrepreneurship, but in fully specified models the effects vanish. The land transfer rate measures the number of transfers of land use rights certificates per hectare in 2003. This is a feature of the rural environment that changed over the past few years. It has had little impact on non-farm entrepreneurship and enterprises—it may have raised revenue but little else—but it is associated with a small reduction in agricultural assets and with sharp changes in the sector of activity of the non-farm enterprises. Moreover, wage employment rises with the land transfer rate. This may be indicating that facilitating land transactions makes the economy more dynamic. (Note that the average land transfer rate was 0.07 with a standard deviation of 0.44: it would be unusual to have a one-unit change in this variable, even if the maximum value of this variable is 12.2.) Table 9.2: Impact of agriculture A. Enterprise performance and growth Average of six enterprise Growth rate of performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Extension index z z z n -11.76 -1.75 z n -9.99 -2.06 z z -8.34 -1.58 Agric wage p p 0.015 3.60 z z zz z z Agric prod. improv. n n -0.158 -2.46 z z zz z z Land quality index n n 0.006 0.38 z z zz z z Land transfer rate p z z p? 33.67 1.51 z z 2.80 0.18 z z 0.74 0.05 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Extension index z z z z 1.647 0.55 z z -1.418 -1.03 p p 1.423 1.40 Agric wage p z -0.151 -1.09 z z -0.196 -0.54 z z 0.136 0.79 p z -0.010 -0.07 Agric prod. improv. z z z p? 10.333 1.78 z z 1.524 0.62 z p? 2.215 0.99 Land quality index n n? 0.076 0.14 z p 1.878 1.38 z z n n? -0.083 -0.15 Land transfer rate z z z z z z 1.178 1.06 p p 3.831 2.08 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Extension index n n -0.197 -1.79 z z -0.366 -1.07 p p? 0.033 1.39 z n? -0.028 -1.12 Agric wage n n -0.002 -0.09 p z -0.034 -0.58 p p 0.004 0.93 p z 0.004 0.76 Agric prod. improv. z z -0.361 -1.30 z z -0.193 -0.23 z z 0.063 1.05 n n -0.156 -2.09 Land quality index n z 0.009 0.13 z z 0.180 0.84 n z -0.007 -0.44 z z -0.012 -0.7 Land transfer rate n n -0.368 -2.19 z z -0.684 -1.09 z z -0.009 -0.26 z z -0.029 -0.73 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Extension index p -0.483 -0.071 -0.328 1.182** -0.300 Agric wage n p p -0.014 -0.058 -0.359* 0.336** 0.095 Agric prod. improv. 65 Land quality index n? 0.487 -1.099 -0.016 0.182 0.446 Land transfer rate n p p -25.006** 8.251** 10.415** 2.140* 4.200** Table 9.3 considers infrastructure. Having an electric power supply in the commune has strong effects on a number of enterprise factors, but with the exception of sector choice, they are statistically not relevant. The large effect on the likelihood of wage employment is statistically significant. Given a mean of 0.99, it appears that only few communes lack electricity, which could drive these results. The cultural post office variable shows up in many models. It appears to raise enterprise growth but reduces the likelihood of operating an enterprise in 2004, and it may reduce survival. Total assets are lower and mining is more likely. The presence of a cultural house is associated with involvement in a non-farm enterprise and lower agricultural assets. A radio relay station is found in communes with better performing enterprises, but enterprise survival is lower. Proximity to a market should be expected to raise enterprise performance, but it does so only in partial analyses. The only effect it really appear to have is on sector choice: it raises the likelihood of trading and lowers that of services. It may also reduce the likelihood of enterprise start-up, which seems counterintuitive. Note however that the full model also contains other market indicators, as shown in Table 9.6 below. Water quality rarely matters. The only significant impact is that it raises agricultural assets. This is a somewhat unexpected result: apparently, households accumulate more agricultural assets when water comes from a well or is piped. Table 9.3: Impact of infrastructure A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Electric power supply p? z z z 109.14 1.44 z z 14.79 0.33 z z 13.65 0.26 Cultural post office z z 0.048 1.07 z z 25.14 1.6 z z 11.61 1.03 z z 15.11 1.29 Cultural house p? z z z -8.59 -0.55 z z 7.46 0.68 z z 5.21 0.45 Radio relay station p p 0.147 2.04 z z z z z z Proximity to market p p? 0.021 0.40 z z z z z z Water (dry season) z z z z 0.26 0.03 z z -4.00 -0.62 z z -6.97 -0.99 Water (wet season) z n -0.028 -1.26 z z z z z z B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Electric power supply p z -4.707 -0.6 z z 29.982 1.35 z z 3.402 0.49 p p 27.184 3.33 Cultural post office z n -3.507 -2 z z -5.850 -1.35 z z 1.756 1.20 z z 3.027 1.45 Cultural house p p 2.935 2.03 z z z z z z Radio relay station p z -2.017 -0.99 z n -15.345 -2.11 z z p z -2.645 -1.24 Proximity to market p p 0.887 0.55 z z z z -3.087 -1.73 z z -2.023 -1.21 Water (dry season) z z z z z z z p? 0.892 1.23 Water (wet season) p? z 0.705 1.01 z z z z 0.956 1.10 z z 66 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Electric power supply n z -0.948 -0.97 z z 2.754 0.73 p z 0.139 0.64 z z 0.063 0.25 Cultural post office n z -0.302 -1.22 z z 0.175 0.27 n z -0.054 -1.15 n n? -0.118 -2.32 Cultural house n n -0.409 -2.07 z z 0.323 0.58 z z 0.021 0.52 z z -0.024 -0.55 Radio relay station n z 0.432 1.52 p z -0.477 -0.54 p z 0.057 0.93 z z 0.033 0.42 Proximity to market n n? 0.176 0.82 z z -0.520 -0.79 p z -0.075 -1.61 z n -0.089 -1.78 Water (dry season) z z -0.083 -0.82 z p 0.429 1.45 z z 0.006 0.30 z z -0.007 -0.28 Water (wet season) z p? 0.228 2.11 z z -0.280 -0.87 z z 0.007 0.29 z z 0.041 1.62 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Electric power supply p n p? 22.226** -38.351** 32.527 -8.918 -7.485 Cultural post office p n? 0.619** -2.302 4.223 -1.431 -1.110 Cultural house Radio relay station n? n? 0.439 -4.328 1.731 0.884 1.274 Proximity to market n? n? p? -1.050* -1.138 5.239** 0.127 -3.178* Water (dry season) n? -0.178 0.085 1.890* -0.421 -1.376* Water (wet season) The next table studies the effect of human resources. In particular, it addresses proximity to and problems at schools (primary, junior secondary, and senior secondary) and health providers (ranging from provincial hospitals to a private nurse). Most full models omit the problem indices in order to avoid multicollinearity, even if the partial analyses might warrant inclusion. Proximity to a junior secondary school seems to lower enterprise growth, but proximity to a senior secondary school raises it. The contradictory signs are a hallmark of multicollinearity, in view of the fact that most other effects are statistically insignificant. The most solid effect of the proximity to a school is the positive impact of a senior secondary school on the likelihood that a household operates a non-farm enterprise. Problems such as inadequate supplies at senior secondary schools may also raise household entrepreneurship. Proximity to health facilities may raise enterprise performance and enterprise growth, but it may also reduce household entrepreneurship. It does shift enterprises from services, manufacturing/construction and mining into trading, but that may also simply be an indication that health facilities are found in larger communities (see Table 9.1). Table 9.4: Impact of human resources A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t cp Coeff t Primary school (prob.) n? z z z zz zz Jun sec school (prob.) n n z z zz zz Sen sec school (prob.) z p z z zz zz 67 Primary school (prox.) p p 0.193 0.48 z p zz zz Jun sec school (prox.) z z -0.024 -0.13 z z -58.65 -1.14 z z -63.96 -1.71 z z -52.97 -1.39 Sen sec school (prox.) p z 0.001 0.02 z p? 55.77 2.02 z z 29.99 1.7 z z 26.15 1.42 Health facility (prob.) z z z z zz zz Health facility (prox.) p p 0.172 1.12 z z 112.48 1.51 z z 72.45 1.49 z z 64.36 1.27 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Primary school (prob.) z z z z z z z z Jun sec school (prob.) z z z z n? z z p? Sen sec school (prob.) p p? 1.403 1.76 z z z z z z Primary school (prox.) p p? 13.116 0.96 z z -43.824 -1.76 z z -8.804 -0.78 z z Jun sec school (prox.) p z 1.789 0.38 p z 19.197 1.53 z z 4.442 0.84 z z Sen sec school (prox.) p p 6.083 3.18 z z n? n? -2.623 -1.26 p? p 2.122 1.06 Health facility (prob.) n n? -1.491 -1.56 z z z z p p 2.509 2.45 Health facility (prox.) p z -7.671 -1.63 z z z z z z -5.811 -1.18 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Primary school (prob.) p z -0.111 -0.47 z z -0.485 -0.68 z z -0.024 -0.46 z z -0.034 -0.59 Jun sec school (prob.) p z 0.045 0.20 z z 0.273 0.41 z z 0.018 0.37 z z 0.023 0.44 Sen sec school (prob.) p z 0.191 1.59 z z -0.279 -0.76 p z 0.037 1.44 p p 0.079 2.67 Primary school (prox.) n n? -2.154 -1.42 z z -3.935 -0.91 p z 0.153 0.45 z n -0.443 -1.33 Jun sec school (prox.) z z 0.194 0.33 z z -1.061 -0.61 p z 0.102 0.81 z z -0.052 -0.38 Sen sec school (prox.) z z -0.145 -0.58 z z 0.097 0.13 p z 0.009 0.16 z z -0.076 -1.25 Health facility (prob.) z z 0.098 0.77 z z 0.070 0.18 z z 0.017 0.60 z z 0.045 1.39 Health facility (prox.) n n -0.594 -0.97 z z -0.728 -0.38 p p? 0.102 0.76 p z 0.144 0.93 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Primary school (prob.) p? Jun sec school (prob.) Sen sec school (prob.) n n p -0.276 2.168* -1.145 -1.784** 1.037 Primary school (prox.) n p? 0.525 -8.147 -8.044 13.777 1.888 Jun sec school (prox.) n 0.745 3.811 -3.538 -3.328 2.310 Sen sec school (prox.) -0.783 -0.098 2.818 -2.209 0.273 Health facility (prob.) n? Health facility (prox.) n? n p -3.011* -7.594 23.601** 0.467 -13.463** Table 9.5 considers various natural and other disasters that have afflicted the community over the previous three years. Each variable is expressed as the percentage of households in the commune that suffered from the specific event. Only the total is entered into the full model, in order to reduce the size of the model. Disasters reduce enterprise performance (revenue, value added, profit). They may reduce the likelihood of engaging or starting an enterprise and, 68 consistent with this, they raise agricultural assets. In other words, these disasters hold back the economic development of a community. The separate indicators suggest that floods, typhoons, droughts, pests and diseases, and other disasters all depress enterprise performance. The effects on other aspects of non-farm enterprises are more varied and less consistent. Note that fire and epidemics are rare events, impacting 0.2 percent of the households, and therefore rarely register in this analysis. Table 9.5: Impact of natural disasters, fire and epidemics A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Fire (per hh) z z z z z z z z Epidemics (per hh) z z z n z n z n Flood (per hh) n? n z z z z z z Typhoon (per hh) n? n z z z z z z Drought (per hh) n n z z z z z z Pest/disease (per hh) n n z z z z z z Other (per hh) n? n z z z z z z All disasters (per hh) n n -0.124 -2.90 z z z z z z B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Fire (per hh) z p? z z z z z z Epidemics (per hh) n? n? z z z z z z Flood (per hh) z z z z z z z z Typhoon (per hh) z z z z z z z z Drought (per hh) n z z z z z z z Pest/disease (per hh) n n? z n? z z z z Other (per hh) z z p p? z z z p? All disasters (per hh) n? z -1.573 -1.21 z z z z -1.967 -1.14 z z C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Fire (per hh) n n z z z z z z Epidemics (per hh) z p z z n z z p Flood (per hh) z z z z z z n n Typhoon (per hh) p p z z z z z z Drought (per hh) z z n z n z z z Pest/disease (per hh) z z z z n z n z Other (per hh) z z z p? z z z z All disasters (per hh) p p 0.368 2.10 n z -0.291 -0.50 n z 0.054 1.39 n z 0.007 0.17 D. Sector choice Partial model Full model 69 IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Fire (per hh) n p p p p Epidemics (per hh) Flood (per hh) n Typhoon (per hh) p? Drought (per hh) p Pest/disease (per hh) Other (per hh) p p p n p All disasters (per hh) n 0.243 -0.378 0.736 1.514 -2.115 Table 9.6 relates to transport conditions. Each variable measures a proximity to an object, ranging from roads, waterways and public transport, to various types of markets and various kinds of towns. Proximity to a road shows up in many models because it is statistically indicative in partial analyses, but in full models it only raises the incidence of entrepreneurship and wage employment. It may raise enterprise survival chances, but paradoxically it may also reduce growth. In some parts of Vietnam, waterways are important for transporting commodities and people, but the present analysis indicates only that enterprises are less likely to engage in manufacturing and construction. In contrast, public transport has many effects: it raises enterprise performance, might raise growth, might raise survival and start-up, raises total assets while at the same time lowering agricultural assets, and shifts enterprises away from mining. When entered together, the daily and periodic market indices almost always have opposite effects. This makes the interpretation more difficult. The clearest effects are in part B of Table 9.6: proximity to a daily market raises the likelihood that a household operates a non- farm enterprise or that it starts one (as in Vijverberg and Haughton (2004) with VLSS 1993/1998 data), but it reduces survival chances. The combination of the last two of these suggests that a market creates a dynamic environment that, together with competitive forces, causes households to more actively search for better income earning opportunities. Being closer to a commune people committee (CPC) building raises entrepreneurship status, enterprise growth and possibly performance. Trading becomes more likely, manufacturing and construction less likely. There is also a hint that it increases business assets and lowers agricultural assets, but that might also be interpreted as stating that communities in proximity to a CPC building are simply larger. Proximity to a smaller town only lowers agricultural assets. Being nearer to a major town starts to raise the chance of a household operating an enterprise; being near to a large town might actually lower it. A major town offers markets; a large town offers alternative employment opportunities that might be more rewarding than one’s own enterprise. Indeed, proximity to a large or major town increases the likelihood of household members holding wage jobs. Table 9.6: Impact of transport conditions A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t cp Coeff t c p Coeff t cp Coeff t 70 Road index p? n? -0.067 -0.53 z z -65.35 -1.41 z z -38.42 -1.12 z z -44.07 -1.22 Waterway index p z z z z z z z Public trans. index p? p 0.002 2.10 p z 2.06 1.03 z z 0.85 0.94 z z 0.95 0.97 Daily market index p p -0.001 -0.01 z z z z z z Periodic market index n n -0.099 -1.63 z z z z z z Wholesale mkt index p p? -0.049 -0.76 z z z z z z CPC building index p p 0.266 1.47 p p 194.60 2.46 p p 120.49 2.61 p p 136.74 2.84 Town distance index z n? 0.001 0.00 z z z z z z Major town index p z 0.025 0.35 z z z z z z Large town index p? p? 0.036 0.47 z z z z z z B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Road index p p 7.636 2.02 p p 13.830 1.34 z z p p 15.077 3.96 Waterway index z z z z z z z n? -3.112 -1.48 Public trans. index z z p? z 0.188 1.14 z z 0.080 1.35 p? z 0.008 0.22 Daily market index p p 9.282 2.78 z n -20.837 -2.51 p? z 6.308 1.90 p z 3.457 1.07 Periodic market index p z 0.733 0.35 z z z z z z Wholesale mkt index p z -2.348 -1.11 z z -8.146 -1.51 z z z z CPC building index p p 17.009 3.24 z z 0.489 0.04 z z p z 5.385 1.04 Town distance index p z 1.697 0.6 z z z z z z Major town index p p 4.538 1.82 z z z z 0.453 0.14 p p 7.280 2.79 Large town index p z -2.259 -0.93 z z z n -7.280 -2.76 p p 8.933 3.49 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Road index n z -0.118 -0.24 p p? 2.380 1.49 p z -0.150 -1.43 z z -0.024 -0.16 Waterway index n z -0.174 -0.65 z z 0.096 0.12 z n? -0.081 -1.41 z z -0.105 -1.63 Public trans. index n n -0.012 -1.51 z z 0.026 1.21 z z 0.002 1.18 z z 0.003 2.01 Daily market index n n -0.884 -2.14 p z 1.995 1.53 p p 0.083 0.92 p z -0.072 -0.71 Periodic market index p p 0.570 2.01 z z 0.006 0.01 z n -0.162 -2.63 z z -0.058 -0.86 Wholesale mkt index n n -0.555 -1.90 z z -1.295 -1.47 p p 0.154 2.45 p p? 0.027 0.38 CPC building index n z -0.919 -1.45 p z 1.080 0.56 p z -0.171 -1.25 z z -0.162 -0.98 Town distance index n z -1.026 -2.67 p z -0.996 -0.86 p z -0.027 -0.33 z z -0.158 -1.7 Major town index n z 0.057 0.17 p z 0.101 0.10 p z 0.061 0.83 z z -0.030 -0.36 Large town index n z -0.008 -0.02 p z -0.329 -0.34 p z 0.010 0.14 z z -0.023 -0.3 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Road index n? 1.066 -9.300 0.633 2.643 4.958 Waterway index n p p 1.035 -7.801** 2.545 2.000 2.221 Public trans. index n -0.049** 0.041 -0.002 0.024 -0.014 Daily market index n p 1.393 -6.174 -8.137 1.832 11.086** Periodic market index p n n -0.207 8.004** 1.223 -4.794** -4.227** Wholesale mkt index CPC building index n p -0.139 -15.953** 17.758** 0.778 -2.444 71 Town distance index p? n 0.106 6.129 -1.549 1.977 -6.664** Major town index n? -0.813 -0.145 0.576 0.007 0.375 Large town index n? p -0.697 6.674* -3.308 -1.187 -1.481 Table 9.7 looks at communication facilities. In some questionnaires, this might include access to telephone services, but the VHLSS 2004 omitted this question. Access to a telephone was included in the VHLSS 2002, where the average proximity index was as high as 0.88. Thus, by 2004, the question had largely become irrelevant. This leaves us with a single variable in this group: the post office proximity index. Partial analyses often indicate its statistical significance, raising enterprise performance, growth, entrepreneurship status, enterprise survival, the likelihood of engaging in trading, and lowering agricultural assets. In other words, proximity to a post office is in some ways an indicator of local economic development. In the full model, it only matters for survival, and then the effect is strong. In particular, survival chances are 27 percentage points higher for an enterprise in a commune with a post office than for one in a commune far away from a post office. Table 9.7: Impact of communication facilities A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t cp Coeff t c p Coeff t c p Coeff t Post office index p p 0.176 1.46 z z -20.47 -0.49 z p? 4.17 0.13 z p? 0.95 0.03 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment IC variable b p Coeff t b p Coeff t b p Coeff t b p Coeff t Post office index p p -2.896 -0.8 p p 27.158 2.77 z z p? p? -5.801 -1.50 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Post office index nn -0.078 -0.16 p z -0.303 -0.20 p z 0.073 0.71 z z -0.004 -0.03 D. Sector choice Partial model Full model IC variable Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Post office index n n p -1.309 -6.471 2.398 2.448 2.934 Table 9.8 is the last one reporting the effect of community variables. It refers to financial conditions, in particular the proximity to banks where residents deposit savings and receive loans. In partial models, these bank indices are frequently useful explanatory variables. But in full models they rarely matter. Only the proximity to lending banks comes up significant: it raises holdings of agricultural assets and lowers holdings of business assets. The latter is rather 72 unexpected. In addition, it reduces the chance that an enterprise is in the service sector while raising the likelihood of manufacturing and construction. Table 9.8: Impact of financial conditions A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit IC variable c p Coeff t c p Coeff t c p Coeff t c p Coeff t Bank (saving) index p z -0.016 -0.43 z z z z z z Bank (loan) index p p? 0.022 0.79 z z 6.83 0.78 z z 7.74 1.1 z z 6.92 0.95 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment Banks b p Coeff t b p Coeff t b p Coeff t b p Coeff t Bank (saving) index p p 0.450 0.38 z z z z p p 0.191 0.15 Bank (loan) index p z -0.234 -0.28 z z -3.082 -1.34 z z p p? 0.912 0.97 C. Household asset position ln(Agro assets) ln(Business assets) ln(Consumer durables) ln(Total assets) Banks c p Coeff t c p Coeff t c p Coeff t c p Coeff t Bank (saving) index n n -0.219 -1.40 p z 0.149 0.32 p p? 0.047 1.41 p z 0.043 1.12 Bank (loan) index n z 0.262 2.22 z n? -0.753 -2.09 p p 0.021 0.83 p z -0.018 -0.59 D. Sector choice Partial model Full model Banks Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Bank (saving) index Bank (loan) index p? n? 0.382 2.052* 0.058 -0.330 -2.162** 9.3 Impact of PCI and MSEFI The PCI and MSEFI indices and their components summarize the many facets of the investment climate, as seen through the eyes of the private entrepreneur, into aggregates. For our analysis, this may or may not be advantageous. On the one hand, it reduces the number of factors to be considered, which, econometrically, may yield sharper indications as to which broad factors matter. On the other hand, a broad factor may dilute the strong effect that a few narrow factors have. The estimated effect of the aggregate factor averages the effects of its narrower components. Thus, the aggregate factor may elicit weak responses even when a few components matter much. This is true especially when components have opposite effects: an aggregate factor cancels those out. For policy makers, a further disadvantage of using aggregate indices is that the indices do not point directly to specific policy measures. For example, the PCI component of “entry costs” comprises fees, costs of time, bureaucratic hassles, and so forth. “Proactive” measures a general attitude among governing agencies. In other words, if an index matters, a policy maker ought to search for the specific factor that drives the aggregate effect. And as argued above, if an index 73 does not appear to have an effect, a policy maker cannot have full assurance that specific factors do not have any impact. With this in mind, let us turn to the analysis of PCI and MSEFI and their components (which in themselves are also aggregates). Tables 9.9 on PCI and 9.10 on MSEFI resemble those discussed in the previous section, with one exception. The effect of these investment climate measures in the full model is reported in columns headed “st.e.”, denoting a standardized effect. Whereas the MSEFI and its components all range from 0 to 100, with a mean of between 60 and 72 and a standard deviation of around 20, the PCI ranges from 39 to 77 with a mean of 57 and a standard deviation of 7.7, and its components average between 3.9 and 6.3 with a standard deviation of between 0.8 and 2.4. Thus a one-unit change in an explanatory variable means different things across the indices. To make the measured effect comparable, the marginal impact is expressed as an effect of a one-half standard deviation of the explanatory variable. This represents a sizeable change that is nevertheless manageable from a policy perspective. Table 9.9 reports on the impact of PCI and its components. A higher PCI value raises enterprise performance substantially. The estimated impact on enterprise growth is positive in all three columns (but never statistically significant). Nevertheless, household entrepreneurship is held back in provinces with a higher PCI: fewer households operate an enterprise, survival chances are poorer, and start-up is less frequent (though in each case not statistically significant). Total assets are higher but only because consumer durables rise: agricultural and business assets are lower in higher-PCI provinces. When PCI increases, urban enterprises shift from manufacturing and trading into hotel/restaurant and services; rural enterprises also shift into trading. Altogether, therefore, the PCI effect is mixed. The various effects would be consistent with the notion that wage employment in higher-PCI provinces grows rapidly, displacing non- farm entrepreneurship; however, the impact of PCI on wage employment is positive but small and statistically insignificant. The aggregate PCI index is a composite of nine subindices. (Note that each index is defined such that higher values are supposedly better.) Six of the nine components raise enterprise performance; three lower it; but none is statistically significant at the 5 percent level quite yet. Econometrically, the problem of multicollinearity may be to blame. Several components have interesting effects. Higher scores on access to land and a lower burden of compliance and inspections raise enterprise growth. The inspections index also encourages household entrepreneurship, enterprise survival, and enterprise start-up. Similarly, it raises agricultural and business assets. The latter are also favorably influenced by better scores on the informal charges and policy implementation indices, and agricultural assets benefit from a proactive government stance. In rural areas, reduced entry costs raises the chance that an enterprise is in the trading or hotel/restaurant business, both of which are more visible and therefore more liable to pay registration and licensing fees. A greater pursuit of private sector development policy leads to fewer manufacturing/construction enterprises and more traders. 23 Counterintuitive effects also appear among the estimation results. A proactive government reduces non-farm household enterprise growth. A better score on entry costs lowers the likelihood of non-farm household entrepreneurship. In provinces where the government 23 Sector choice in rural areas shows several significant effects involving a decrease or increase in mining. But note that the mining sector is small and dependent on natural resources. This may hamper the reliability of the estimation results. 74 pursues private sector development policies, enterprise survival and enterprise start-up may be lower. Lower entry costs and greater access to land reduce most types of assets, and reduced favoritism towards state-owned enterprises and private sector development policies may also lower agricultural assets. Several of these results might indicate that higher scores on PCI components encourages the development of larger enterprises, perhaps not owned by households but rather by corporations, cooperatives or multinationals, which offer wage employment and displace household entrepreneurship. This may be the case with the proactive attitude and SOE favoritism index, but the private sector development policy index appears to lower wage employment. The authors of the PCI index constructed three structural endowment variables as well. Among structural endowments, better infrastructure and closer proximity to markets raises enterprise performance, but while infrastructure raises the likelihood of private entrepreneurship as well as wage employment, market proximity lowers entrepreneurship and does little to wage employment. Both lower agricultural and business capital, and infrastructure raises total capital and consumer durables. In all, the overall impact of PCI index appears strongly related to these two components. Better provincial scores on the human capital endowment component lowers enterprise performance, somewhat surprisingly. It may raise revenue growth but not growth in value added or profits. As one might expect, wage employment rises with the human capital index; non-farm entrepreneurship might be enhanced, but the positive estimates are statistically weak. The only other impact is found in a sectoral shift among rural enterprises towards manufacturing. Table 9.9: Impact of PCI. A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit c p st.e. t c p st.e. t cp st.e. t cp st.e. t Aggregate PCI p .. 0.065 7.13 z .. 5.57 1.53 z .. 2.16 0.74 z .. 0.86 0.30 Components Entry costs p p -0.020 -1.60 z z 4.08 0.86 z z 1.81 0.49 z z -1.17 -0.32 Land p p 0.006 0.37 z p 17.69 2.37 z z 8.61 2.04 z z 7.57 1.75 Transparency p p -0.010 -0.77 z z -3.00 -0.50 z z -0.61 -0.14 z z 0.45 0.10 Inspections z z 0.006 0.53 z p? 9.23 2.14 z z 6.58 2.11 z z 4.98 1.66 Informal charges n n 0.002 0.15 z z 5.01 1.14 z z 1.82 0.59 z z 1.83 0.59 Implementation z z 0.016 1.20 z z 11.26 2.22 z z 2.24 0.57 z z 2.32 0.60 SOE favoritism p p -0.019 -1.48 z p? 8.03 1.69 z z 5.07 1.41 z z 4.28 1.18 Proactive p p 0.033 1.79 z z -25.71 -2.76 z z -13.12 -2.16 z z -12.03 -1.97 Dev. policy p p 0.018 1.67 z z 11.04 1.41 z z 6.33 1.46 z z 6.60 1.47 Structural endowments Human capital n n -0.038 -3.29 z z 10.03 1.95 z z 0.85 0.25 z z 0.92 0.27 Infrastructure p p 0.099 7.24 z p 7.71 1.33 z z -0.94 -0.24 z z -2.08 -0.54 Market proximity p p 0.045 4.45 z z -0.52 -0.11 z z -0.31 -0.08 z z 1.37 0.35 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment 75 b p st.e. t b p st.e. t bp st.e. t b p st.e. t Aggregate PCI z .. -0.445 -1.29 z .. -0.603 -0.72 z .. -0.751 -1.61 p .. 0.085 0.21 Components Entry costs z n -1.053 -2.25 z z 0.586 0.51 z z -0.685 -1.14 p? z -0.422 -0.75 Land p z -0.119 -0.21 z n? -0.254 -0.18 z z 0.173 0.23 p z -0.364 -0.54 Transparency p z -0.097 -0.21 z z -0.652 -0.54 z z -0.068 -0.12 z z -1.156 -1.84 Inspections p p? 1.140 2.79 p? p 2.126 2.24 p p 0.973 1.89 z z -0.279 -0.54 Informal charges n z 0.247 0.62 z p? 1.155 1.19 z z 0.046 0.09 n n? -0.205 -0.39 Implementation z z 1.123 2.4 z z 0.970 0.83 z z 0.082 0.15 z z 0.427 0.73 SOE favoritism z n? -0.514 -1.03 z z 0.427 0.36 z z 0.698 1.07 z z 1.000 1.52 Proactive z z 0.083 0.12 z z -0.046 -0.03 z n -0.491 -0.59 p z 1.289 1.48 Dev. policy z z -0.292 -0.63 z n -1.733 -1.58 z n -1.074 -1.8 z z -1.956 -3.00 Structural endowments Human capital p z 0.462 1.03 z z 1.246 1.14 z p 0.680 1.19 n p 1.763 3.07 Infrastructure p z 1.063 2.03 z n? -0.733 -0.54 p z 0.996 1.45 p p 3.261 4.49 Market proximity z n? -0.724 -1.77 z n? -2.399 -2.38 z z -0.428 -0.82 p p 0.373 0.75 C. Household asset position ln(Agro assets) ln(Business assets ln(Consumer durables) ln(Total assets) c p st.e. t c p st.e. t c p st.e. t c p st.e. t Aggregate PCI n .. -0.152 -2.52 z .. -0.264 -1.82 p .. 0.023 2.49 p .. 0.022 1.95 Components Entry costs n n -0.309 -3.96 z n -0.241 -1.27 p z -0.013 -1.02 p z -0.050 -3.34 Land n n -0.283 -3.11 p n -0.114 -0.50 p p? -0.028 -1.92 p z -0.043 -2.4 Transparency n z -0.004 -0.05 z z -0.084 -0.44 p z -0.018 -1.45 p z -0.043 -2.87 Inspections p p 0.235 3.40 z z 0.314 1.93 p z -0.009 -0.78 p z 0.022 1.66 Informal charges p p 0.195 2.85 z p 0.474 2.96 n z 0.026 2.46 n z 0.039 2.97 Implementation p p 0.143 1.94 p p 0.496 2.78 n z 0.008 0.67 z p 0.027 1.91 SOE favoritism z z -0.123 -1.46 z z -0.199 -1.00 p p 0.004 0.34 p p? -0.022 -1.41 Proactive n z 0.353 3.07 p z 0.132 0.48 p p 0.031 1.69 p p 0.078 3.53 Dev. policy n n -0.139 -1.72 z n -0.143 -0.79 p p 0.001 0.06 p p 0.014 0.91 Structural endowments Human capital p z -0.041 -0.55 n n? 0.007 0.04 p z 0.002 0.20 z z 0.011 0.82 Infrastructure n n -0.304 -3.42 z n -0.445 -2.15 p p 0.069 5.20 p z 0.056 3.07 Market proximity z n -0.241 -3.36 n n -0.610 -3.59 p z 0.006 0.52 p z 0.012 0.87 D. Sector choice Partial model Full model Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Urban areas Aggregate PCI .. .. -1.535** -1.967** 1.514** 1.989** Components Entry costs .. n p .. -1.060 -1.014 -0.026 2.101** Land .. n? p .. 0.749 -2.115 0.384 0.982 Transparency .. n p? p .. 0.752 -1.393 0.473 0.169 Inspections .. .. 0.468 -0.427 0.796 -0.837 Informal charges .. .. -0.818 -1.044 1.102 0.760 76 Implementation .. .. 1.038 0.118 -0.522 -0.634 SOE favoritism .. n p .. 0.672 -0.616 -0.072 0.016 Proactive .. n? p .. -1.433 1.231 0.340 -0.138 Dev. Policy .. n .. -0.435 0.107 0.441 -0.113 Structural endowments Human capital .. .. 0.843 -0.163 -0.520 -0.160 Infrastructure .. n? p .. -1.239 0.310 0.283 0.646 Market proximity .. .. 0.159 -0.257 -0.206 0.303 Rural areas Aggregate PCI -0.019 -3.391** 0.939* 1.490** 0.980** Components Entry costs n p p p? -0.251 -2.780** 1.590** 1.098** 0.342 Land n? n p p -0.578** -1.062 -1.002 1.302** 1.339** Transparency n 0.556** 1.158 -1.246 -0.644 0.175 Inspections n p n n? -0.905** 0.355 0.251 0.330 -0.031 Informal charges n? -0.521* 0.977 0.218 -0.474 -0.199 Implementation n? p -0.214 1.226* -0.160 0.494 -1.346** SOE favoritism p? n p 0.909** 0.423 -0.224 0.153 -1.260** Proactive n p p -0.799* -0.581 0.103 0.537 0.740 Dev. policy p n p p p 0.620** -2.847** 1.950** -0.265 0.541 Structural endowments Human capital p n n -0.067 2.746** 1.109 -1.476** -2.313** Infrastructure n p p? -0.274 0.871 -0.116 0.427 -0.907 Market proximity n -0.175* -0.125 0.540 -0.725 0.484 As with the PCI subindices, the MSEFI components are scaled such that higher values are better. Enterprise performance rises with the MSEFI index: a one-half standard deviation improvement in MSEFI raises performance by 2.5 percent. Enterprise growth appears to rise as well, although none of the estimated effects is quite statistically significant. Entrepreneurship status, enterprise survival and enterprise start-up may all be reduced (but again, this is statistically weak evidence), and employment may rise. Assets are not impacted, with the exception of a drop in agricultural assets. Sector choice is hardly influenced either. Surprisingly, all six MSEFI components have a significant impact on enterprise performance. Four of the six have a significant positive effect; the other two (lower cost of finance, and ease of registration procedures) have a significant negative effect. However, none of the indices matter for growth. The (weak) aggregate MSEFI effect remains therefore unexplained. Normally one might expect that a factor that raises enterprise performance would also encourage entrepreneurship, but actually, it also matters how this factor impacts the attractiveness of wage employment. In our case here, economic policy does raise both performance and entrepreneurship status; the negative influence on enterprise survival and start- up would therefore indicate that economic policy contributes to a satisfactory state of stability. The other three contributors to better performance have weak negative effects on the likelihood of entrepreneurship. Better infrastructure appears to lower both entrepreneurship and wage employment, as does reduced theft and corruption. A higher labor index might lower 77 entrepreneurship but actually raises the chances of household members holding a wage job. Of the two factors that lower enterprise performance, an improved score on finance raises both entrepreneurship and wage employment. Thus, overall, employment benefits from a well- functioning financial market. As to the second factor, fewer registration hurdles may well lower both entrepreneurship and wage employment. It is not clear why this might be so. Infrastructure lowers agricultural and business capital (as it also did with the PCI-related measure of provincial infrastructural endowments). Easier registration raises agricultural and business assets but lowers consumer durables and total assets. More favorable economic policies shift assets from agricultural to business uses and to consumer durables. Reduced corruption and theft increases total assets, mostly through agricultural assets. Sector choice is impacted to some extent by MSEFI components but not necessarily in ways that compare with the PCI indices. For example, unlike the PCI effect, ease of registration reduces the size of the service sector in favor of manufacturing/construction. In rural areas, better finance increases the number of manufacturing/construction enterprises at the cost of services; more favorable economic policies in turn reduces manufacturing/construction in favor of hotel/restaurant and services, but reduction in theft and corruption work the exact opposite direction; and better labor conditions also reduces manufacturing/construction in favor of hotel/restaurant. Table 9.10: Impact of MSEFI. A. Enterprise performance and growth Average of six enterprise Growth rate in performance measures Revenue Value added Profit c p st.e. t c p st.e. t c p st.e. t c p st.e. t Aggregate MSEFI p? 0.025 3.19 z 4.692 1.32 z 3.317 1.40 z 3.575 1.35 Components Infrastructure p? p 0.034 2.57 z z 2.801 0.44 z z 2.714 0.59 z z 4.639 1.03 Finance n n -0.095 -6.42 z z -11.393 -1.52 z z -2.068 -0.42 z z -2.915 -0.60 Registration z z -0.030 -2.62 z z 3.653 0.77 z z 1.403 0.42 z z 2.559 0.71 Labor p p 0.055 4.68 z z 3.691 0.85 z z -0.209 -0.06 z z -1.995 -0.52 Economic policy p p 0.044 3.41 z z -5.520 -1.03 z z -1.282 -0.34 z z -1.067 -0.28 Corruption z p 0.037 3.42 z z 12.231 1.69 z z 2.968 0.71 z z 2.337 0.57 B. Entrepreneurship status, enterprise survival, enterprise start-up, and wage employment Entrepreneurship Enterprise survival Enterprise start-up Wage employment bp st.e. t bp st.e. t bp st.e. t b p st.e. t Aggregate MSEFI z .. -0.365 -1.23 z .. -0.710 -1.05 z .. -0.367 -1.08 p? .. 0.427 1.20 Components Infrastructure z z -0.744 -1.55 z z -0.788 -0.75 z z -0.847 -1.48 z z -0.906 -1.58 Finance z z 1.098 2.24 z z 0.117 0.09 z z 0.441 0.71 z p? 1.876 3.13 Registration n z -0.639 -1.53 z z 0.720 0.67 z z 0.344 0.67 z z -1.148 -2.23 Labor z z -0.348 -0.84 p z 1.419 1.36 z z 0.330 0.63 p p 1.298 2.67 Economic policy z z 1.071 2.27 z z -1.395 -1.35 z n -1.328 -2.20 p z -0.048 -0.09 Corruption z z -0.714 -1.73 z z -0.618 -0.64 z z 0.693 1.24 z z -0.589 -1.11 78 C. Household asset position ln(Agro assets) ln(Business assets ln(Consumer durables) ln(Total assets) c p st.e. t c p st.e. t c p st.e. t c p st.e. t Aggregate MSEFI n .. -0.107 -2.30 z .. -0.048 -0.40 z .. 0.003 0.34 z .. 0.005 0.52 Components Infrastructure n n -0.225 -2.99 z z -0.374 -1.95 z z -0.008 -0.60 z z 0.021 1.37 Finance n z -0.024 -0.31 z z -0.205 -1.00 n z -0.007 -0.48 n n -0.051 -3.19 Registration z z 0.209 3.14 z p 0.520 2.99 n n? -0.034 -2.77 n n -0.038 -2.77 Labor n n? 0.024 0.38 z n? -0.291 -1.73 p z 0.006 0.50 p z 0.020 1.49 Economic policy n n -0.209 -2.88 p z 0.333 1.78 z z 0.038 2.79 z n? -0.020 -1.38 Corruption z z 0.109 1.59 z z -0.029 -0.17 z z 0.011 0.90 z p 0.065 4.64 D. Sector choice Partial model Full model Mi. Ma. Tr. Ho. Se. Mining Manuf&Con Trading Hotel&Rest Services Urban areas Aggregate MSEFI .. .. 0.166 0.711 -0.695 -0.182 Components Infrastructure .. .. 0.344 -0.175 -1.029 0.860 Finance .. .. 0.877 0.504 -0.809 -0.572 Registration .. .. 1.492** -0.069 0.276 -1.699* Labor .. n p? .. -2.459** 0.308 0.407 1.745* Economic policy .. .. -0.940 -0.288 0.838 0.391 Corruption .. .. -0.259 0.332 -0.048 -0.025 Rural areas Aggregate MSEFI 0.124 -0.481 -0.746 0.819** 0.284 Components Infrastructure n? p p? 0.137 0.002 -0.846 -0.149 0.856* Finance p 0.281 1.986** -0.748 -0.180 -1.338** Registration p p 0.346** -0.587 0.581 -0.143 -0.198 Labor n p p -0.438** -1.801** 0.578 1.298** 0.362 Economic policy n n? p p -0.406** -2.774** -0.385 1.766** 1.800** Corruption p 0.555** 2.033** -0.958 -0.749* -0.881* MSEFI differs from PCI in that PCI is based on responses of larger-scale private entrepreneurs to a specialized questionnaire, whereas MSEFI is derived from responses of operators of household enterprises to questions in the VHLSS multipurpose questionnaire. One may thus hypothesize that the MSEFI is more relevant for the operation of non-farm household enterprises than the PCI. The calculation of the standardized effects permits a direct comparison of the PCI and MSEFI effects. With respect to enterprise performance, growth, entrepreneurship status, survival and start-up, the direction and magnitude of the effects is similar. PCI has a stronger effect on assets and sector choice. 79 Given that PCI and MSEFI are hardly correlated (r=0.029; see Section 7.1), it is feasible to enter both in the same regression model. The average effects on performance are the following: a half-standard deviation increase in PCI raises enterprise performance by 6.4 percent (with t=6.98), and a half-standard deviation in MSEFI increases enterprise performance by 2.0 percent (with t=2.44). These effects are very close to those reported in Tables 9.9A and 9.10A. The same happens with regression outcomes for enterprise growth and entrepreneurship status. Thus, the MSEFI and PCI indices are truly complementary: they impact performance facets of micro and small household enterprises along different channels, while yielding similar impacts. 9.4 Summary We close this analytical, policy-focused section with a summary of the evidence, concentrating now not on the investment climate variables but on the non-farm enterprise outcomes. Table 9.11 lists the variables that proved to be most important in explaining the variation in enterprise performance, entrepreneurship status, wage employment, and ownership of business and agricultural assets. The table separates community variables from PCI and MSEFI components, because data sources and samples differ: community variables are measured through the VHLSS questionnaire in rural communes; PCI indices come from an extraneous source (Vietnam Chamber of Commerce and Industry 2005) and describe 40 provinces; and MSEFI indices summarize VHLSS 2004 responses of non-farm enterprise operators, covering all provinces of Vietnam. Thus differences between the columns as to which investment climate factors matter may derive from the variation in the definition of the factors and from the difference in the coverage of the sample. Since the results have been discussed in detail in the previous subsections, we abstain from further detailed comments here. A few investment climate facets appear frequently: population size of the community; proximity to major or large towns; proximity to markets; infrastructure; registration; and policy implementation or economic policy. Table 9.11: Important determinants of non-farm enterprise performance and entrepreneurship Outcome variable Community variables PCI components MSEFI components Enterprise Population (+) Infrastructure (+) Infrastructure (+) performance Agricultural wage (+) Market proximity (+) Finance (–) Agric. prod. improvement (–) Human capital (–) Registration (–) Disasters (–) Labor (+) Economic policy (+) Theft and corruption (+) Enterprise growth Population (–) Land(+) Access to agric extension (–) Inspections (+) Proximity to CPC building (+) Proactive stance (–) Non-farm Population (+) Entry costs (–) Finance (+) entrepreneurship Factory in area (+) Inspections (+) Economic policy (+) Proximity to senior secondary Policy implementation (+) school (+) Infrastructure (+) Proximity to CPC building (+) Proximity to major town (+) 80 Enterprise survival Investments in health and school Inspections (+) facilities (–) Market proximity (–) Proximity to daily market (–) Proximity to post office (+) Start-up of a non- Population (–) Economic policy (–) farm enterprise Investments in health and school facilities (–) Investments in utilities (–) Proximity to large town (–) Wage employment Factory in area (+) Private sector development Finance (+) Electric power supply (+) policy (–) Registration (–) Road (+) Human capital (+) Labor (+) Proximity to major town (+) Infrastructure (+) Proximity to large town (+) Agricultural assets Population (–) Entry costs (–) Infrastructure (–) Investments in health and school Land (–) Registration (+) facilities (–) Inspections (+) Economic policy (+) Land transfer rate (–) Proactive stance (+) Water quality (wet season) (+) Infrastructure (–) Disasters (+) Market proximity (–) Proximity to daily market (–) Proximity to small town (–) Proximity to lending bank (+) Business assets Investments in utilities (+) Informal charges (+) Infrastructure (–) Proximity to lending bank (–) Policy implementation (+) Registration (+) Infrastructure (–) Market proximity (–) 10. Counting household enterprises: AHBS and VHLSS In this last section of the report, we return to an issue that concerns the validity of the Vietnam Household Living Standards Survey for a study on household enterprises. In particular, does the VHLSS cover the population of household enterprises adequately? As it happens, the General Statistics Office of the Government of Vietnam conducts an annual census of household businesses under the name of the Annual Household Business Survey (AHBS). Its purpose is to account for the amount of production produced in and employment provided by enterprises that are owned and operated by private households. Thus, the AHBS and VHLSS have an overlapping interest in this aspect and can inform each other about accuracy in capturing the non-farm household enterprise segment in the economy of Vietnam. The VHLSS suggests existence of around 9.34 million non-farm enterprises in Vietnam; the AHBS data measures about 2.9 million units. This is a large gap. Of course, the timing of the surveys differs slightly—AHBS in 2003 and VHLSS in 2004—, which might explain a small part of the gap. However, the difference in the timing is not really one year: while the AHBS counts enterprises as of 1 October 2003, the VHLSS is administered during the Spring of 2004, seeking retrospective information over the previous 12 months. The question arises what the reasons are behind this discrepancy and whether that impacts the accounting for the production and employment in the private household enterprise segment of the economy of Vietnam. 81 There are four reasons for this difference (Vijverberg 2006). First, the VHLSS surveys each separate household enterprise that occurs in the sampled households, whereas the AHBS conducts its survey on the basis of the concept of a business household, which aggregates every different own-account activity that takes place in a given household. Second, the AHBS uses a different, more restrictive set of inclusion criteria than the VHLSS. Third, the VHLSS enumerates all activities that have taken place over the twelve months prior to the time of the survey visit. In contrast, the AHBS takes a snapshot of activities on a given date. Fourth, after making adjustments for these three factors, it still appears that the AHBS’s count of household businesses is incomplete. Let us consider each of these reasons in turn. The first reason means that a comparison of 9.34 million VHLSS household enterprises and 2.9 million AHBS units is in fact void. The latter should be compared with the number of business households in Vietnam as estimated by the VHLSS, which equals 7.50 million. That reduces the gap by 1.84 million units but still leaves a large gap of 4.6 million. Second, when AHBS criteria of what constitutes a business household are applied to the VHLSS count, between 897 and 1275 of the 3621 business households in the VHLSS sample drop out. These criteria are the following: i. The business household must operate at least one enterprise that generates monetary revenues. ii. Responses to questions about the enterprise must be complete. In particular, because the enterprise questions are split over two modules, the household respondents (i.e., the entrepreneur that operates each enterprise) must address both modules. iii. Enterprises in the household must be in operation for at least three months during the year, unless they are new enterprises. iv. The household must operate its enterprise from a fixed location, unless they are in the transport sector. v. Enterprises operate a significant part of the month. In particular, if an activity is in operation for more 11 or 12 months per year, it must operate at least 10 days per month; if an activity is in operation between 6 and 10 months, it must operate at least 15 days per month; and if an activity is in operation at most 5 months per year, it must operate at least 20 days per month. The fourth criterion in this list gives rise to some ambiguity. The VHLSS inquires about the location of the enterprise, allowing for a maximum of three responses. 24 An enterprise is assumed to operate from a fixed location if the entrepreneur provides only one response and this response is not an ‘unfixed place’. The business household is then assumed to operate from a fixed location if the household enterprise operates from a fixed location. The ambiguity arises when the household operates more than one enterprise. A loose interpretation of this location restriction requires that at least one enterprise has a fixed location; a stringent interpretation requires that all enterprises operate from the same fixed location. The difference in interpretation is not insignificant: it removes an additional 379 business households from the sample. 24 The possible responses are at home, industrial zones, trading zones/centers, markets, other independent stores, other fixed locations, and unfixed places. 82 The third reason impacts enterprises that operate seasonally or have recently closed down. The AHBS counts business households where the economic activity was ongoing as of the first day of October of the given year. Seasonal enterprises that are active only from, for example, March to August are not counted in the AHBS database, but with its retrospective focus the VHLSS does include them. Since the VHLSS database omits information about the dates at which the enterprise was operating, the adjustment to the VHLSS count is made probabilistically and leads to a drop of another 230.9 business households under the loose location restriction and 222.3 units under the stringent location restriction. Together, the second and third reasons reduce the VHLSS count of business households from 7.50 million to 5.24 million under the loose location restriction and 4.50 million under the stringent location restriction. Thus, of the gap of 4.6 million business households, the AHBS inclusion criteria explain between 49 and 65%. There still remains a gap of an estimated 2.34 million or 1.60 million between the AHBS and VHLSS counts, depending how the AHBS inclusion restrictions are interpreted. This residual gap is therefore an estimate of the undercount in the AHBS census of business households (the fourth reason). What does all of this mean for the accounting of production and employment in the private household enterprise segment of the economy? According to information gathered with the VHLSS, the AHBS undercount (the fourth reason above) implies that the value added in the business household sector is understated by between 5.4 and 80.7% and yearround employment by between 16.0 and 80.7%. These ranges are wide because it is a priori unknown which types of enterprises were missed in the AHBS. The low end of the range assumes that the AHBS missed all of the smallest and none of the larger units and that the strict location restriction assumption was appropriate; the high end assumes that the AHBS missed units completely at random and that the loose location restriction was more appropriate. The question of what constitutes a large and small enterprise unit also plays a role: is it measured through employment or value added? Overall, the percentage of the uncounted-for value added and employment probably lies in the middle of the given ranges. The restrictive AHBS rules (the second and third reasons above) also lead to an underestimate of the value added and employment contributed by the household enterprise sector. In particular, if the loose location restriction is appropriate, value added is understated by 22.7% and yearround employment by 25.7%; under the stringent location restriction, these percentages are 47.3 and 54.8, respectively. 25 Note that if the stringent location restriction is a closer approximation of the AHBS rules, the estimated undercount in the AHBS is smaller, implying a smaller understated value added and employment, but the restrictiveness of the AHBS rules then also leads to a larger underestimate of the value added and employment. In other words, the difference in the 2.9 million business households in the AHBS and 7.5 million business households in the VHLSS may be apportioned in different ways to a pure undercount (the fourth reason) and an underestimate due to restrictive AHBS rules (the second and third reasons). Only if the AHBS managed to survey primarily large enterprise units is the overall understatement of value added and employment somewhat smaller. 25 These corrections are on top of the adjustments for undercount as reported in the previous paragraph, i.e., multiplicative rather than additive. 83 Non-farm household enterprises are often not well documented. In most countries, the vast majority of them are not registered, and Vietnam is no exception. With the existence of two micro-level data sources Vietnam’s situation is rather unique. Given the different methodologies, a discrepancy in the measured size of the private household enterprise sector is not unexpected. The AHBS count could have exposed an undercount in the non-farm private economic activities in the VHLSS (after adjustments), but as it is, the count through the VHLSS proves to be more exhaustive. This information can be used to sharpen the AHBS methodology. References De Janvry, Alain, and Elisabeth Sadoulet. “Rural poverty in Latin America: determinants and exit paths.” Food Policy, 2000, 25, 389-409. Hare, Denise. “The determinants of job location and its effect on migrants’ wages: evidence from rural China.” Economic Development and Cultural Change, April 2002, 50:3, 557- 579. Hoang Thi Thanh Huong, “A comparison of investment climate using VHLSS 2002 and VHLSS 2004.” Background Report, The World Bank, March 2006. Hoang Thi Thanh Huong, “Performance of non-farm household enterprises (NFHEs) in VHLSS 2004 and VHLSS 2002.” Background Report, The World Bank, March 2006. Johnson, D. Gale. “Can agricultural labour adjustment occur primarily through creation of rural non-farm jobs in China?” Urban Studies, December 2002, 39:12, 2163-2174. Liang, Zai, Yiu Por Chen, and Yanmin Chen. “Rural industrialization and internal migration in China.” Urban Studies, December 2002, 39:12, 2175-2187. Mwabu, Germano, and Erik Thorbecke. “Rural development, growth and poverty in Africa.” Journal of African Economies, 2004, 13:Suppl.1, 116-165. Nguyen Chien Thang, “Households engaging in non-farm business activities in the VHLSS 2004,” Background Report, The World Bank, March 2006. Nguyen Chien Thang, “Recent entry into non-farm household business activities in the VHLSS 2004.” Background Report, The World Bank, March 2006. Nguyen Chien Thang, “Sector choice among non-farm household enterprises in VHLSS 2004.” Background Report, The World Bank, March 2006. Nguyen Ngoc Que and Wim Vijverberg, “Household assets and household income: Vietnam in 2002 and 2004.” Background Report, The World Bank, June 2006. Nguyen The Quan, “Labor skill in the VHLSS 2004.” Background Report, The World Bank, June 2006. Phung Duc Tung, “Importance of non-farm household business activity in the local economy: VHLSS 2004 and VHLSS 2002.” Background Report, The World Bank, March 2006. 84 Reardon, Thomas, J. Edward Taylor, Kostas Stamoulis, Peter Lanjouw, and Arsenio Balisacan. “Effects of non-farm employment on rural income inequality in developing countries: an investment perspective.” Journal of Agricultural Economics, May 2000, 51:2, 266-288. Van der Sluis, Justin, Mirjam Van Praag, and Wim P.M. Vijverberg, “Entrepreneurship Selection and Performance: A Meta-analysis of the Impact of Education in Less Developed Countries.” World Bank Economic Review, 2005 Vietnam Chamber of Commerce and Industry (VNCI) (2005). The Provincial Competitiveness Index of the Business Environment in Vietnam. Summary Report, May 2005. Vijverberg, Wim P.M., “Measuring the Investment Climate in Rural Areas.” Agriculture and Rural Development, The World Bank, August 2003. Vijverberg, Wim P.M., “Non-farm household enterprises: comparison of VHLSS and AHBS concepts.” Background Report, The World Bank, June 2006. Vijverberg, Wim P.M., and Jonathan Haughton. “Household Enterprises in Vietnam: Survival, Growth, and Living Standards.” In P. Glewwe, N. Agarwal, and D. Dollar, eds., Economic Growth, Poverty and Household Welfare in Vietnam, Ch. 4. (Washington, DC: The World Bank, 2004), 95-132. Vu Thi Kim Mao, “Non-farm household enterprises in VHLSS 2004: connections with business organizations and clubs, contacts with government agencies, sources of information, hurdles to operation and management.” Background Report, The World Bank, March 2006. World Bank. The Rural Investment Climate: It Differs and It Matters. Agriculture and Rural Development, June 2006. World Bank. Reaching the Rural Poor: An Updated Strategy for Rural Development. 2002. Zhao, Yaohui. “Labor migration and earning differences: the case of rural China.” Economic Development and Cultural Change, July 1999, 47:4, 767-782. Zhao, Yaohui. “Causes and consequences of return migration: recent evidence from China.” Journal of Comparative Economics, 2002, 30, 376-394. 85