Country Study Series '. , 56879 ,, , . l\lanufacturing in Kenya Under Adjustment Final Report on the Round n RPED Survey Department of Economics, Goteborg University Department of Economics, University of Nairobi May 1':'15 . . The views and interpretations expressed in this study are solely those of the authors. They do not necessarily represent the views of the \Vorld Bank or its member countries and . should not be attributed to the \Vorld Bank or its affiliated organizations. ·. .. Table of Contents Preface 1. Introduction 1 2. Productive Efficiency and Productivity Change 4 3. The Acquisition of Technology 36 4. The Manufacturing Labor Force 59 5. Financing and Investments 68 6. Infrastructure, Regulations and Business Support 128 7. Impact of Adjustment 157 8. Finn Growth 172 9. Concluding Remarks 179 References 182 . Preface This is the final report on the second round (wave IT) of a three round survey of Kenyan manufitcturing firms undertaken in Kenya within the World Bank's Regional Programme of Enterprise Development. The survey was carried out by the Economics Departments of Goteborg and Nairobi Universities in collaboration with the Kenya Association of Manufitcturers. Interviews were conducted by Peter Kimuyu, Gerrishon Ikiara, Mutsembi Manundu, Wafula Masai, Mohamud Jama, Njuguna Ndung'u, E.M. Kabuu, Simon Thiga, Anders Isaksson, Jorgen Levin, Lennart Hjalmarsson, and myself Data entry was done by Domisiano Mwabu Kirii and Geoffrey Opicha Okallo. The basic data analysis was done by Renato Aguilar. While the report presented here is the joint responsibility of the entire team, there has been some division of labor in writing. Lennart Hjalmarsson and Waher Ochoro wrote on production and market performance (Chapter 2). Wafula Masai and Haji Semboja analysed technology acquisition (Chapter 3). Renato Aguilar wrote about labor (Chapter 4). Clas Wihlborg and Anders lsaksson wrote on financial markets and investment (Chapter 5), while Genishon K Ikiara and Mutsembi Manundu wrote a background paper for this. Peter Kimuyu analysed infrastructure, regulations, and the business support system (Chapter 6), while Njugtma Ndungil and :myselfwrote on the impact of adjustment (Chapter 7). Renato Aguilar and myself wrote the, chapter on firm growth (Chapter 8). Simon Thiga provided information about changes in the business environment, while I coordinated and edited the work and wrote the remaining parts. Rick Wicks corrected our English. We would like to express our gratitude to the firms interviewed for their collaboration and patience: Goteborg, 22nd of May, 1995 AmeBigsten ii .. : .' . 1. INTRODUcrION By African standards Kenya has a relatively large manufacturing sector, but its share of GDP has increased very little over the last two decades. It has not been dynamic enough to function as an engine of growth for the whole economy, and it has not contnbuted much to foreign exchange earnings. The sector has been inward-oriented, with limited technological progress. The policy environment which evolved after independence was mainly conducive to manumcturing for import-substitution, while the refollD process which began in the 1980s has not generated any dramatic changes in the manufacturing sector so far. Investment response has been muted. The withdrawal of donor support in 1991 initiated a serious economic crisis for Kenya. Political turmoil and ethnic clashes before and after the elections in December 1992 also had serious repercussions on the economy. Uncertainty about government policies, as well as the shortage of foreign exchange, hampered economic growth. 1992 and 1993 were thus years of stagnation: GDP grew by just 0.5% and 0.1 %, respectively (Economic Survey, 1994). The manufacturing sector grew by only 1.2% and 1. 8% in those years, way below its growth trend. Manufacturing capital fonnation picked up somewhat in 1993, but still remained about 30% below the level reached in 1990. Thus, 1993, which is the reference period for the Wave n questionnaire, was one of the worst years in post-independence Kenya. At the time ofthe interviews in May and J1me 1994, the economy had recovered somewhat. The 1994 growth rate is estimated' at about 2% (EIU, 1994). Earlier reforms had initially focused an macroeconomic issues, but has since been broadened to include substantial structural and institutional reforms. Many of the issues we were concerned with in the first round of the survey had thus been changed since the project started. More ambitious economic reforms were now underway, and relations with donors were on the mend, ahhough many donors were still worried about the politics of the government. There were still many violations of basic democratic principles, with consequences also for investor confidence and thus for investment in the economy. Although there were some signs of stabilization and economic recovery at the time of the interviews, the economy remained in crisis, and short-term prospects were bleak. The overall aim of the project is to analyse factors that affect the efficiency and dynamics of business in Kenya. As noted in earlier reports, we concentrate on the following: 1) the structure of economic incentives and regulatory policies that may distort or constrain the efficient allocation of labor and capital; 2) the determinants of transaction costs for enteIprises, e.g., information Hows, property rights, legal structures, and infrastructure; 3) factors that may inlu"bit the acquisition of technological capability; and 4) the operation of business support services in finance, technical training, and market information. The primary sample for the second round of the survey was the 224 firms that were successfully interviewed in Round 1 (see last year's basic report for details of sampling procedures). Ofthose, 20 refused to be reinterviewed, and 11 had closed down, leaving us with 193 firms. We replaced some of the lost firms with new ones, giving us a new sample of214 firms: 51 in food, 58 in wood, 49 in textiles and garments, and 56 in metal. We did 137 interviews in Nairobi, 37 in Mombasa, 21 in Nakuru, and 19 in Eldoret. 156 firms may be characterized as formal, while 58 are informal The original sample selection was done on the basis of size, with firms stratified by employment levels. Firms that dropped out were replaced by firms as similar to the original ones as possible. The sample was chosen to be representative of the four sub-sectors, but not to be representative of the four towns covered by the survey. Inferences for the towns should therefore be avoided. In this report we analyze the second round of interviews, undertaken in May and June of 1994. 1 Chapter 2 provides an extensive discussion ofproductivity and productivity changes, 1 Earlier analyses from our Kenya surveys include a background paper (Departments of Economics, 1993a), two reports on Wave 1 (Departments of Economics, 1993b, 1994), and a basic data report from Wave 2 (Departments of Economics. 1995) 2 substantially extending the discussion in last year's report. Chapter 3 looks at technology acquisition by the firms, while Chapter 4 updates our analysis of the labor force. Chapter 5 analyzes the financial sector, while Chapter 6 evaluates problems relating to . infrastructure, regulations, and business support services. In Chapter 7 we discuss the impact of adjustment on the firms, while Chapter 8 analyzes firm growth. Chapter 9 summarizes and concludes. 3 2. PRODUCTIVE EFFICIENCY AND PRODUCTIVITY CHANGE 2.1 Introduction This chapter investigates the stmcture of technology in Kenyan manufacturing in 1994. In particu1ar, we focus on technical efficiency measures, the efficiency distn"bution of firms in the sample, and changes in productivity from 1992 to 1993. The scale and substitution properties of a traditional Cobb-Douglas production function are also evaluated. To fBci1itate and maintain comparability, the form and content of this chapter conform, to some extent, with the corresponding chapter in the 1993 survey. 2.2 Data As in the previous report, we have selected firms which had positive values for the measures of output, capital and labor. Table 2.1 presents 1993 summary statistics for output, capital and labor for the sample :firms in the food, metal, textile and wood industries. Output is measured in Kenya Shillings (Kshs), capital is measured by replacement value, and labor is the number of full-time equivalent persons employed in the :firm. Large standard deviations may reflect differences in production technologies among finns. 4 Table 2.1 Output, capital and labor, by sector, 1993 (in Kshs ,000) ... Minimum Maximum Mean SlaDdard dev. Food Output(Kdls) 22.85 4,414,600.0 199,s18.83 673,s83.09 (44fums) Capital (Kdls) 5.00 600.000.0 58,s97.73 113,770.27 Lab o x I XII>! J X /( X (Input) The input saving measure (i) shows what proportion of the observed input would have been necessary for the observed output, if the production unit in question had been at the efficient frontier. The output increasing measure (ii) shows how much output could have been produced with the observed input by a production unit at the efficient frontier. These measures are constructed so that efficient units will have a value of 1, while 8 inefficient ones will have less than 1. For example, a fIrm at point J in Figure 2.1 will be input saving efficient (under VRS), since X/Xl = 1. Production at point K will be inefficient both under input saving and output increasing measures, since XJIXK < 1 and YK/Y L < 1. With CRS technology, input saving efficiency coincides with gross scale efficiency (iii), as can be seen for point K in Figure 2.1, for instance, where both measures are calculated as X1IXK · Once input saving, output increasing and gross scale efficiencies are calculated, pure scale efficiency measures (input and output corrected) can be derived as in (iv) and (v) above. For a production unit i, scale properties are as follows: If Eli > ~i. the unit is performing with increasing returns to scale. If Eli < ~i, the unit is performing with decreasing returns to scale. If Eli = Eli, the unit is performing with constant returns to scale. 2.5 Measurement of Productivity The Malmquist (1953) (productivity) index was introduced into consumer theory as a ratio between two deflation factors (or proportional scaling factors) deflating two quantity vectors onto the boundary of a utility possibilities set. This deflation or distance function approach was later applied to the measurement of productivity in a general production function framework by Caves, Christensen, and Diewert (1982) and in a non- parametric DEA framework by Fiire, Grosskopf, Lindgren, and Roos (1989). Since the distance function is the inverse of Farrell's original measure of technical efficiency, and since Farrell technical efficiency is obtained from the DEA-approach, it is more convenient to use the Farrell measures directly, without reference to the 9 distance function. We will apply the same approach as Berg et a1. (1991) and Forsund (1991). The Malmquist productivity index for a production unit is the ratio between Farrell measures calculated relative to two different frontiers at two different points in time. The index can be decomposed, so that the change in total factor productivity (TFP) may be separated into the shift of the frontier (frontier technical change effect) and the shift relative to the frontier (catching-up-effect). Figure 2.2 illustrates the construction of the Malmquist index for a constant returns to scale frontier production function. In this case, both input saving and output increasing technical efficiencies coincide with gross scale efficiency. Figure 2.2 Construction of a Malmquist index (CRS) Output f I t+1 t+1 y t+1 ; ---;..-- - , t , _"_ .... ______ ,_J..._ I I I , ____ '_, I ": I I I I o A B C D E F Input P is a production unit observed twice, at t and t+ 1. Between these points in time, the 10 frontier has shifted from f t tD ft +1' At time t, the gross scale effiency of unit P measured against ft is p.t = OCtOD, where the frrst index t denotes the frontier year, and the second index t denotes the observation year. When measured against ft+t , gross scale efficiency is Et+l,t = OAtOD. Correspondingly, at time t + 1, the gross scale efficiency of P relative tD ft+ 1 is Et+1,t+1 = OB/OE, and when measured against f t · it is El.t+l = OFtOE. The Malmquist input-based productivity index (MJ, with frontier technology f as a reference base, is defined as E I,t+l Mt = - - Et,t and, with frontier technology ft+l as a reference base, it is defined as E 1+1,1+1 M 1+1 = --- E 1+1.1 We will concentrate on Mt , which measures productivity change relative to the base year 1. If Mt is greater than 1, productivity change has been positive. As mentioned earlier the Malmquist index can be decomposed into two components; shift of the frontier (Mfj , where i = t, t+ 1), and shift relative to the frontier (MC, the catching-up effect). The shift of the frontier between times t and t+ 1 is defined as ij = t,t+1 i*j The shift relative to the frontier production for unit P between times t and t + 1 is 11 DB DE E f+l.'+1 MC - -- = DC E,·t DD For fulJy efficient production in both years, Me = 1. The total index then measures only the shift of the frontier. 1 2.6 Estimation of Efficiency Scores In the case of non-parametric frontier production functions, total factor productivity (TFP) measurement based on the Malmquist index is a natural approach; the index requires neither profit maximization nor cost minimization, only quantity data. The calculation of the Malmquist index is based on the efficiency measures derived from the DEA-model. DEA is a linear programming (LP) technique for the construction of a non-parametric, piecewise linear convex hull to the observed set of output and input data; see Charnes and Cooper (1985) for a discussion of the methodology. The DEA approach defines a non- parametric frontier which may serve as a benchmark for efficiency measures. I Fare and Grosskopf (1990) define a revised Malmquist index as the geometrical mean of M, = E '+1.1+1 I E f,r+l E t,t I:' t.t '\ I:' ,+1.'+1 I:' 1+1,r The catching-up-index (shift relative to frontier) remains the same. but the shift of the frontier is then the geometrical mean of the distance between the frontiers for both obselVations. Le., MF, and MF'+I' This definition of the Malmquist index requires knowledge of the frontier technology in both years while, the definition we use only requires knowledge of the base year technology. 12 The estimated non-parametric frontiers serve as the basis for measurements of productive efficiency. The measures applied are Farrell type ray measures; for a generalization of Farrell's measure into input saving, output increasing and scale efficiency, see Forsund and Hjalmarsson (1974), (1979) and (1987). In the DEA approach, Farrell efficiency measures can be defined in the usual way. Farrell provided a methodology by which technical efficiency could be measured against an efficiency frontier, assuming constant returns to scale. The production set obtained is represented by a convex set of facets, i. e. the production frontier obtained is the boundary of the free disposal convex hull of the data set. Farrell's original approach, of computing the efficiency frontier as a convex hull in the input coefficient space, was reformulated by Chames et al (1978) into calculating the individual input saving efficiency measures by solving an LP-problem for each unit under the assumption of constant returns to scale. Computation of efficiency scores Compared to Farrell's approach, DEA offers a more operational framework for the estimation of efficiency; efficiency is calculated separately and directly for each production unit in turn, while at the same time the location of the corresponding linear facet is determined. Calculating efficiency measures as defined above is trivial as long as the production activity consists of only one input and one output. In order to handle more than one input/output, it has been shown, notably by Charnes, Cooper and Rhodes (1978) that a linear programming (LP) problem can be solved for one unit at a time. The input saving measure under VRS is found by solving the following LP-problem for each production unit, k, with output Ytand input Xt, where At is a vector of weights, Ale;, which determine the reference point: 13 (2) subject to the following restrictions: (2a) (2b) (2c) (2d) where m is the number of outputs, n is the number of inputs and N the number of units. Restriction (2a) means that the reference unit lllust produce at least as much output as unit k, while restriction (2b) implies that the efficiency adjusted volume of input used by unit k must at least amount to the input volume used by the reference unit. Restriction (2c) is the condition for VRS. If this restriction is omitted, eRS is implied. E\ and ~ then coincide, and both coincide with ~ in the case of VRS. ** Ar detta stycke ok?*** When efficiency is measured along a ray from the origin, a micro unit may seem to be fully efficient, although it is not fully efficient in the sense that it is dominated by another unit (regardless of assumptions about scale property). In empirical applications, this can. be controlled for by inspection of the slack variables. The output increasing efficiency measure is found by restricting the reference point on the unknown frontier to the same amount of input(s) as observed for unit k, and solving the following LP-problem: 14 .' (3) El Yrt ~ "f::; AtiYrJ' r =1, ···· m (3a) 2t (3b) (3c) AtlO, j=l ·...· N (3d) where fa is the output increasing efficiency measure for unit k. The rest of the variables are defined as in (2)-(2d). Restriction (3a) means that the efficiency-corrected volume of output «lf~ Yde) must be less than or equal to the amount of output produced by the reference unit. Restriction (3b) means that unit k must use at least as much inputs as the reference unit for the same amount of output. Restrictions (3c) is, again, the condition for VRS. The observed outputs of unit k will now be efficiency-adjusted proportionally upwards to be less than or equal to output at the frontier reference point. where at least one output equals the reference point. Since scale inefficiency is due either to decreasing or to increasing returns to scale, one can easily determine the case for the E, calculations by inspecting the sum of weights, S, (4) with CRs technology. If this sum is less than one, we have decreasing returns to scale (both at K and at the adjusted point J at the VRS frontier, see Fig. 2.1), and if it is larger than one we have increasing returns to scale. Some caution is in order concerning DEA as a technique for efficiency measurement. Since DEA yields relative efficiency measures and defines a production unit (in this case 15 a truck) as inefficient relative to another unit by comparing combinations of inputs and outputs, units operating with input-output quantities sufficiently far from the other units at either end of the size distribution may be identified as efficient due to the lack of comparable units. However, problems of this kind are minimal if the sample size is large relative to the numbers of inputs and outputs. This is because larger samples tend to . decrease the average level of efficiency, due to the positive probability of including more efficient outliers in the sample. 2.7 Cobb-Douglas Results Tables 2.3a and 2.3b show the OLS parameter estimates of the unrestricted (NCRS) and the constant returns to scale (CRS) versions of the Cobb-Douglas model The elasticities of output with respect to the two inputs are positive and significant in both the unrestricted and the restricted models. The tests for constant returns to scale (Table 2.3a) reject this hypothesis at the 1% level of significance in the food, metal, and textile sectors. 0n1y in the wood sector is the CRS hypothesis supported. In the unrestricted model, the responsiveness of output to labor is greater and more significant than the responsiveness of output to capital, for the food processing, metal, and textile sectors. The reverse is the case in the wood sector. This is evidence of :fairly large increasing returns to scale in the food, metal, and textile sectors. There would seem to be some advantage in promoting firm growth for greater productivity in these sectors. The high level of diversity within the various sectors, howev~r, recommends caution. And the wood sector does not offer much prospect from scale increases. 16 Table lola OLS parameter estimates in the case of unrestricted technology, by sector \.mI'lSIliacd CO Food Metal Textiles Wood Non-CRS N 44 45 42 48 Log A 3.37 1.47 1.43 0.78 (3.68) ( .63) (1.95) (0.69) A 2362.17 29.53 26.75 6.08 Elasticity w.r. t 0.13 0.29 0.40 0.65 K (0.77) (2.55) (2.44) (3.01) ( u ) Elasticity w.r.t 1.40 1.30 0.96 0.46 L (4.41) (5.16) (3.87) (1.24) 0' ) Returns to scale 1.54 1.60 1.37 1.11 AdjR-sq. 0.66 0.79 0.79 0.42 Test results for the restriction u + 6 = 1 Numerator: 18.85 27.04 16.99 1.30 OF: 1 1 1 1 F Value: 7.49** 12.16** 7.74** 0.17 Denominator: 2.52 2.22 2.20 7.58 OF: 41 42 39 45 Prnh > p. o OO() (} 001 o (}O~ () "R ~: T-statistics are in brackets. ** indicates rejection at the 1% level of significance. 17 Table 2.3b Ol.S parameter estimates in the case of CRS technology, by sector Restricted CD Food Metal Textiles Wood CRS N 44 45 42 48 Log A 4.08 2.11 1.70 1.01 (4.32) (3.56) (2.1~ ~1.0~ A 12105.12 129.53 49.60 10.30 Elasticity w.r. t 0.36 0.51 0.59 0.67 K (2.16) (4.68) (3.59) (3.24) Elasticity w.r.t 0.64 0.49 0.41 0.33 L (3.86) (4.4~ (2.5~ J1.6~ Adi "R.,:n o til 074 0.7ti 0.43 Note: T-statistics in brackets Table 2.3c compares the results with those from last year. In some cases the changes are large and it is obvious that the input elasticities are sensitive to the sample selected. In the case offood and metal there is a decrease in capital elasticity and a large increase in labor elasticity in both sectors. There are strong increasing returns to labor in 1993. In the case of textiles. changes from last year are fairly small. while in the case of wood the changes are reversed, with a large increase in capital elasticity and a large decrease in labor elasticity. However, in all sectors the total scale elasticity exceeds one in both years. 18 Table 2.3c A comparison of input elasticities between 1992 and 1993, NCRS, by sector Food Metal Textiles Wood Elasticity of capital 1992 0.67 0.37 0.35 0.08 1993 0.13 0.29 0.40 0.65 Elasticity of labor 1992 0.41 0.92 0.S2 1.37 1993 lAO 1.30 0.96 0.46 Elasticity of scale 1992 LOS 1.29 1.17 1.45 1993 1.54 1.60 1.37 1.11 2.8 Frontier Results Table 2.4 shows the mean efficiency scores. In terms of input saving (El) or output increasing (E2) efficiency, food shows the lowest mean values, while metal shows the highest. The potential for increasing output, given observed input levels, is higher than the potential for saving input, given observed output levels. This result implies that most firms are below optimal scale, ie., they are located in ranges of increasing returns to scale. Consistent with this observation, gross scale efficiency (E3) is very low, with about equal values for food, textiles and wood. Here, metal has a much higher value than the other sectors. A comparison between gross scale efficiency and pure scale efficiencies (E4 and E5) reveals the re1a$e impact of technical innefficiency and scale inefficiency. Removing all technical inefficiency still leaves the input-corrected efficiency values (E4) at fairly low levels, while the impact is much larger on the output-corrected values (E5). This is a good illustration of the impact on the efficiency results of the scale properties of the production functio~. Values of S less than one indicate suboptimal average firm size, a value of one would 19 indicate optimal size; and a value larger than one would indicate super-optimal size. Table 2.4 Mean efficiency scores Measure Food Textiles Wood Metal El 0.26 0.36 0.30 0.49 E2 0.19 0.20 0.23 0.44 E3 0.16 0.15 0.16 0.35 E4 0.53 0.49 0.53 0.66 E5 0.89 0.81 0.82 0.91 S 0.13 0.09 0.14 0.92 Efficiency distributions are shown in Figures 2.3 to 2.6. In addition, the distnbution of the Scale index, S, is also shown. Production units are ranked on each figure in ascending order of efficiency. A specific unit may have different rankings for different measures, so direct comparison of individual units across the figures is not pOSSIble. Food The general impression given by the efficiency distributions is of extreme skewness, with a few fully efficient and rather efficient units, and a large tail of units with very low efficiency scores. Skewness is most extreme in output-increasing and gross scale efficiencies. As Figures 2.3c-f reveal, in spite of a very low gross scale efficiency and an extremely skewed scale index distnbution, with almost all firms of suboptimal size, elimination of technical inefficiency would also eliminate most of the overall inefficiency. This is especially true of output-corrected pure scale efficiency (£5), in which case not much inefficiency remains attributable to scale; while with input-corrected pure scale efficiency (E4) a large share of inefficiency remains even after elimination of technical inefficiency. 20 Figure 2.3a-c El - E3 Input saving efficiency: Food 1993 1 0.9 0.8 0.7 - w 0.6 0.5 0..4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 ':Sl 39 41 Unit Output augmenting efficiency: Food 1993 1 0.9 0.8 0.7 0.6 N w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 ':Sl 39 41 Unit Scale efficiency: Food 1993 1 0.9 0.8 0.7 0.6 M w 0.5 0..4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 1921 23 25 27 29 31 33 35 37 39 41 Unit 21 FIgure 2.3d-f E4. E5 and S Pure scale efficiency: Food 1993 1 0.9 0.8 0.7 0.6 .., w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 Unit Pure scale efficiency: Food 1993 1 0.9 0.8 0.7 0.6 It) w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 1921 23 25 27 2931 33 35 37 39 41 Unit Scale indicator: Food 1993 1 ,------------------------------------ 0.8 +---------------------- 0.6 +------------------- 0.4 + - - - - - - - - - - - - - - - - - - 0.2 + - - - - - - - - - - - - - - - - = 1 4 7 10 13 16 19 22 25 28 31 34 37 40 Unit 22 Textiles The textile efficiency distributions, shown in Figures 2.4a-£: are fairly similar to those of the food sector. Input-saving efficiency is higher, while output-increasing and scale efficiency levels are about the same. The textile industry has the most extreme size distnoution, with only three firms of optimal size, and an extensive, extremely thin tail Wood The efficiency distributions ofthe wood industry, shown in Figures 2.5a-£: are very similar to those of textiles. The gross scale efficiency distnoution has a somewhat thicker part at the right end, and the scale indicator is somewhat less extreme, even though there is still an extensive, very thin tail. Only one firm is of super-optimal size. Metal The metal industry, shown in Figures 2.6a-£: shows a much higher efficiency level than the other sectors, thick on the right with fiUrly fat tails. Moreover, the size distnoution of firms is radically different: About 15 firms are of optimal size or larger, though, the tail is still very thin. 23 FJgUre 2.4a--c Input· saving efficiency: Textile 1993 0.9 0.8 0.7 0.6 ..- 0.5 w OA 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit Output augmenting efficiency: Textile 1993 1 0.9 0.8 0.7 0.6 ~ 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit Scale efficiency: Textile 1993 1 0.9 0.8 0.7 0.6 CO\! w 0.5 0.04 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit 24 FJgUre 2.4d-f Pure scale efficiency: Textile 1993 . 1 0.9 0.8 0.7 0.6 i%i 0.5 A... 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit Pure scale efficiency: Textile 1993 0.9 0.8 0.7 0.6 It) w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit Scale indicator: Textile 1993 1 0.9 0.8 0.7 0.6 tI) 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Unit 25 " Figure 2.58-C. Input saving efficiency: Wood 1993 1 0.9 0.8 0.7 0.6 ... w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 Unit output augmenting efficiency: Wood 1993 1 0.9 0.8 0.7 0.6 '" w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 1315171921 2325 2729 3133 35 37 39 4143 45 47 UnIt Scale efficiency: Wood 1993 1 0.9 0.8 0.7 0.6 C') w 0.5 0.4 0.3 0.2 0.1 0 13 57911131517192123252729313335373941434547 Unit 26 Figure 2.Sd-f. Pure scale efficiency: Wood 1993 1 0.9 0.8 0.7 0.6 IZ 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 111315171921232527 29313335 37394143 4547 Unit Pure scale efficiency: Wood 1993 0.9 0.8 0.7 0.6 an UJ 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 1315171921 23 25 27 29 3133 35 37394143 45 47 Unit Scale indicator: Wood 1993 1.2 -r-----------------------, 0.8 + - - - - - - - - - - - - - - - - - - - - U) 0.6 + - - - - - - - - - - - - - - - - - - - - - 0.4 +--------------------- 0.2 + - - - - - - - - - - - - - - - - - - - - = 1 3 5 7 9 11 13 15 171921 23 25 27 29 31 33 35 373941 43 45 47 Unit 27 Figure 2.6a-c. Input saving efficiency: Metal 1993 0.9 0.8 0.7 0.6 ... w 0.5 0." 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 2729 31 33 35 3739 ..1 .ot3 Unit Output augmenting efficiency: Metal 1993 1 0.9 0.8 0.7 0.6 ~ 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 ..1 .ot3 Unit Scale efficiency: Metal 1993 1 0.9 0.8 0.7 0.6 C"l w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 .ot3 Unit 28 FIgUre 2.6d-f. Pure scale efficiency: Metal 1993 0.9 0.8 0.7 0.6 iZ 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 1921 23 25 2729 31 33 35 3739 41 43 Unit Pure scale efficiency: Metal 1993 0.9 0.8 0.7 0.6 It) w 0.5 0.4 0.3 0.2 0.1 0 1 3 5 7 9 11 13 15 17 1921 23 25 2729 31 33 35 3739 41 43 Unit Scale indicator: Metal 1993 8T 7 6 5 tn 4 3 2 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 Unit 29 2.9 Productivity Change The rate of total factor productivity change is measured by the Malmquist index and decomposed into the catching up, Me, and frontier shift, MF, components. A value greater than one indicates a productivity increase while a value less than one indicates productivity regress. Malmquist results are shown in Figures 2.7 to 2.10. The overall Malmquist index is given in ascending order. Each production unit has a unique identifying number, so that the same number refers to the same firm in the corresponding catching up and frontier shift diagrams. Since the panels are not complete the number of comparable finns vary between the industries. Food The most succesful firms in the food industry have experienced a rapid rate of productivity growth. However, 10 of 16 firms had negative productivity growth and, moreover, at a rather "high" rate. There has been a rapid movement of the frontier outwards for most units, Therefore, it is not surprising that average efficiency is faDing. For most finns, the catching up index is less than one, ie., most firms are lagging behind the frontier movement. Overall, the results indicate productivity regress for most finns. Textiles The distribution of the overall Malmquist index for textiles is fairly similar to that of the food industry, but frontier progress is much slower and more uniform; note the difference in scale on the vertical axis. The catching up effect is also more important here than in the food sector. Wood " The wood industry has experienced the highest rate of productivity growth of the four sectors. All firms in the panel had a very large frontier shift and, as would be expected, the catching up effect is less than one for almost all firms. 30 .' Figure 2.78-<:'. .' Malmquist index: Food 1992193 3.5 3 2.5 .. :::E 2 1.5 1 0.5 - - -. . o - ···· I I I II 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Units Malmquist catching up index: Food 1992193 1.4 1.2 1 0 0.8 :::E 0.6 0.4 I 0.2 0 · I I · 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Units Malmquist frontier shift index: Food 1992193 5 4 u.. 3 :::E 2 1 o I ··I I · · 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Units 31 Figure 2.8a-c. Malmquist index: Textile 1992193 2.5 2 1.5 :E 1 0.5 · 0 ·. 11.11.111.1.111.1 1 3 5 7 9 11 13 15 17 19 21 Units Malmquist catching up index: Textile 1992193 2.5 2 o 1.5 :E 1 III III .... II 0.5 o 1.,III,a I 1,1,1,1,1 1 3 5 7 9 11 13 15 17 19 21 Units Malmquist frontier index: Textile 1992/93 1.6 1.4 1.2 1 u.. :E 0.8 0.6 0.4 0.2 ... ,1,1 0 1 3 5 7 9 11 13 15 17 19 21 Units 32 Figure 2.9a-c. Malmquist Index: Wood 1992193 5 4 3 :IE 2 1 0 1 3 5 7 9 11 13 15 17 Units Malmquist catching up index: Wood 1992193 1.4 1.2 1 (J 0.8 :IE O.S 0.4 0.2 .111.11 0 1.1 I 1 I II I I I 1 3 5 7 9 11 13 15 17 Units Malmquist frontier shift index: Wood 1992193 5 4 LL 3 :e 2 1 0 1 3 5 7 9 11 13 15 17 Units 33 Figure 2.10a<. Malmquist index: Metal 1992193 3 2.5 2 :i: 1.5 1 0.5 0 1 3 5 7 9 11 13 15 17 Units Malmquist catching up index: Metal 1992193 2.5 2 U 1.5 :i: 1 0.5 · ,I ,I o II 1 ·· I 3 5 7 9 11 13 15 17 Units Malmquist frontier shift index: Metal 1992193 3 2.5 2 LI.. :i: 1.5 1 JIll _iii 0.5 o I III ,II I I 1 3 5 7 9 11 13 15 17 Units 34 Metal Behind the facade of an evenly sloping productivity change distnoution, there is a mixed pattern, of progressive catching up components partly neutralised by regressive frontier shifts. Overall, only 6 of 18 firms have realised a positive productivity growth. 2.10 Conclusions In this chapter we have investigated a few aspects of industrial structure, technology, efficiency and productivity change in Kenyan manufacturing. Compared to more industrialised economies, a major difference in industrial structure is the large range of simultaneously existing technologies with substantial differences in productive efficiency. The large spread between best-practice and worst-practice creates a huge potential for input saving or output increasing rationalisation measures. To some extent, the realisation of this productivity potential also requires an increase in firm sizes, although the most important potential is in adoption of best-practice technology at observed output levels. Since the size distribution is very skewed, and so are the productive efficiency distributions one should expect a fiUrly low competitive pressure. Therefore, market power should be a problem One would expect that an increased exposure to interntional competition will decrease the existing spread in the efficiency distnoutions. 35 3. THE ACQUISITION OF TECHNOLOGY 3.1 Introduction Like other Sub-Saharan countries, Kenya is moving toward a more liberal economic regime, with an increase in competition, particularly for private manufacturing firms. However, the prospects for substantial improvements in enterprise dynamics and overall economic welfare depend upon successful exploitation of more efficient technologies. The acquisition of new technology in Kenya has been a gradual and little coordinated or regulated process, largely dependent on direct foreign investment activities (Economics Departments, 1994). This chapter will examine some of the means of technology acquisition, such as foreign licenses, technical assistance and the employment of expatriates. Then we analyze the capability stock (managers' knowledge and skills). Finally, we provide an econometric analysis of the acquisition offoreign technology. 3.2 The Nature of Technology Acquisition 3.2.1 Foreign Licenses Ucense agreements cover a wide range of areas, such as know~how for the implementation of productive processes, the granting of rights for (exclusive) patent exploitation, technical assistance, blu~prints, and the use of brand names. Our sample included only 15 manufactming firms holding foreigiJ. technical licenses. In comparison to last year, two firms had droPped and three firms had added foreign technical licenses. Of the 15 current foreign licenses, 12 operate in the 'low technology~traditiona1' industrial sectors (5 in food, 4 in wood, and 3 in textiles). and only 3 in the 'non~traditional' industrial sector (metal products). Thus, foreign licensing is not currently a common means of obtaining industrial technology. 36 Since most manufilcturing is in the traditional and/or consumer goods sectors, it may be that these enterprises have built up a competitive local position with relatively simple technologies. Thus, it may be that Kenyan firms are not importing technology via licensing agreements to any great extent, because their relatively unsophisticated technology is adequate for their purposes. Some licensing agreements are concluded as part of joint venture agreements between domestic enterprises and foreign technology suppliers. A study by Seyoum (1990) on technology licensing in East Africa showed that many such licensing agreements fail to provide for improvement of design and engineering capability. This may limit the ability of local licensees to adapt the technology to local conditions and to innovative in other ways. The limited use of foreign licenses may also be attnouted to the relatively low level of foreign ownership, contractual provisions inhibiting technological self-reliance, and investment structures. 3.2.2 Technical Assistance and Expatriates Provision of technical assistance and services is one of the major components oftechnology licensing agreements in East African, and many firms are aware of its potential importance as a source offoreign technology. It is important because the transformation from being a low- productivity firm to being an export-oriented firm directly linked to the ability to upgrade technology from well-established sources. 20 firms (9 in metal, 5 in textiles, 3 in food, and 3 in wood) received technical assistance, including the majority of:firms with foreign licenses. Some technical assistance personnel are part o( foreign aid packages. Foreign professional expertise may be utilized to administer the firm, to establish training programs, to provide design and engineering services or marketing and financial services. 44 firms (14 in food, 8 in textiles, 8 in wood and 14 in metal) received technical assistance. Table 3.1 shows that a rather small number of firms (only 21) had expatriates in managerial 37 posts, only 19 firms engaged expatriates as technicians, and only 4 employed expatriates in other operative capacities. This may suggest that Kenyan firms do not perceive administrative and operative-skill constraints as serious. Still, many of the metal firms engaged expatriates in technical positions. Last year we noted that the hiring of expatriates is influenced by costs, government labor policies, expatriate qualifications and their ability to train local staff: and technological constraints. Coughlin and Ikiara (1991) found that changes in the number of expatriates in textiles resulted partly because firms invested in entirely new processes, for which they could not immediately obtain properly skilled personnel locally. Contractual arrangements are usually for a short period (one to five years) with provision for renewal in case it is considered necessary by the local firm Expatriate contracts are not renewed when there is effective transfer of technical skills and know-how to the local enterprises or when suitable local personnel becomes available. However, Tostenson and Scott (1987) argue that the transfer of technical know-how to Kenyan counterparts and colleagues remains generally unsatisfactory. Table 3.1 Number of expatriates in different positions Managerial Technical Operative Other capacities Food 7 6 1 Textiles 6 2 Wood 3 4 1 Metal 5 7 1 1 ALL 21 19 1 3 3.2.3 Skills of General Managers and Production Manager Skills Capability stock refers to managerial and technical knowledge and skills. The quality of knowledge and skiIIs available to a :fum is a critical factor influencing its performance. High level managerial, administrative and technical personnel are inevitably needed to modernize and to carry industrialization forward. The rate of modernization is associated with the 38 capability stock and its rate of accumulation. Not all firms in the sample had a general manager, a production manager or a plant manager. Generally, smaI1 firms, in both the formal and informal sectors, have only the owner-manager performing most management functions, probably assisted by a supervisor at the production level Larger and more complex firms have organizational systems with hierarchical levels of authority and respollSloility, and relationships between different departments. 150 firms had general managers, and 48 (about one quarter) had production managers. Table 3.2a shows the levels of education of general managers and of production or plant managers. 21% of the general managers had primary education or less, 48% secondary education, 14% had non technical university education, 12% technical university education, and about 5% had post-graduate education, with 4% having studied abroad. The metal sector has the best educated general managers, and the larger firms have a higher proportion of highly educated managers as one would expect. The level of education of production managers is often high. 37% of production managers had technical university education, 27% had secondary education, and 14% had post-graduate education level 39 Table 3.2a Education level of general managers and production managers - - . - .. ~.~ ~- ... - ~ .. -.--.~ .... - NONE PRIMARY SECONDARY UNIVERSITY-NT UNIVERSITY-Tech POSTGRAD-Kmya POSTGRAD-Abroad N 0/.. hi O/A hi 0/.. ... IlL ~ _0/.. hi ~ N 0/.. GENERAL MANAGER Food I 2.6 4 10.3 22 56.4 6 15.4 5 12.8 1 2.6 Wood 3 7.9 11 28.9 15 39.5 5 13.2 3 7.9 1 Textiles 5 15.2 20 60.6 4 12.1 2 6.1 I 2.6 1 3.0 Mital 7 17.5 IS 37.S 6 15.1 8 20.0 3.0 4 10.0 ALL 4 2.7 27 18.0 72 48.0 21 14.0 18 12.0 2 1.3 6 4.0 PRODUCTION MANAGER. Food 1 6.7 4 26.7 3 20.0 S 33.3 1 13.3 Wood 1 9.1 2 18.2 6 S4.5 2 18.2 Textiles 2 16.7 2 16.7 S 41.7 1 8.3 1 .16.7 Mital 1 9.1 2 \8.2 6 54..5 2 18.2 ALL 3 6.1 3 6.1 13 26.,5 ,5 10.2 18 36.7 1.00 2.0 6 12.1 40 Regarding general managers' education levels there are important differences between informal and formal firms. About 49% of the informal firms' general managers have only primary education, while 43% have secondary education, and 5% have university education. Of the formal firms' general managers, 8% have only primary education, while 50% have secondary education, and about 40% have university education. General managers who attended postgraduate studies abroad are found only in formal firms. General managers and production managers have almost equally long average experience in the business (see Table 3.2b). The average number of years of service in the present business was 9.5 for general managers and 10.8 for production managers. The general managers in wood were by far the most experienced, with an average of 14.6 years in the business. The experience of managers in the textile sector was on the average about half of this, perhaps because many wood firms were established earlier than most textile firms. Wtth regard to production or plant managers, those in textiles had the longest experience, averaging more than 15.4 years, followed by those in food with about 9.6 years. Formal firms had the longest serving general managers, with an average of 16.4 years, while the average period in informal firms was about 9.0 years. Table 3.Zb Average years of general managers and production managers in the present business Food Wood Textiles Metal All N Years N Years N Years N Years N Years General 38 4.6 39 14.6 33 7.5 40 9.2 150 9.5 managers Production 15 9.6 10 7.3 11 15.4 12 5.9 48 10.8 managers The Kenyanization campaign for public sector management called for stepped-up training for public setVants. But the training policy did not give adequate consideration to the needs of the private sector, where indigenous managerial expertise to run commercially viable industrial and commercial enterprises continues to be in short supply. 41 9% of the general managers in our sample, and 22% of the production managers, came ftom abroad (see Table 3.2c). The proportion varies with the size ofthe :firm. There were no foreign production managers in the wood working finns. There was only one foreign general manager in the informal finns. The percentage of Kenyan managers in formal finns was higher than last year. However, there is still limited access for Kenyans to managerial positions in finns owned by non*Kenyans, which may be a matter of lack of trust or experience. Table 3.2c Origin of general managers and production managers GENERAL MANAGER PRODUCTION MANAGER KENYA OTHER KENYA OTHER N OJ.. N % N % N % Food 38 97.4 1 2.6 12 80.0 3 20.0 Wood 35 89.7 4 10.3 10 100.0 Textiles 29 87.5 4 12.1 9 75.0 3 25.0 Metal 35 87.5 5 12.5 7 58.3 5 41.7 ALL 137 90.7 14 9.3 38 77.6 11 22.4 3.3 A Probit Model of Technology Acquisition We use a probit model to investigate how various factors relate to the probability of an enterprise acquiring foreign technology. The theoretical foundation is the analysis of acquisition of technology presente~ in LaIl (1984 and 1987), Dahlman and Cortes (1984) and Tei~el (1993a and 1993b). Last year's analysis suffered ftom some problems of interpretation. This year we refine the analysis. Like last year, we set up a probit model to analyze whether an enterprise attains the searching level in the acquisition offoreign technology. We assume that this occurs via the imports of machinery with embodied technology, foreign direct investment, technical and management consultancy, and through the use of experts or licensing. We investigate 42 whether the probability of acquiring foreign technology defined like this is influenced by variables such as (1 ) the fum's size, organization and peIformance, (2) the skill require-me- nts of the foreign technology, (3) the distance of the existing technological base from the technology frontier, (4) market structure, (5) direct policy and regulatory incentives, and (6) institutional arrangements. The dependent variable in this model is a single dichotomous variable (Y). The independent variables (X) consist of continuous, dummy, sequential categorical and non-categorical variables. This situation does not allow us to employ classical regression analysis without estimation and interpretation problems (see Maddala, 1983). We therefore construct a binary qualitative response mode. Note that this model asSumes that all firms evaluate the explanatory factors identically. In principle, any proper continuous probability distn'bution defined over the real line will suffice. The choice ofthe distn'bution function determines the type of analysis. We assume a normal distribution and thus use a probit mode~ which is the simplest and most commonly encountered binary choice model. There were 214 observations (firms). This is enough to carry out a successful probit analysis on the acquisition oftechnology based on the above model. All our binary variables were coded as 0 for YES and I for OTI:lERWISE before running the SAS Probit procedure. The SAS Probit equation is defined as: Prob(Y=O) = - F(.X'P). The analysis uses labor weights, which gives each employee in any fum the same probability of being included in the sample. In the maximum likelihood estimation, the terms in the likelihood function and its derivatives, not the data values themselves, are multiplied by the weighting variable. 43 3.4 Defmitions of Variables The variables included in the model are defined and constructed in the following way: A Endogenous Variable: A1 Acquisition of technology: ATXO = 0 ifthe fum has conducted any form of AT searching activities, or = 1 otherwise. B Exogenous Variables: B.l Organizational and ownership structure: SUBS= 0 ifthe fum is a subsidiary of a domestic fum or holding company, or = 1 otherwise. MNC = 0 if the fum is a subsidiary of a multinational coIporation, or = 1 otherwise. LOCC = 0 if the fum is neither SUBS nor MNC, or = 1 otherwise. B.2 Output, capital-labor ratio, energy and management inputs: OUTPUT is the value of the fum's output in 1993 in Kenyan Shillings. KL RATIO is the capital labor ratio. MANRATIO is the ratio of number of employee in the management to total labor. 'ENERGY is the total costs of a fum for the use of energy. , B.3 Market structures: DDCON = 0 ifthe finn faced domestic demand constraints, or = 1 otherwise. 44 .' COMPI = 0 if the firm had competition from imports, or = 1 otherwise. COMPD = 0 ifthe firm had competition nom local firms, or . = 1 otherwise. EXPO = 0 if the firm exports, or = 1 otherwise B.4 Technology and investment regulations: TECREG = 0 if restrictions on payment of technology licenses and royalties were important for the operation of a firm, or = 1 otherwise. INVREG =0 if restrictions on capital requirements, access to domestic finance, and approval of foreign loans are important for the operation of a firm, or = 1 otherwise. B.5 Technical and business support institutions: BSS 1 = 0 ifthe fum received any training assistance or information from government programs or business support institutions, or = 1 otherwise. BSS3 ::; 0 ifthe fum received any technological assistance or information from government programs or business supporting institutions, or = I otherwise. B. 6 Skill requirements: GMEDU = the highest completed level of education of the general manager (1 None, 2 Primary, 3 Secondary, 4 University (general), 5 University (technical), 6 Postgraduate (domestic), 7 Postgraduate (abroad) 45 PMEDU = the highest completed level of education of the production manager (1 None, 2 Primary, 3 Secondary, 4 University (general), 5 University (technical), 6 Postgraduate (domestic), 7 Postgraduate (abroad) GMEXP = the experience of the general manager, ie., number of years in the present business. PMEXP = the experience of the production manager, ie., number of years in the present business. B.8 Nature of products produced: SEC} = 0 ifthe firm belongs to sector 1, ie., food, or = 1 otherwise. SEC2 = 0 if the firm belongs to sector 2, ie., textiles, or = } otherwise. SEC3 = 0 if the firm belongs to sector 3, ie., wood, or = } otherwise. SEC4 = 0 ifthe firm belongs to sector 4, ie., metal, or = 1 otherwise. .. 46 Table 3.3 Probit model summary statistics VAllTART.F MTNTMlJM MAYTMlTM MEAN" !':TnnFV OlITPUT 0 4414600000 65188017.61 318600915 MANRATIO 0 1 0.134 0.186 GMEXP 0 50 10.150 11.072 PMEXP 0 32 2.414 5.779 PMEDU 0 7 0.963 1.930 ATXO 0 1 0.491 0.501 GMDU 0 7 2.346 1.857 BSSI 0 1 0.911 0.285 BSS3 0 1 0.972 0.166 DDCON 0 1 0.682 0.467 COMPI 0 1 0.509 0.501 LOCC 0 1 0.075 0.264 COMPD 0 1 0.808 0.395 KL RATIO 0 5250000 736342.41 1023593.92 SEC! 0 1 0.762 0.427 SEC3 0 1 0.771 0.421 SEC4 0 1 0.738 0.441 EXPO 0 1 0.234 0.424 INVREG 0 1 0.893 0.310 ENERGY 0 39025400 1168787.25 3663837.55 3.5 Test Statistics We utilize the Pearson Chi-Square test statistic (PCS) to get a rough idea of how well our prohit model performs. At the 5 % level. the theoretical value for the Pearson Chi-Square test statistic is 232. The computed PCS is 161. Since the computed value is less than the theoretical one, the fit of our model seems good. We also note that the Log-Likelihood Value of the model is -338. 47 ·. Table 3.4 Goodness-of-Fit tests and Log-Likelihood values Value OF Pearson Chi-Square 161.5 194 L.R Chi-Square 169.8 194 Log-Likelihood Value - 338.2 194 Several test statistics such as t-ratios, the signs of parameters, and Log-Likelihood Value (lLV) were used to investigate the importance of each explanatory variable: Some possible variables were excluded since they did not perform well in preliminary estimation rounds. Though it seems reasonable that the mean of the error terms is zero, the variance of the error terms associated with large :firms is larger than that of small firms. This may happen because small :firms are unlikely to engage in extensive AT, while large :firms have the leverage (assets, liquidity, economies of scale) to more extensively use this competitive strategy. We used a Lagrangian Multiplier Test to test models with and without heteroskedasticity correction: Heteroskedasticity does not seem to be a problem in modeling factors determining AT in Kenyan manufacturing with the above mentioned maximum likelihood estimation method. 3.6 Maximum Likelihood Parameter Estimates It has been noted (Green, 1992, Davidson and Mackinnon, 1993) that the direct parameter estimates ofthe above probit model are not necessarily the marginal effects, vvhich we are often interested in. In this case the marginal effect of a continuous explanatory variable (Xc) is given by product of the pparameter of the respective variable and the standard normal density function. The result yields the derivative of the probabilities with respect to a particular independent variable, ie., the marginal effect of that variable. In the case of a dummy variable (Xd), the partial derivatives or marginal effects are known 48 to be less meaningfuL But it is still important to analyze the effect of the dummy variable (Xd) on the whole distribution by computing Prob(Y=l) over the range ofptX (using sample estimates) and with two values of the binary variable. This is computed as the difference between two probability functions. ie., Prob(Y=I) when dummy variable Xd=l and Prob(Y= 1) when Xd =0. In this sense the marginal effect measures the effect of a dummy variable on predicted probabilities. It is obvious that these marginal effects vary with the vames ofX For simplicity, in interpreting the estimated model we will use the means of the regressors and their relevant standard errors. Table 3.5 gives the maximum likelihood direct coefficient estimates, standard errors, t- values and their corresponding Chi-Square statistics. The significance and signs for most ofthe coefficients are as expected. Only two coefficients (out of21) are found insignificant at the 5% leve~ the technology and information support services variable (BSS3), and the intercept term. The output (OUTPUT), capital-labor ratio (KL_RATIO), energy use (ENERGY), general manager education (GMEDU), and experience (GMEXP), production manager experience (PMEXP), organization status (LOCC), export (EXPO), market competition problems (COMPD and COMPI), and sector types (SEC} and SEC3), are significant and have positive signs. However, the sizes of (OUTPUT, KL_RATIO and ENERGY) coefficients were rather small compared to others. The effects of market demand constraints (DDCON), production manager education (PMEDU), restriction on capital requirements (INVREG), business supporting service (BSSl), and management- labor ratio (MANRATIO) coefficients are found negative and very significant. 3.7 Marginal Effects of Exogenous Variables Table 3.6 presents the marginal effects of exogenous variables on the search for foreign technology in Kenyan manufacturing. The first column lists the included variables, the second column shows the mean values of marginal effects and the third column presents the standard errors of mean marginal effects. · 49 Output petformance: The output variable has a significant positive marginal effect on the search for technology. This variable reflects both the size and the performance of the firm. . The results confirm the suggested link between technology and broader economic performance (Bhalla, 1975, Katz, 1983, Economics Departments, 1994, Bigsten, et aI, 1994). Large firms have an advantage in the AT process in comparison to small firms, because larger firms are more likely to have specialized resources to devote to the various search and acquistion fimctions. Organizational and Ownership Structure: The present study examined the impact of organizational and ownership structure on AT. We constructed several dummy variables, such as private, joint, state, foreign and non-foreign ownership. These variables were found insignificant in explaining the AT process. Bigsten et al (1994) showed that the probability of AT searching was positive ifthe fum is a sole proprietorship, a simple partnership, or a limited liability enterprise. That study showed that the probability of acquiring foreign technology is positive ifthe fum is neither a subsidiary of a domestic fum nor a subsidiary of a multinational coIporation (LOCC). The resuh suggests that the probability of acquiring foreign technology increases with a decrease in organizational complexity. Like other similar studies in developing countries (Herbert-Copley, 1990), we may argue that both results show that there is no necessary correlation between ownership structure and the acquisition of technology. 50 .' Table 3.5 Direct coefficient estimates VARTARTP II "ESTIMATE ~TnF.RR T-RATIO (,l·iMU)T TA RF. INTERCEPT 0.750 1.155 0.649 0.422 OlITPUT I 3.351E-09 1.288E-09 2.602 6.764 KL RATIO 6.848E-07 9.639E-08 7.105 50.478 ENERGY 1. 791E-06 4.314E-07 4.164349 17.342 SEC} 1.042 0.299 3.488 12.168 SEC3 1.620 0.284 5.704 ·32.539 SEC4 i -0.431 0.147 -2.929 8.581 MANRATIO -1.711 0.344 -4.970 24.705 GMDU 0.280 0.057 4.939 24.396 PMEDU -0.403 0.074 -5.471 29.927 GMEXP 0.016 0.007 2.252 5.073 PMEXP 0.012 0.002 6.666 44.451 LOCC 3.738 1.546 2.419 5.849 EXPO 1.674 0.470 3.563 12.694 I DOCON -0.736 0.181 -4.071 16.572 COMPI 0.598 0.159 3.770 14.210 COMPD 0.575 0.245 2.349 5.519 INVEREG -1.274 0.628 -2.028 4.112 BSS} -1.011 0.306 -3.299 10.886 BSS3 -1.889 0.981 -1.926 3.709 Capability Stock: Another suggested firm specific variable in the AT process is the capability stock. We constructed various variables such as the education and experience levels of the general manager and production manager and workers as proxies for the capability stock variables. Last years study suggested that the type of education rather than the level was important in the AT process. Table 3.6 shows that the education of general managers (GMEDU) and the experience in the business of both general managers (GMEXP) and production managers (PMEXP) have 51 significant positive relationships to the search for foreign technology. However, the estimated marginal effect of the education of production managers (PMEDU) is negative and significant. Market Structures: The Kenyan manufacturing sector is still inward oriented. Domestic competitive pressure is often low (KAM, 1988). Pricing is mainly done on the basis of markup over costs. Demand constraint. or lack of sufficient demand in the domestic market (DDCON), has a significant negative relationship to the search for foreign technology. The variable COMPI, reflecting the effect of competition pressure from imports, can be used as a proxy in assessing the second round effects of import hoeralization on AT. It is knO'WD that import competition pressure, which is the first round effect of import hbera1ization, introduces product quality and quantity changes in the manufacturing sector, reducing monopoly behavior at firm level and increasing cost effective production technologies. Certain product lines and products may be discontinued, with priority placed on new competitive product lines and products. Domestic firms are likely to consider quality and cost reduction production technologies. 52 Table 3.6 Marginal effects on the AT search process I VA'RIA'RLl" rn'R~Tn T-VALUE INTERCEPT 0.203 0.320 0.633 3.403 OUTPUT 9.05SE-I0 2.660E-lO KL RATIO 1.850E-07 7. 641E-08 2.421 ENERGY 4.853E-07 8. 824E-08 5.500 SEC 1 0.281 0.127 2.218 SEC3 0.438 0.182 2.407 SEC4 -0.117 0.059 -1.960 MANRATIO -0.462 0.206 -2.247 GMDU 0.076 0.032 2.388 PMEDU -0.109 0.044 2.483 GMEXP 0.004 0.002 1.746 PMEXP 0.003 0.001 2.564 LOCC 1.010 0.498 2.029 EXPO 0.452 0.209 . 2.159 DDCON -0.199 0.089 -2.224 COMPI 0.162 0.075 2.161 COMPD 0.155 0.089 1.744 INVEREG -0.344 0.204 -1.687 BSSI -0.273 0.136 -2.001 R~~1 -0 ,\10 o lQ? _, n'i1 The significant positive marginal effects associated with the variables COMPI and COMPD suggest that the probability of:firms acquiring foreign technology increases with domestic competition. Competition both from local :firms and from imports exerts pressure for new production technologies. This market effect is a consequence of the present import trade hoeralization policies. For a long time, domestic manufacturing operated behind high walls, protected from competition by tariffs, quota systems and other measures. Trade liberalization induced a 53 ·. competitive spirit among domestic industries which produce similar products, in terms of quality, price, and efficiency of delivery. Protectionist policies had created an environment which was not conducive to the acquisition oftechnology. Technical and Business Supp0l11nstitutions: Firms use technical support setvices to obtain skills from the outside in preference to the slow and costly route of developing and maintaining in-house expertise. Technical institutions are expected to provide specific services, such as quality contro~ repair and maintenance, technical assistance, personnel training, instrument cahoration, and other productivity techniques aimed at stimulating in- house skills. These institutions include the Bureau of Standards, Ministry of Technical Development, Kenya Industrial Training Institute (KITI), Ministry of Technical Training and Applied Technology, Kenya Industrial Research and Development Institute (KIRDI), Directorate of Industrial Training, Kenya Polytechnic, Kenya Institute of Sciences and Technology (KIST), Kenya Textile Training Institute and other institutions. Few firms in our sample used these technical services, pethaps because of the poor performance of these institutions, they perform routine work, or do not deal with actual industrial problems. These institutions do no produce technology outputs which are of relevance to and satisfY the needs of manufacturing firms. Soon after independence, Kenya also established a wide range of business support institutions to provide financial, managerial and technical services and information for faster industrialization. These institutions include the Investment Promotion Centre, Kenya National Chamber of Commerce, Kenya Association of Manufacturers, Kenya External Trade Authority, Kenya Industrial Property Office, Kenya Industrial Estates, Export Promotion Council, financial institutions, etc. Support services include planning, preparation, appraisal and setting up of projects. Other industrial services include provision of premises at low rents, repair and tool manufacturing facilities at the organizations' technical service centers, and provision of extension services on management, technical problems, etc. Few firms in our sample had received any kind of support from these institutions. The probit model (Table 3.6) shows that both BSSI and BSS3 have a significant negative effect on the AT search process. This suggests that local technical 54 .' .' efforts or support services are substitutable for the acquisition of foreign technology. Technology and Investment Regulations: It is important to examine the role of investment and profit restrictions, e.g., access to domestic finance, capital requirements and approval of foreign loans in detennining the operation of the firm and therefore the AT process. Our data shows that, for the relatively well-established domestic firms, access to domestic finance is not much of a problem. Problems ~~ to access to the domestic capital market were most acute in the textile sector. The MNC's feh restricted in their domestic borrowing. This is a result of a rule which restricts companies which are not majority-owned by Kenyans. The probit estimation showed that the investment restriction variable (INVREG) is significant and has a negative effect on the acquisition of foreign technology in manufacturing. This suggests that the restrictions on access to domestic finance and approval of foreign loans are an obstacle or a limiting factor in the acquisition of technology. State of technology, nature ofproducts produced and other factors: The capital-labor ratio (K_LRATIO) and ENERGY variables are used as proxies for the existing production technological base. The effects of both K_LRATIO and ENERGY on the AT process are positive and significant at the 10% level The administrative intensity variable (MANRATIO) has a negative and significant effect on the acquisition of foreign technology. This suggests that poor domestic management may be a problem. We also examined whether the nature of product or the type of manufacturing play a significant role in the AT process in Kenya. Table 3.6 shows that food (SEC 1) and wood (SEC3) 'are the most active sectors in searching for foreign technology. Interfinn cooperation and coordination, strategic alliances between firms, vertical linkages among input suppliers and cnstomers may play a double role in the acquisition of technology at the finn level (Aliinge, 1987, and Romano, 1990). These linkages may act as a source oftechnological information. On the other hand, linkages with foreign firms may 55 constitute a specific form of acquisition of technology, if it is effectively utilized at the mastery and productivity levels. Network size, degree of formality, and effectiveness differ from:firm to :firm, and are complex in nature in a developing economy. We were not able to construct a specific technological linkage variable. However, anecdotal evidence suggests that technological linkages are important in the acquisition of foreign technology, especially in the competitive environment. Due to lack of data, this study did not examine the effects of management personnel - behavioral characteristics (e.g., preferences, risk aversion attitude, awareness and perceived net benefit of modern technology) on the acquisition offoreign technology. However, we may argue (with Atuahene-Gima, 1992) that these personnel manager characteristics may have an effect on AT search activities, depending on the overall importance of the firm's specific characteristics, institutional support, market structure and government regulations. 3.8 Conclusion This chapter has examined the acquisition oftechnology in Kenyan manufacturing. Modern technical capabilities are core elements of structural transformation and industrial development. Expanding technical capabilities opens up possibilities for structural change and accelerates economic growth and other dynamics of enterprise development. However, there are specific limitations in the acquisition of advanced technology in Kenyan manufilcturing. The countly has no systematic and coordinated technology policy. It relies on implicit technology policies reflected in other policy instruments, such as tariffs, customs duties, pricing and protectionist measures. The country lacks a legal and institutional framework for handling technology issues openly and directly. About 42% of the sample firms this year had invested in new plant equipment, 45% had invested in used equipment, and 13% in mixed old and new equipment. This suggests that there is currently a good chance to start manufilcturing with used equipment in Kenya. Most 56 ofthe pJant equipment in Kenyan manufacturing is foreign. Certainly, there are possibilities for acquiring modem technology through direct investment in manufacturing, but Kenya's technology-related policies, especially those regarding machinery and equipment importation, were designed to facilitate the importation of whole pJants under the import- substitution strategy pursued in the 1970s. The policies allowed low or no taxes on machinery, but heavily taxed engineering raw materials (Coughlin and lkiara, 1991). This distortion encouraged importation oftumkey plants, inlnoiting the local fabrication of such machinery or spare parts. Commercial transfer of licensed technology usually takes place through firm-to-firm arrangements. The transfer ofindustrial property rights, including the provision oftecbnical assistance and services, may be in connection with the purchases and import of machinery or other capital goods that embody technology or as a result of direct foreign investment. Technology is also acquired through the dissemination and utilization of published technological information, the education and training of personnel at foreign technology institutes or R&D centers, 'and the exchange of information and personnel through tecbnical cooperation programs. The technology search process is also conducted via technical and management consultancy and other formal and informal channels.. There were few firms in the sample holding foreign technology licenses. Foreign licensing is not currently a common means of obtaining industrial technology. Kenyan firms do not import technology in this form because of their relatively unsophisticated technology level and because most manufacturing activities are in the traditional consumer goods sectors. Only a small number of flIlIlS, m!JStly in metal products, had expatriates in managerial, technical or other operative capacities. This may suggest that there are few administrative- and operative-skill constraints in Kenyan manufacturing. SmaD manufacturing firms engage relatively less-educated managerial, administrative and technical personnel. 61 % of small firm general managers had secondary education or less. The level of education ofthe production managers was often higher than that of the general managers. However, production managers have roughly the same lenght of experience in 57 the business as general managers (1 0 years). Low education and experience, or low management skills may limit a firm's ability to translate what is obtained in technical documentation into viable production processes. Competition between firms and among developing economies in the coming decades will be waged not only on the basis of cost and quality, but also on the basis of new production lines, processes and technologies. Kenyan manufacturing no longer has a protected domestic market, and thus improved product quality is vital to gaining a competitive edge and sustaining industrial development. The acquisition of foreign technology has a key role to play. The acquisition of technology in Kenyan manufacturing has so far relied mainly on direct investment and on imported plant equipment, with a minor role for technical assistance. Foreign licensing is not a common means of obtaining industrial technology, and relatively few :firms employed expatriates in operative capacities. Econometric analysis suggests that the probability of an entequise searching for foreign technology is intluenced by firm specific characteristics, such as size, organizational structure and output performance, and by extema1 filctors, such as market structure, technical and business supporting institutions, government policies and regulations. Industrial technology policies encouraging acquisition of foreign technology would be more effective with freer trade, first, simply because barriers to such access would be reduced, and secondly because the larger market created by more open trade and economic integration in Southern Amca would enhance the pay-off to technical advance based on frontier technical knowledge and information. 58 4. THE MANUFACfURING LABOR FORCE 4.1 Introduction Also in rOWld 2 we intetviewed a number of workers in each finn. Thus, in addition to our sample ofindustrial firms, we again have a sample of industrial workers. The survey largely parallels the previous one, so there is no need for a new comprehensive analysis of the type provided in the previous report. Still, we present some basic data analysis for comparative purposes. 4.2 TheData The swvey included inteIViews 'With a sample of workers vvithin each finn. We chose these workers in a random way vvithin each firm, tIying to get an adequate representation of different categories of workers, production as well as administrative. The classification into formal and informal firms was made from data provided by the Central Bureau of Statistics. Table 4.1 shows some basic statistics for several main variables. Hours worked and age are much less variable than are wages and experience. Wages are measured in Kshs per hour. Two experience variables are used, namely, general experience and firm-specific experience. The former is simply measured as the time since the worker left school; both are measured in months. 59 Table 4.1 Severa) main variables in the labor market N Mean Std C.V. Wage(Kshs) 1138 23.46 33.77 143.9 General experience 1142 113.26 93.42 82.5 (months since leaving school) Specific experience 1142 72.55 68.86 94.9 (months with firm) Hours worked/week 1139 51.42 9.77 19.0 Age (years) 1151 31.51 8.69 27.6 4.3 Gender Distribution Our sample contains a relatively small number of women. Table 4.2 shows the gender distn'bution of our sample and the corresponding weighted average for the four manufacturing sectors targeted by our swvey. 81.8% of manufacturing workers are estimated to be male, and only 18.2% female. Thus industrial labor force is dominantly male. Table 4.2 Gender distribution of labor force Sample Interference N % % Male 985 85.6 81.8 Female 166 14.4 18.2 Total I.151 100.0 100,0 Table 4.3 shows the gender distnlmtion of industrial workers by different characteristics of the firms employing them. The table show this distn'bution both by rows and by columns. For example, looking at the rows, 87.24% of workers in formal firms are male, and 12.76% are female. In informal firms, 74.07% are male and 25.93% female. Of all the male workers, 62.54% are employed in formal firms, and 37.46% in informal ones. Offemale workers, 60 .' 41.10% are employed in formal firms, and 58.90% informal. Thus most workers are male, and most ofthem are employed in formal firms. Formal firms employ proportionally more male workers than do informal ones, vvhile participation of females is higher in informal firms than in formal ones. These resuhs are very similar to those obtained in the previous survey. Table 4.3 Gender distribution by category Male Female Row Column Row Column FonnaI firms 87.24 62.54 12.76 41.10 InfonnaI firms 74.07 37.46 25.93 58.90 }·5 64.90 14.52 35.10 35.28 6-20 82.06 25.07 17.94 24.61 21·75 85.38 5.79 14.62 4.45 76-500 87.03 19.35 12.97 12.95 501+ 87.47 35.27 12.53 22.71 Food 85.54 35.69 14.46 27.10 Textiles 62.75 21.85 37.25 58.26 Wood 96.51 21.49 3.49 3.49 Metal 89.41 20.97 10.59 11.15 Workers are predominantely male in all sizes of firms., but not surprisingly, the smallest firms, those with only 1-5 workers, have the highest proportion offemale workers. All sectors have more male than female workers. While very few women work in the wood industry, the highest proportion of female workers now seem to be in textiles. This differs somewhat from the previous survey, vvIlen the food sector had the highest proportion of female workers. There are also clear and significant differences in the behavior of men and women in the labor market. Table 4.4 shows these differences, which are estimated for the average person in the labor market, without consideration of possible differences in education, training, etc. The table shows the gender distribution of several main variables of the labor 61 market with mean, standard deviation, and coefficient ofvariation. The differences ofthe men's and women's means of the different variables were tested, first under the assumption of equal variances, then under the assumption of unequal variances. The hypothesis of equal variances was itself also tested; the test statistic associated with this null hypothesis is reported in column F. The following column shows the probability of values larger than F. The differences in the means have a high degree of statistical significance. Thus, there is considerable empirical evidence supporting the hypothesis of different wages for men and women. The hypothesis of different behavior, performance and results for women and men in the labor market seems to be reasonable at first glance. The labor market seems to be gender-segmented, at least for the four industrial sectors considered in the sample. Wage differences are large: Hourly wages of male workers are 36% larger than those of female workers. But men have much longer general experience than women, about 75% more, and about 42% more specific experience. This suggests that women have a lower stability in the labor market, affecting their on-the-job training possibilities, and leading to lower wages. But it could simply reflect the difference in the average age of male and female workers, which is about six years. Possibly most of the gender discrimination, observed in the labor mm:ket as wage differences. does not originate in the labor market but before it. That is, gender discrimination could be caused by differences in education, training, and general cultural patterns that differentiate the age at which women come into the labor market and the length of their association with it. 62 Table 4.4 Gender distribution by main variables 1~ Mean Std C.V. Test t Prob>ltl F Prob>F Wage Male 979 24.59 35.14 142.9 Equalvar. 2.242 0.0251 2.357 0.0000 Female 159 18.13 22.89 126.2 Unequal var. 3.027 0.0027 General Esp. Male 978 123.22 92.04 74.7 Equalvar. 6.845 0.0000 1.125 0.1725 Female 164 70.47 86.78 123.1 Unequal var. 7.139 0.0000 Spec. Esp. Male 978 76.86 68.24 88.8 Equalvar. 3.966 0.0001 1.021 0.4200 Female 164 53.99 68.95 127.7 Unequal var. 3.937 0.0001 Roun worked! week Male 975 52.18 9.92 19.0 Equalvar. 5.096 0.0000 1.627 0.0001 Female 134 48.03 7.78 16.2 Unequal var. 6.051 0.0000 Age Male 985 32.53 8.60 26.4 Equalvar. 7.886 0.0000 1.351 0.0079 Female 166 26.95 1.40 21.4 Unequal var. 8.174 0.0000 In order to further examine the difference observed in the perfotmance of men and women in the labor market, we estimated wage profiles for these workers. Age and education are common variables in wage equations based on human capital theory. The experience variable included here is experience in the actual firm, and can be related to the on~the·job training hypothesis. We used the consistent estimator of the matrix of covariance proposed by White (1980), instead ofthe usual least squares estimator, because we assumed that we had heteroscedaticity in this stratified sample. 63 Table 4.5 Wage profile Variable Parameter estimate Standard error Chi-squared Prob>Chi-squared Intercept 0.410162 0.885097 0.215 0.6431 Age 0.125017 0.055675 5.042 0.0247 Age squared -0.001437 0.000773 3.454 0.0631 Education -0.071827 0.053044 1.834 0.1757 Education squared 0.004747 0.002556 3.449 0.0633 Experience in fum -0.000625 0.031858 0.000 0.9844 Experience squared 0.000313 0.001 0.098 0.7542 Adjusted R-squared 0.1549 Heteroscedasticity 62.8931 0.0000 Table 4.5 shows our estimates for all the workers in the four industrial sectors. The model does not fit the data very well In fact, it fit worse than for the 1993 survey: Adjusted R- squared equalled only 0.1549, compared to 0.2044 in the previous year. Many parameters were also not estimated with a satisfactory degree of statistical significance. The intercept is sman and estimated with a low degree of statistical Significance. Neither for education nor for experience were we able to estimate the usual inverted-U profile, and neither of the associated parametres is highly significant. But the rate of return to age was estimated at 12.0% annually and with a fairly high degree of significance. For the 1993 sample we estimated this parameter to be only 3%. We also estimated separate wage profiles for men and women, to see if there truly are gender differences in their performance and behavior in the labor market (Table 4.6). The adjusted R-squares were 0.1154 for men and 0.3105 for women. Thus, the model fits the data for women much better than the data for men. Once again, there is seems to be a high degree ofheteroscedasticity for both sub-samples. 64 Table 4.6 Wage profJIe by gender Male Female Variable Parameter Standard Prob>Chi- Parameter Standard Prob>Chi- estimate error squared estimate error squared Intercept 1.147176 1.024824 0.2630 -1.635334 1.829257 0.3713 Age 0.095618 0.062531 0.1262 0.243171 0.118829 0.0428 Age squared ·0.001038 0.000851 0.2225 ·0.003449 0.003032 0.0556 Education ·0.103454 0.063779 0.1048 ·0.001110 0.075187 0.9882 Education sq. 0.005683 0.002991 0.0574 0.002403 0.003861 0.5336 Experience 0.000555 0.034403 0.7330 -0.050200 0.068274 0.4622 Experience sq. 0.000152 0.001061 0.8861 0.004395 0.003266 0.1785 Adjusted R-sq. 0.1154 0.3105 Heterosced. 49.2702 0.0026 61.9027 0.0001 We estimated inverted-V profiles for age in both sub-samples, however, onJy for women did our parameters have a reasonably high significance level. The rate of return to age appears to be very high for women, much higher than for men. Again, intercepts are small with low statistical significance. In both sub-samples we estimate the effect of education with an unexpected negative sign. Only in the case of men could we estimate a positive effect of experience. The significance of education and experience parameters is low in both su~samples. We tested the null hypothesis that the corresponding parameters of men and women are equal. The corresponding Chi-squared statisticl was 5.2726. The probability of values higher than this statistic is 0.6267. Thus, we found poor empirical evidence supporting the hypothesis that there are different statistical structures for men and women in Kenya's industrial labor market. Note that these results contradict those obtained in the previous survey: Data obtained in 1993 did not allow us to reject the hypothesis of different statistical structures for men and women in the labor market. These resuhs, disagreeing as they do with the results form the previous survey, cast doubts on the robustness of our model POSSlbly we had an inadequate sample due to some sample 1 This lost is fonnally equivalent to the Chow lost. 65 selection effect. It is also often difficult to get good estimates from samples with a low education level: It seems that a miniumm level of general education is needed before education really matters. The sub--sample of women is also too small. How~er, it seems clear that age is a main determinant of the wage level 4.4 Labor Market Behavior In order to get a more integrated picture of the labor market, we re-estimated our wage fimction including a number of discrete variables controlling for different effects. Some of these variables are binaty, such as gender and fonnal or informal firms. Other variables have several values, such as industrial sector and ethnic group of the owner. In order to get an estimable function, one value was deleted for each variable and dummy variables were created for the rest. Once again, the matrix of covariance was estimated using the procedure suggested in White (1980). The inchlsion of a number ofnew variables improved the Adjusted R-squared value slightly, from 0.1549 to 0.2142. Again, there is strong empirical evidence in favor of the hypothesis ofheteroscedasticity. We found an inverted U-profile for age with a high level of statistical significance. A similar profile was now found for experience, but with a low level of significance. We did not find such a profile for education. The dummy for gender has a low degree of significance. There is a significant negative effect of working in all sectors as compared with working in the metal sector. The parameter for informal firms shows the expected negative sign, but it has a low degree of statistical significance. The results, again are poorer than those oflast year. 66 Table 4.7 Wage profde including control variables Variable Parameter estimate Standard error Chi-squared Prob>Chi-squared Intercept 1.727585 0.923785 3.497 0.0615 Age 0.132779 0.054836 5.863 0.0155 Age squared -0.001463 0.000746 3.846 0.0499 Education -0.069233 0.049551 1.952 0.1624 Education sq. 0.004145 0.002411 2.956 0.0856 Experience in firm 0.003602 0.027000 0.018 0.8939 Experience sq. -0.000128 0.000871 0.022 0.8832 Female 0.026934 0.129164 0.043 0.8348 Food -0.545702 0.121591 20.142 0.0000 Textiles -0.257216 0.133910 3.690 0.0548 Wood -0.167553 0.127803 1.719 0.1898 Informal firms -0.054804 0.109487 0.251 0.6167 Asian owner -0.430770 0.319620 1.816 0.1777 European owner -1.456815 0.448745 10.539 0.0012 African owner -1.166235 0.145307 64.417 0.0000 Adjusted R-squared 0.2142 Heteroscedasticity 145.2615 0.0001 4.5 Conclusions The 1994 sample was somewhat smaller than the 1993 sample. The data collected in 1994 has turned out to be less useful for testing empirical hypotheses about the labor market. The lack of robustness of the regression models casts some doubt on the suitability of the sample. We undertook the same analysis of segmentation as we did on Round 1 data: The results were similar but of poorer quality, so we do not report them here. The main issues of segmentation and discrimination, as well as the role of education. could be better analyzed with the help of a specifically-designed sample. 67 5. FINANCING AND INVESTMENT 5.1 Introduction Like last year, the pwposes of this section are to evaluate whether inefficiencies in the financial system in Kenya hinder the progress of viable projects, and to draw conclusions with respect to the influence of economic and regu1atOl)' policies on financial sector behaviour. After developing theory and hypotheses in section 5.2, we compare the asset and liability structures of this yeats sample with last yeats in section 5.3, focusing on firms that re~onded in both years. We wish to look for differences which may indicate that responses in either year were less than truthful. Special attention is also paid to the use of collateral when explaining access to the formal loan market. In section 5.4, data for informal loans, trade credits, and advance payments is presented. An overview of recent investments and their financing is provided in section 5.5. In section 5.6 we tum to econometric analysis of the 1994 interview data, focusing on explanation of the following aspects offirms' activities: 1. the use of bank loans and trade credits in the financing of ongoing operations, 2. the financing of recent investments, and 3. investment activity during the past year. In last yeats report "start up financing" was analyzed as well, but since approximately 90% of the firms are the same as last year, we do not repeat that analysis. Section 5.7 presents a comparison with last yeats results, as well as policy implications. 68 5.2 Theory and Hypotheses! We wish to associate different forms offinancing with the characteristics of individual firms. We choose characteristics for analysis based on theories offinancial institutions and financial contracts. These theories describe the firm as a "nexus of contracts" among various stakeholders. The contractual relation between shareholders and lending banks, for instance, is explained in the theory by the existence of asymmetric information. Information avai1ability, as well as enforcement mechanisms for contracts - be they explicit or implicit - take prominent roles in the analysis. Implicit contractual arrangements are very important for the financing of firms in countries like Kenya, where legal institutions supporting contract enforcement tend to be less effective than in the industrialized market economies.:2 Self-enforcing mechanisms for contract enforcement, such as risk to the :firm's reputation, are therefore relatively more important. It is expected that external financing is more acceSSlDle for firms iftheir contractual obligations are more easily enforceable and/or if self- enforcement mechanisms are strong. We use various characteristics of firms to obtain proxies for enforceability of contracts. The variables we use in seeking to explain investments and financing are divided into the following categories: 1. Variables related to expected profitability (see section 5.2.1); 2. Variables related to the riskiness of a firm and its investments (see section 5.2.2); 3. Variables related to the aVailability of information about a firm (see section 5.2.3); 4. Variables related to the enforceability ofa firm's contracts (see section 5.2.4); 5. Variables related to the regulatory environment (see section 5.2.5). 1 References to the theoretical literature we build on can be found in last year's report. 2 Last year's report contains a brief description of financial and legal institutions in Kenya. 69 The variables we are trying to explain in formal analysis are the following: A.. Forma/loans (see section 5.3): Formal loans are given by banks and so called non- bank financial institutions. We explain which firms have formal loans among their liabilities and the ratio of formal loans to total assets. B. Trade credits; payables (see section 5.4): We want to know how important trade credits are, and what characteristics tend to exclude firms from obtaining them One issue is whether trade credits substitute for formal loans when a firm has limited access to the formal loans market. c. Advance payments (see section 5.4): A firm's financing needs are increased by advance payments made, so our main interest is to analyze which type of firm is required to pay in advance. D. Bank loanslor financing olrecent investments (see section 5.5): While the formal loans variable above assesses formal loans in the overall capital structure, this vuiable is more representative of current access to the loan market. E. Investments (see section 5.5): Ifinvestments are constrained by limited access to the formal loan market, then we expect that the variables explaining investments are vet)' similar to variables explaining access to the loan market. Exact definitions of all variables, as well as references to questions in the survey questionnaire, are given in Table 5.1. Hypotheses regarding the effects f,fthe independent variables on each of the dependent variables are summarized in Table ~.2 and discussed in more detail next. 70 ," Table S.la Dermition of dependent variables II Variables Dermltion _Questionnaire ref. Bmkloan = 1 if investment in land. Page 27. question 4d. (probit) buildings, or equipment was financed by bmkloan; == 0 otherwise. Fonnalloan I The ratio of formal loans to total financing Page 32, question 3. total financing is the sum of overdrafts and loans from Page 32, question 8.1. (OLS) banks or nonbank financial institution, Page 32. question 8.2. Trade credit = 1 ifthe firm cWTently owes at least one Page 31. question 1. (probit) supplier trade credi~ = 0 otherwise. Trade credit I The ratio of trade credits owed to suppliers Page 31, question 1. total financing to total financing. (OLS) Advance = 1 if the firm CWTently has a claim (ie., Page 31, question 2. payments pre-paid goods or services) on a supplier, (probit) = 0 otherwise. Investor = 1 if the firm bas invested in land, Page 27. question 1. (probit) buildings, or equipment since the last interview, = 0 otherwise. Investments in The amount invested in Jand, buildings, or Page 27, question 2. equipment I total equipment during 1993, as a share of total assets. assets. {OT.~~ 71 . . T a ble 5tb Defi .. 0 f' d epen den t varIa bles lDltion m Variables Definition Questionnaire ref. Profit I total The ratio of gross profits total assets. Page 15, question 34. assets Spare capacity Q: 'How much more (%) compared to now Page 15, question 39. couldyou produce with existing equipment?' Perceived lack of = 1 if lack of demand was considered one of Page 43, question 2. demand the 3 biggest problems; == 0 otherwise. Debt Sum of overdrafts, informal and formal loans, Pages 27-34. mwe~tsand~an~payments~w~ Debt-equity ratio The ratio of debt to total assets. Pages 27-34. Collateral! total The ratio of collateral for formal loans as a Page 32, question 9. assets share of total assets. Level of local = 1 if competition from local firms is Page 42, question e. competition considered large or severe; (scale 0 or 1) = 0 otherwise. Degree of A scale from 1 to 6: the higher the score the Page 18, question 5. product less dWersified is the firm. diversification Capital/labor The ratio of the repl~ent value ofland, Page 14, question 26 ratio (KIL) buildings, and equipment to total labor cost and 27. Inventory / sales The ratio of the inventory of ,goods in pr~', Page 14, question 17, ratio 'finished goods', and raw materials in the end 22, and 24. of the year', to sales. Exporter == 1 if the firm is an exporter. Page 16, question 42. == 0 otherwise. Well-educated Number of unWersity-educated staff as a share Page 23, question 6, staff of total employment in the firm. Rows 3-5. Advertising I Ratio of promotion and ~ertising to sales. Page 15, question 32. sales ratio Formal/ informal == 1 if formal firm; Page ii. = 0 if otherwise. Li~nse problems = 1 if 'difficulty to obtain li~s' was Page 43, question 2. regarded as one of the 3 biggest problems this year, = 0 otherwise. License time The average number of days it took to obtain a Page 39, question 13. li~se? To be continued 72 Price controls = 1 if 'price controls' were regarded as one of Page 43, question 2. problems . the 3 biggest problems this year, =0 otherwise. Labor problems = 1 if 'labor regulations' were regarded as one Page 43, question 2. of the 3 biggest problems this year, = 0 otherwise. Bankruptcy costs The sum of the scores of the fonowing Page 40, question 19, questions: rows a and b a: 'government restrictions on sell-ing the enterprise...· b: 'the le2al orocess ofbankruotcv'. Lack of access to = 1 if access to imported raw mate-rials was a Page 42. question f imported raw large or severe problem; = 0 otherwise. materials Table S.le Defmition of scale variables Variables Dermition Questionnaire ref. Total Assets Sum in Ksh of inventories of raw materials. of Various. goods in process, and of finished goods), sale value of (land, buildings, and equipments), and informal lending, trade credits given, advance payments given. Sales Sales of manufactured goods, in Ksh. Page 13, question 13. 73 Table 5.2 Hypothetical relationships between dependent and independent variables in Probit and OLS analysis Independent variables Dependent variables Fonnal loans! Trade credits Advance payments Bank loans investments Total assets received made for financing OLS Probit OLS Prabtt Prabit Probtt OLS I. Expected profitability Gross profits/assets ratio + + ? + + + Spare capacity + Perceived lack of demand + 2. Riskiness offirm investments Debt-equity ratio N/A + Level of local competition + Degree of product diversification + Capital labor ratio + ? ? 3. A l·tti/ability of information Exporter + + + + + + Fonnallinfonnal + + + + + + 4. Enforceability ofcontracts Collateral + + + + + + Inventories + + + + + + Well-educated staff + + + + + + Advertising/sales ratio + + + + + + 5. Regulatory environment License problems + + + License time + + + Price control problems + + + Labor problems + + + Bankruptcy costs + + + Lack of raw imported materials + + + 74 5.2.1 Variables Related to Expected Profitability On the assumption that there is little variation in the economic conditions of firms in cross- section analysis, we included only proftis in this category last year. Although the general level of economic activity was low and economic conditions were considered unfavorable by most managers, it is still posSlDle that the variation across firms and industries was substantial. Thus, this year we also include spare capacity and perceived lack of demand as explanatory variables. There is no obvious theoretical relation between debt in the capital structure and current profits, except that for firms with equal returns on total assets, the profits after interest payments must be negatively related to debt. The effect of current profits on trade credits is ambiguous. Higher profits could increase the willingness of suppliers to give credits. At the same time, high profitability reduces the need for credits. In probit analysis we expect the first effect to dominate, because low profitability could cut access to trade credits completely. Similarly, we expect that trade credit is increasingly available as profitability increases, but at the same time the firm's ability to choose among various sources of funds improves with profitability. Therefore we do not have an unambiguous hypothesis for the impact of profits on the use of trade credits. Advance payments are generally required by a producer committing resources to a customer with little reputation. A buyer's low profitability stengthens a §Cller's incentive to ask for advance payments. Thus we expect a negative sign for profits in the analysis of advance payments made. However, there is a possibility in a severely credit constrained system, that highly p~ofitable firms lend to others by paying in advance. We expect that gross profits positively affect both investments and the supply of bank loans for investments, on the grounds that current profits are correlated with expected future profits and with the present value of the firm. This effect is expected to dominate in both OLS and probit. However, there is a possibility that higher profits would reduce the demand for bank.financing of new projects. For highly credit-constrained firms, this effect should 75 be negligible, however. High spare capacity, as wen as perceived lack ofdemand, are expected to affect investments negatively..The effect of these variables on the financial variables is expected to be the opposite of the current profit variable. The signs for these variables in Table 5.2 reflect hypotheses, when the supply ofvarious forms of credit is a constraining factor for the :firm. hi a highly credit-constrained economy, it can be expected that high spare capacity reduces the willingness of banks to supply loans. Both the availability of formal loans to finance investments, and the total amount offonnalloans, are expected to be negatively affected by high spare capacity and lack of demand. The willingness of other firms to supply trade credits is expected to be reduced similarily. although firms might want more trade credits. Conversely, suppliers may require more advance payments. 5.2.2 Variables Related to Risk Ifthe availability offinancing is a constraint, then a higher level of risk, which increases the pOSSIbility of bankruptcy, should reduce access to all kinds of financing, and should indirectly reduce investments Conversely, the need to pay in advance is expected to increase with risk. Risk is assessed in Tables 5.1 and 5.2 using the following variables: The debt/equity ratio, J collateral/asset ratio, level of local competition, degree ofproduct diversification, and capital/labor ratio. The risk: facing the fimis creditors increases with debt. We also expect that high competition and low product diversification increase the riskiness of the firm. A high capitaVlabor ratio could increase risk, if a large proportion of capital investments are 3 Note that the debt/equity ratio eannot be used to explain the ratio of formal loans to total assets, as these are simply two different ways of deseribing the same finaneial oondition. 76 sunk: costs. On the other hand, a large capital stock could make the firm more attractive as a borrower, if the physical capital can be offered as collateral. The latter possibility is investigated by nmoing probit and OLS regressions with collateral as one of the independent variables. It will be discussed in more detail below. 5.2.3 Variables Related to Availability of Information FlID1S with publicly well-known characteristics are expected to face relatively little rationing in capital markets, and their investment projects are expected not to be constrained by financial considerations. Thus, we expect higher investments, more bank-financing of investments and a higher proportion of debt in their capital structures. Well-known firms should also not be required to pay in advance to the same extent as others. The effect on trade credits is ambiguous, however. On the one hand, the supply of trade credits to well- known firms should be high, but given access to bank-financing, the need for trade credits should be lower. The dunnny variables that are used as proxies for public information-availability are exporter andformal/informal. In the analysis for last years report, we also used a dummy for incorporation as a limited liability company to express information and enforcement possibilities. This information is lacking for new firms in the sample this year. 5.2.4 Variables Related to Enforceability of Contracts We use variables for collateral, inventories, well-educated staff, and the advertising/sales ratio as proxies for enforceability of contracts. Loan contracts are more easily enforced ifthe borrowing firm has assets that can be claimed in case of default and ifthe firm has intangible assets at stake in contract performance. The "collateral" variable captures the situation when the fum has pledged assets to the fulfillment 77 of a specific contract. The capita1llabor ratio discussed above and inventories are variables that reflect the existence of assets that can be claimed by creditors as a group. Therefore, if inventories are marketable, we can expect that firms with relatively high inventories are less credit constrained than others. An ambiguity exists, however, for the relation between inventories and access to loans, because high inventories in a particular period could be a reflection of a decline in sales. In this case, the sign for the effect of inventories on bank loans is reversed. We control for management's perception of''low demand" in the analysis in order to capture a posSlole role for inventories in enforcement. The most important mechanism for self-enforcement of contracts is reputation. A fum with its good reputation as a reliable business partner at stake has a lot to lose from reneging on financial contracts. It will therefore be rationed to a lesser extent than other firms. We expect that a firm with relatively differentiated products and with strong brand-name recognition has the most to lose from loosing reputation by reneging on contracts. We use the share of universitely-trained employees as a proxy for product differentiation, and the amount spent on advertising and promotion relative to sales as a measure of brand-name recognition. These variables are expected to be positively related to formal loans, trade credits, bank loans for financing recent investments, as well as investments assuming that access to finance is a constraint on investments. The same variables would be associated with a decrease in advance payments. 5.2.5 Variables Related to the Regulatory Environment Expected returns on projects are reduced by regulatory constraints increasing costs at the planning stage, during operation, or at abandonment. The variables we employ to assess the fum's perceptions of the regulatory environment are license problems, license time, price controls problems, labor problems, bankruptcy costs, and lack ofaccess to imported raw materials. 78 A fum facing difficulties obtaining operating licenses or requiring a long time to obtain licenses incurs costs at the pbmning stage of a project. Price controls, regulation of the terms oflabor contracts and limited access to raw materials tend to reduce the profitability of an ongoing project. Govemment restrictions on the sale of an enterprise, or large costs associated with bankruptcy proceedings, increase the costs of abandonment. Our hypothesis is that all the variables reducing a project's profitability will tend to reduce investments as wen as the availability offinancing. Thus, the ratio of formal loans to assets, trade credits, the existence of bank loans for financing of recent investments, and investment itself are expected to be negatively related to all the variables for regulatory constraints. Advance payments are expected to be required to an increasing extent as the regulatory constraint becomes more severe. 5.3 The Asset-Liability Structure Table 5.3a+b show the liability structure of the 193 firms surviving in the sample from last year. 4 We show data for surviving firms in order to compare with the previous yeats figures. This yeats figures are not influenced by the 21 additions to the sample. Therefore, we do not present an additional table for the asset-liability structure of the whole sample. The calculation of total assets is described in Table 5.1. Retained earnings is a residual The general picture is the same as last year. Micro and small firms have very little access to formal loans (Table 5.3a),5 Trade credits and informal loans also do not contribute substantially to the financing ofthese films, Medium and macro films have the highest shares of debt "financing. The highest formal loan share is 24%, for macro firms. 4 Two firms have been deleted from the sample; because their loans were much larger than their total assets, There were six firms excluded last year for this reason. 5 Micro firms have up to 5 employees. Small firms have between 6 and 20 employees. Medium firms have 21 ~ 75, large firms 76 ~ 500. and macro firms have more than 500 employees. 79 The share of financing by "angels" - meaning friends and relatives, as opposed to more institutionalized infonnallenders - is low in all groups. The highest share appears for non- Asian, non-African micro-firms (3%), and for African small firms (2%). More institutionalized infonnalloans are obtained only by African micro-firms (3%) and large firms (1.5%). The share of trade credits is larger than last year. As we suspected then, our data underestimated the true extent of trade credit financing. Medium and large firms rely most heavily on trade credits; the shares for these two groups are above 5% regardless of ethnicity of owner. Large African firms standout with a trade credits share of 35%, compared to 9.3% last year. Non-Asian, non-African macro-firms also have a high share (15%), like last year, reflecting the fact that subsidiaries of multinational firms obtain intra- company credit. We will return to the issue of trade credits below. Advance payments received appear larger than last year for some groups. African micro- firms in particular received payments in advance, 5% this year compared to 0.3% last year. While large Afiican firms now rely on trade credits for financing, as discussed above, small African firms now seemingly have to provide those trade credits (14%) (Table 5.3b), up from 0.2% last year. Asian micro-firms show very small shares for trade credits on both sides ofthe balance sheet. All other categories show trade credit receivables at or above 5%, but the disparity between receivables and payables is most pronounced for African micro- firms. These firms are clearly in a weak financial position, which they partially relieve by requiring advance payments. Last yeat.s informallending was virtually non-existent in our data for all groups except for non-Afiican, non-Asian small firms. This year, informal lending is substantial only for Asian micro-firms (12%) and large African firms (13%). These figures are puzzling, and perhaps unreliable. Summarizing, the low share of formal loans in the liability structure is striking across alJ 80 fum-sizes and ethnic groups. Informal loans are used by micro and large African firms in particular, and they are provided to the greatest extent by large African and Asian micro- firms. Trade credits are a net source of financing for large, non-African non-Asian firms, while African micro-firms provide large trade credits to customers. To a lesser extent, the latter firms also receive advance payments. 81 Table 5.3a Liabilities (as percent oftotal assets). 193 firms 'sUlviving' from last year, by size offinn. -_ - I - .. , i~. . '. MICRON=40 SMALLN=46 MEDIUMN=49 LARGEN=33 MACRON=3 ~x· II ,. , Asian African Other Asian African Other Asian African Other Asian African Other Asian African Other '. , N=4 N=35 N=l N=17 N=28 N=l N=41 N=6 N=2 N=17 N=6 N=O N=1 N=O N=l Loans by "angels" 0.0 1.0 3.0 0.2 2.0 0.0 0.09 0.16 0.0 O.S 0.0 --- 1.8 ---- 0.0 Other informal 0.0 3.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.5 ...__..... 0.0 -_..... 0.0 loans Fonnalloans 0.6 2.0 0.0 8.7 7.0 0.0 19.0 8.0 3.6 10.0 9.0 --......- 24.0 -- ....... 22.0 Trade credits 0.09 0.3 0.0 3.S 4.0 0.0 7.0 S.7 5.0 6.0 35.0 -.......... 0.5 --- IS.O (payables) Advance payments 0.03 5.0 0.0 0.4 2.0 0.0 2.8 0.03 1.6 2.0 1.0 ............. 0.0 ---- 0.0 received Equity and 99.0 89.0 97.0 87.0 85.0 100.0 71.0 86.0 89.8 82.0 54.0 _ _- ..... 74.0 ---- 63.0 1 retained earnings Note: These figures are all mean values. 82 Table S.3b Assets (as percent oftotal). 193 firms 'surviving' from last year, by size offirm. MICRON=40 00 I SMALLN=46 I MEDIUM N=49 I LARGEN=33 I MACRON=3 Asian African Other Asian African Other Asian African Other Asian African Other Asian African Other N=4 N=35 N=l N=17 N=28 N=l N=41 N=6 N=2 N=27 N=6 N=O N=2 N=O N=l Trade credits II 0.6 14.0 5.0 7.0 7.0 12.0 14.0 7.6 12.0 10.0 18.0 -_...... 9.0 -.....-- 28.0 (receivables) Advance payments given II 0.021 0.41 0.0I 0.31 0.81 0.0 I 2.0 I 0.0 I 2.61 1.51 1.7 1 --I 0.0 1 ----- 1 0.0 Informal lending II 12.0 I 0.21 0.0 I 0.51 1.51 0.0 I 0.3 1 0.021 0.061 0.071 13.0 I ---- I 0.02 1 ----- I 0.0 Real assets I 87.3 I 86.0 I 95.0 I 92.0 1 90.61 88.0 I 83.0 I 92.0 I 85.0 I 88.0 I 67.0 I ---- I 91.0 I ---- I 72.0 Note: These figures are all mean values. Empty space means that there were no observations in that subset. The dataset used here is a different one from that used hereafter. This dataset is slightly smaller than last year (N= 193 vs N=224). but the firms are the same as last year. It only includes firms providing answers in all categories. Note also that two ftrms were deleted since their trade credit to total assets ratios were larger than 1. 83 Table 5.4 Use of formal loans, by size of flrDl and type of loan Micro Small Medium Large Macro Average N=45 N=49 N=51 N=40 N=4 Overdraft 8.89 51.0 82.4 90.0 100.0 66.5 Formal 15.56 57.1 82.4 92.5 100.0 69.5 loans. incl. overdraft Formal 6.67 20.4 27.4 37.5 100.0 38.4 Loan.excl overdraft Note: Numbers are size-weigbted pen:entages offirms in each lim group. Other formal loans include loans from banks, NBF!s, the government, or any other formal source. Table 5.4 shows the percent finns of different sizes with overdraft facilities or other formal loans. All macro-finns, and almost all medium and large .firms have access to overdrafts, on the other hand. Only 15% of micro-finns, and just over half of small.firms, have access to the formal loan market. Table 5.5 shows the use of collateral for formal loans by firm size, and helps to explain the financial difficulties of the micro finns. Only four micro-finns in the sample have formal loans, and of them only two, or 50%, provided collateral, compared to about 80% of the larger firms. Both mean and medium values of collateral are generally many times the value offormalloans, but the micro-finns are a notable exception. It is clear that these firms are generally unable to provide assets as collateral. Table 5.6 shows the use of collateral for formal loans by manufacturing sector. The differences across sectors are relatively minor, and are most likely explained by the size- distribution within each sector. 84 Table 5.5 Collateral for formal loans, by size of fU"m. Micro SmaU Medium Large Macro Average N=4 N=12 N=16 N=15 N=4 Percent of loans SO.O 7S.0 87.S 3.33 7S.0 76.16 eoUatera1ized Ratio of eollateral to 1.96 23.14 4.87 5.48 1.15 7.32 loan value, mean Ratio of eollatcral to 0.40 2.94 3.32 3.03 2.15 2.37 loan value, median Standard deviation 1 40.S4 3.77 7.42 2.16 13.17 Table 5.6 Collateral for formal loans, by sector Wood Teniles Metal Average N=15 N=9 N=14 Percent of loans 76.92 100.0 88.89 64.28 82.52 eollatcra1ized Ratio of 2.67 18.02 2.29 27.96 12.74 eoUatcral to loan value, mean Ratio of eollatcral to 2.42 3.38 3.70 2.22 2.93 loan value, median Standard deviation 1.83 29.9 7.13 22.0 15.22 5.4 Informal Loans, Trade Credits and Advanee Payments Table 5.3a+b showed the proportion of informal loans, trade credits and advance payments in total financing. This section descooes shares of firms with informal loans, trade credits, and advance payments with respect to firm size, formal registration offirm, industrial sector, and location; we also desco'be the use of collateral and interest rates for informal loans. Table 5.7a shows the shares of firms receiving informal loans by size group. Overall, one firm out ofive has borrowed from an angel or another informal source. According to Table 5.3a these loans represented less than one percent of total financing. Thus, informal loans are generally quite small 85 Table 5.7a Use of informal loans, by size offll'Bl and type of loan Miero SmaU Medium Large Macro Average N=5l N-a9 N-51 N=41 N-a Infonnalloans, 30.77 28.57 7.84 9.76 25.0 20.39 incl angels lnfonnalloans 7.69 0.0 0.0 2.44 0.0 2.03 excl. an Is Note: The figures represent the proportion offirms in a certain s~ group that CUJTeDtly holds an informal loan from either an angel. money lender, informal group, supplier, client, or any other informal source. No infonnalloans in the sample were given against collateral (Table 5. 7b). The interest rate figures given are probably unreliable. Micro-firms report an extremely low rate (1.6%), while small firms report an extremely high average rate (281 %), both 'With large standard deviations. The reported interest rates are below market rates for most size categories, possibly because "angels" (friends and relatives) offer subsidized rates. Table 5.7b Use of coUatera) and interest rates for informal loans, by size of fIrm SmaU Medium Large Macro Average N=6 N=4 N=4 N-l Percent of 0.0 0.0 0.0 0.0 0.0 0.0 loans collateralized Interest rate, 1.60 281.0 9.80 10.80 24.0 65.44 mean Interest rate, 35.0 2220.0 10.0 8.70 454.74 Std. Dev. Note: The table shows the mean interest rates paid in a given sector, as well as its standard deviation. This sample n~essarily smaller than in table 4a since it is derived from current outstanding loans. Tables S.8a+b show that loans by angels are especially important for infonnal firms (42%). Other infonnalloans are held by only 5% of informal firms, and 3% of all firms. 86 Table 5.8a Use of informal loans by formal and informal fll'lDs Informal Average N=57 Informal loans, 10.71 42.1 26. meL angels Informal loans, 1.43 5.26 3.34 exel. Table 5.8b Use of collateral and interest rates in informal loans Formal Informal Average N=13 N=l1 Interest rates, mean 21.0 103.0 62.0 Interest rate, 24.0 1724.0 874.0 Stand. dey. The sectoral distn'bution of informal loans is shown in Table 5.9 and the geographical distribution in Table 5.10. The biggest difference among industries is the average interest rate. The very high mean rate that was observed for small firms in Table 5.7 appears here only in the metal industry and only in Nairobi. But the validity of these figures is uncertain because of the low response rate to this question in all industries. 87 Table 5.9a Use of informal loans, by sector (percent offirms, size-weighted, within each group). Wood Te:stiIes Metal Average N=Sl N=44 N=S5 Informal loans, 13.04 21.15 25.0 20.0 19.8 incl. angels Informal loans, 4.35 0.0 4.54 1.82 2.68 excl. eiS Table 5.9b Use of interest rates for informal loans Food Wood Tu:tiIes Metal Average N=4 N=S N=7 N...a Interest rate, 8.0 0.0 1,2 303.0 78.05 mem Interest rate, 8.0 0.0 10.20 1852.0 467.55 Stand. dey. Table 5.10a Use of informal loans, by location (percent offirms, size- weighted within each group) airobi Mombasa Nakuru Eldoret Average =121 N=37 N=ll N=18 Informal Ioms, 19.83 27.03 14.28 11.1 18.06 incl. mgels Informalloms, 1.65 5.4 0.0 5.56 3.15 excl. Table S.10b Use of interest rates for informal loans, by location Nairobi Mombasa Nakuru Eldoret Average N=15 N=7 N=O N=l Interest rate, 179.0 1.40 38.0 54.60 mean Interest rate, 1403.0 14.0 67.0 371.0 Stand. dey. 88 The shares of firms receiving and giving trade credits and advance payments are shown in Table 5.11. Most if not all medium, large, and macro-firms receive and give trade credits. But only 16% of micro-firms and 47% of small firms receive trade credits, while 47% and 720/0, respectively, provide trade credits. These figures confirm the earlier conclusion (Table 5.3a and b) that providing trade credits may be a financial burden for many micro- firms, and even for small firms. The percentage of micro-firms and small firms receiving advance payments is also large (39% and 25%, respectively) helping to relieve the financial burden on those firms. Table 5.11 Use of trade credits and advance payments, received and given, by size of fmn (percent of firms, size-weighted, within each group). Micro SmaU Medium Large Macro Average N==50 N=53 N=53 N==44 N=4 Trade credits, 16.0 47.17 69.81 75.56 100.0 61.71 received Advance 8.33 9.62 15.38 23.26 0.0 11.32 payments, paid Trade credits, 47.06 72.22 90.74 84.09 100.0 78.82 given Advance 39.22 24.53 20.75 34.88 0.0 23.88 payments, received The overall share of firms receiving advance payments is 24%, while only 11 % pay in advance. This disparity could imply that larger firms help finance smaller firms in this way, but this inteIpretation is contradicted by the relatively high share of firms receiving advance payments in almost all size groups. Table 5.12 shows trade credits and advance payments given and received by formal and informal firms. Many more informal firms provide than receive trade credits. Informal firms are mostly small and micro-firms, so this result correlates with the previous one. Again, the burden is at lest partially relieved by advance payments received. 89 Table 5.12 Use of trade credits and advance payments received and given, by formal and informal rwms (percent offirms, size-weighted, within each group). Formal Informal Average N=149 N==5S Trade Credit. 65.33 16.67 41.0 received Advance Payment. 15.17 9.26 12.22 paid Trade Credit. given 82.24 50.91 66.58 Advance Payment. 26.17 36.36 31.26 received The tables indicate that trade credits do not effectively a substitute for fonnalloans for firms lacking access to the latter. Another indication that this is true is that the simple correlation between trade credits received and fonnalloans as shares of financing is nearly zero (.09). It seems, however, that some firms use trade credits from one supplier to finance advance payments to other suppliers; the correlation coefficient between trade credits received and advance payments made is .48. 5.5 Recent Investments and Their Finaneing In this section we focus on the shares of firms which invested and on the shares of firms which obtained various kinds of financing for investments during 1993. Table 5.13 shows this financing by size of firm. 90 Table 5.13 Financing of 1993 investments, by size of firm (percent of:firms reporting use of each source) Micro Small Medium Large Macro Average N=16 N=14 N=23 N=24 N=3 Retained 66.67 68.75 53.33 67.86 75.0 66.32 earnings Personal 27.78 6.25 10.0 3.57 0.0 9.52 savings Angels 0.0 0.0 0.0 0.0 0.0 0.0 Formal loans 0.0 6.25 20.0 10.71 25.0 12.39 Overdrafts 0.0 6.25 10.0 10.71 0.0 5.39 Trade credits 0.0 0.0 3.33 7.14 0.0 2.09 Money 0.0 6.25 0.0 0.0 0.0 1.25 lenders Holding 0.0 0.0 0.0 0.0 0.0 0.0 company Sale of 0.0 0.0 0.0 0.0 0.0 0.0 equity New 0.0 0.0 0.0 0.0 0.0 0.0 partners Other 5.56 6.25 3.33 0.0 0.0 3.03 Note: The figures are average percentages and due to roundings and multiple answers they may not sum to 100%. 91 Table 5.14 Financing of 1993 investments, by formal and informal ftrms (percent· of:firms reporting use of each sources) ~~ Formal Informal Average I - :-: :\: '~~-::'->r': ,<; ":;:,'- . .~ N=60 N=lO ---~ Retained earnings 64.86 59.09 61.98 Personal savings 6.76 22.73 14.74 Angels 0.0 0.0 0.0 Bank loan 13.51 4.54 9.02 Overdrafts 8.11 4.54 6.32 Supplier credits 4.05 0.0 2.02 Money lenders . 0.0 4.54 2.27 Holding company 0.0 0.0 0.0 Sale of equity 0.0 0.0 0.0 New partner 0.0 0.0 0.0 Other 2.70 4.54 3.62 Note: The figures are average percentages and due to roundings and multiple answers they may not add up 1000/0. 92 Table 5.15 Financing of 1993 investments, by location ofrIrDls (percent offirms reporting use of each source) Food Wood Textiles Metal Average N=19 N=23 N=l1 N=27 Retained 52.0 64.0 53.33 77.42 61.69 earnings Personal 12.0 4.0 13.33 12.90 10.56 savings Angels 0.0 0.0 0.0 0.0 0.0 Bank loans 20.0 8.0 0.0 6.45 8.61 Overdrafts 12.0 8.0 13.33 0.0 8.33 Supplier 4.0 4.0 13.33 0.0 5.33 credits Money 0.0 4.0 6.67 0.0 2.67 lenders Holding 0.0 0.0 0.0 0.0 0.0 company Sale of equity 0.0 0.0 0.0 0.0 0.0 New partner 0.0 0.0 0.0 0.0 0.0 Other 0.0 8.0 0.0 3.22 2.8 Note: The figures are average percentages and due to roundings and multiple answers they may not add up 100%. The retained earnings and personal savings figures combined roughly equal the corresponding shares of retained earnings in the overall liability structure shown in Table S.3a. This correspondence exists in spite of the fact that Table 5.13 is reporting percentages offums using particular sources offinancing for current investing, while Table 5.3a reported the relatiye value of various sources of financing overall. This correspondence holds for other forms of financing as well. More than 90% of micro-firms financed recent investments with personal savings or retained earnings. The corresponding totals for larger firms are 63-75%. No micro-firms obtained overdrafts or other bank loans, while 13-30% of larger firms did. Supplier credits financed some investments ofmedium and large firms. Money lenders were a source offinancing for 93 6.25% of small firms, but otherwise neither angels nor money lenders lenders financed recent investments. Other kinds of informal bOITowing probably are included in the last line, "Other". These sources have been accessible for 5-7 percent of micro and small firms. Table 5.14 shows financing of 1993 investments by formal and informal firms. Less than 10% of informal firms used overdrafts or other bank loans to finance recent investments, while another 10% obtained loans from money-lenders or "other" informal sources. Table 5.15 shows financing of 1993 investments by sector. Fonnalloans (overdrafts or other bank loans) were received by 6% of metal firms and 32% of food firms, with wood and textiles in the middle. These differences between sectors probably relate to the size of firms in each sector. Tables 5.16 shows the shares offirms, by size group, which invested in 1993, by type of investment. 45% ofall firms invested in 1993, ranging from 25-30% for small and micro- firms to 75% for the relatively few macro-firms. Most if not all investment in each size group was for equipment. Only a slightly smaller proportion of informal than of formal firms invested in 1993 (Table 5.17), a smaller difference than could be expected given the difference in access to formal financial markets. 1993 may not be a representative year, however, because of the recession and a general reluctance to invest. 94 Table 5.16 1993 investments by size offll'm Micro SmaU Medium Large Macro Average N=S3 N'==:!5 N-56 N=46 N=4 Percentage of 24.53 20.0 37.5 52.17 75.0 41.84 firms investing in equipment only Percentage of 30.19 25.45 41.07 52.17 75.0 44.78 firms investing in equipment, buildings or land in 1993 Note: The figure is the proportion of firms in each category that invested in equipment, and that invested in 1993. respectively. Table 5.17 1993 investments by formal and informal fll'ms Formal lDformal Average N=l56 N=58 Percentage of firms investing 35.90 27.59 31.74 in equipment only Percentage of firms investing 38.46 34.48 36.47 in equipment, buildings or land in 1993 Note: The figure is the proportion offirms in each category that inv~sted in equipment, and that invested in 1993, respectively. The propensity to invest varied across sectors, with the metal sector investing most frequently, and textiles the least willing or able to invest: Only 20% of textile firms invested. The high rate of investment in the metal sector is a surprise, since Table 5.15 showed that the metal industry used bank financing least frequently (6.5%), while about 20% of textile firms borrowed using either overdrafts or from money lenders. Perhaps relatively high retained earnings and own savings in the metal industry enabled metal firms to invest, while textile firms were constrained both by lack of retained earnings and by insufficient access to other sources. 95 Table 5.18 1993 investments by sector Food Wood Textiles Metal Average N==51 N=58 N==49 N=56 Percentage of:firms 31.37 39.66 20.41 41.07 33.13 investing in equipment only Percentage of:firms 37.25 39.66 22.45 48.21 36.89 investing in equipment, buildings and land in 1994 5.6 Results of Probit and Regression Analyses Tables 5.19-23 show the results of probit and OLS analyses. The dependent variables are formal loans as a share of total financing (Tables 5. 19a-b); the proportion offirms using bank loans for recent investments (Tables 5.20a-b); the proportion of firms obtaining trade credits (Tables 5.21a-b); trade credits as a share of tota! financing (Tables S.21c-d); the proportion offirms making advance payments (Tables 5.22a-b); the proportion offirms investing during 1993 (Tables 5.23a-b); and investments in 1993 as a share of tota! assets (Tables 5.23c-d). 96 .' 5.19a Formal loans for 1993 investments, OlS without collateral Variables Coefficients Constant -0.09"'·· Spare capacity 0.0003 Level of local competition -0.05 Degree of product diversification 0.07··· License time -0.000015··· Number of Observations N==141 Goodness of Fit Adj.-R2 == 0.231 Test for Autocorrelation Durbin-Watson = 1.867 Test for Heteroscedasticity ~ is homoscedasticity) Breusch-Pagan = 1183.96·· Reslrictions Ho: ~=O= 32.178··· ··· = significant at 1 % (reject nuD hypothesis) .'" =significant at 5 % (reject null hypothesis) ... =significant at 10 % (reiect nuD hypothesis) 97 Table S.19b Formal loans, OLS Variables Coefficients Constant 0.11 Profitltotal assets -0.13 Collateral/total assets 0.16** Degree of product diversification 0.04* Inventory/sales ratio -0.05 License time 0.003** Bankruptcy costs -0.02 Number of Observations N=33 Goodness of Fit Adj._R2 = 0.779 Test for Autocorrelation Durbin-Watson = 1.896 Test for Heteroscedasticity <.Ho is homoscedast. ) Breusch-Pagan = 41.380** Information Criteria Aleaike = 0.011 Restrictions Ho: ~=O= 17.291*** *** = significant at 1 % (reject null bypothesis) ** = significant at 5 % (reject null hypothesis) . = significant at 10 % (reject nuU hypothesis) 98 Table 5.20a Bank loans for 1993 investments, Probit Variables Coefficients Marginal effects Constant -1.07 - Spare capacity 0.01·· 0.003 Debt-equity ratio 3.41· 0.72 Capital.labor ratio -1.24 -0.26 CollatcralltotaI assets -0.07* -0.01 Number of observations N=20 Goodness of Fit Pseudo-R2 = 0.844 Predictions Ratio = 0.85 Test for Heteroscedasticity Wald-test = 0.799 uted explanatory value last year (limited liability, age of firm, size of firm, and ethnicity) were not included among the independent variables this year. They either did 6 The debt-equity ratio is very low by the standards of developing countries with functioning financial markets (Hussain, 1995). 121 not exist for the whole sample, or they do not have a clear theoretical interpretation. In addition to the strong resuhs for collateral this year, some other risk, infonnation, and enforcement variables seem to have explanatory value for sources of financing; for instance, the strength of local competition helped explain the availability of financing. Variables expressing availability of infonnarlon and the strength of self-enforcement mechanisms primarily explain access to trade credits and their volume relative to assets. Business condition variables seem to playa lesser role, although correlation with the collateral variable obscures the resuhs. Regulatory conditions explain less this year than last. Finally, we repeat the policy implications from last year's report: "... property rights to land for individuals should be both expanded and strengthened in the sense that owners should be able to transfer real property without consent from so-called Land Control Boards. These boards can veto the transfer of land to a bank after a borrower's failure to repay. The bankruptcy process should also be improved. The duration of the process and lack offaith in the legal system implies that lenders face great uncertainty about recovering loans. Informal financial institutions do not seem to playa large role.... It is possible that providing legal protection for informal financial transactions. in the form of court enforcement of contracts, could help the informal sector expand... " The importance of institutional reform, strengthening property rights legislation and other aspects ofthe legal system, is heightened by our finding that :firms with insufficient collateral for formal loans are further financially constrained by lack of access to trade credits, by having to offer trade credits themselves, and by needing to pay in advance. 122 TECHNICAL APPENDIX SA.l. Heteroscedasticity in OLS Since OLS estimation is well-known, we will only discuss the heteroscedasticity test. It is questionable whether there exist any good tests for heteroscedasticity at all, because the form ofheteroscedasticity is usually unknown, and when a test actua1ly detects non-minimized error variances, one cannot be sure that it is not some other mis-specification that has been detected. Therefore in almost all cases we have let the software itself determine the functional form. As ;Mentioned in the text, we have employed the Breusch-Pagan (B-P) test for a general form of heteroscedasticity. The B-P test is relevant for a wide class ofhypotheses that the varimce is some function of a linear combination of known variables. Its strenghth is that it does not require prior knowledge of the functional form. However, this generality is also its weakness since more powerful tests can be employed if we know the functional form (Kennedy, 1994). The test statistic is calculated as one-haIfthe explained sum of squares from a linear regression of u}/fJ 2 on a constant and the variables that are thought to be responsible for the heteroscedasticity. In this case, utis the OLS residual and fJ2 is the average of il 2 · The error is asymptotically distributed as chi-square, with the number of degrees of freedom equal to the number of responSlole variables. The null hypothesis is homoscedasticity. SA.2. Probit analysis and relevant tests7 When the dependent variable is a dummy OLS is no longer appropriate, and we use probit instead Probit analysis assumes that there is an unobservable underlying response variable Y j * , and that this variable can be defined by the regression relationship (1) where xjis the vector of explanatory variables, Wis the vector of parameters, and ujis the vector of residuals. What we observe is only the dummy variable y defined by y=l i/Yi > 0 (2) y=O otherwise 7 Most of this section is in all essence taken from Maddala (1983). 123 It follows that Prob(yj=l)=Prob(u j >-p 'Xi) (3) =l-F( -p 'Xi) where F is the cumulative distribution for u'" In our case, the values ofy are just realizations of a binomial process. Furthermore, y varies with XI (that is, y varies from trial to trial). This means that our likelihood function looks like this: (4) ;=0 i=} This is the probability of obtaining a loan, for example, multiplied by the probability of not obtaining a loan. In our probit modeL the functional fonn for F in equation (4) depends on our assumptions about ulin equation (1). We assume Ui to be normally distributed8, and therefore we obtain the expression: (5) 'The estimated coefficients themselves are rather meaningless. In the univariate case on1y the signs provide any information. To be able to say something about magnitudes, marginal effects, the partial derivatives, must be evaluated. 9 The marginal effect from, for example Xib on y is calculated as: (6) where xij is the jth element of the vector of explanatory variables, and P is the jth element of j 8 If one assumes that the oritical values are distributed as a byperbolie>-secant-square (in which case the cumulative distribution F is a logistic function), then the logit model would be the appropriate one. That is, we sbould use the logit if we have many extreme values. 9 Note that marginal effects should be evaluated for each observation, not at the sample mean of the explanatory variables. The second way of calculating marginal effects would be correct only if the model were linear, while the first way is correct for both linear and non-linear models. Also note that these two calculations are identical in the case of a linear model. 124 p.We can now say by how much the probability that y = 1 changes when an explanatory variable changes by one unit. There is, as usual, one condition for this inteIpretation: it is not enough that the variable is significant, but the marginal effect itself must also be significant. Therefore, we need to compute the corresponding standard errors, using the following formulas: 10 Asy.Var.[PD] ::: G j · Var[p] . G/ (7) where PD stands for partial derivatives, and (8) Here, ct>i is the density evaluated at Jix i " and I is the identity matrix. After all these calculations have been done for each and every observation, we simply take the means of the marginal effects and the standard errors. It should be mentioned that there is no use in calcu1ating marginal effects for a 'dummy variable, since the estimated coefficient itself already is the slope of the estimated equation. Although there are many suggestions for goodness-of..fit measuresll , none is universally accepted for probit analyses. We used the pseudo-R2, developed by McElvey and Zavoina. 12 For the pseudo-R2 , we first calculate (9) where lis the inverse of MillIs ratio.ll We then compute pseudo-R2 as: 10 This is the asymptotic covariance matrix for marginal effects evaluated at the mean, However, given that the limiting covariance matrices are identical, we can use this covariance matrix when marginal effects are evaluated at every observation as wen, 11 See for example Amemiya (1981) and Maddala (1983). 12 The pseudo-R2 is taken from Greene (1992). 13 The inverse of Mill's ratio is simply the density function for the standard norma) divided by the cumulative density function. 125 R2= (n-l) var(~) (10) (n+(n-l) var(~)) \\'here n is the number of observations. Our second goodness-of-fit measure is the predictive ability ofthe model, simply calculated as the proportion of correctly predicted outcomes. Probit models are sensitive to mis-specification (heteroscedasticity, omitted variables, etc.). Not only will the estimators be inefficient in that case (as in OLS), but they will also be inconsistent. This is, of course, a serious problem and must be corrected for. We employed the Wald-test, which can be used to test for heteroscedasticity, among other things. Although it is asymptotica1ly chi-square, it has unknown small-sample distnDUtiOn, which is a weakness. 14 Another weakness is that we have to make an assumption regarding the nature of heteroscedasticity. The number of tested restrictions in this case is the number of responSlDle variables, detennines the degress of freedom We calculate the test as (Greene, 1990): (11) Our assumption ofheteroscedasticity is of the multiplicative form: Var(u)=[e y/Zr% (12) where u is the vector of residuals again, z is the vector of responSlole variables, and y is the vector of parameters of z. Finally, we also calculate two information criteria, Akaike and Schwartz. An example of their application is when one reduces a model. To see whether parsimony improves after the exclusion of one variable, one can compare the information criteria before and after. If the information criteria decrease, greater parsimony has been achieved. IS The Akaike information 14 The same is true for the Lagrange Multiplier test and Likelihood Ratio test as well. 15 Two points have to be made here: First, the infonnation criteria should not be the only reasor- for accepting a reduction of the model. Second, neither Akaike nor Schwartz infonnation criteria is valid tor 126 criterion (AlC) is calculated as: AlC= -2 LogL + 2(k + s) (13) where k is the number of ordered values for the response, and s is the number of explanatory variables. The Schwarz information criterion (SC) is calculated as: SC= -2 LogL + (k + s) log(N) (14) where N is the total number of observations, and k and s are defined as above. comparison of two different models; they are entirely model-specific. 127 6. INFRASTRUCTURE, REGULATIONS AND BUSINESS SUPPORT 6.1 Introduction This chapter explores how entrepreneurs perceive changes in the state of infrastructure, changes in regulations to which firms are subject, and changes in access to and the use:fu1ness of business support services. Infrastructure services, licencing and taxation regimes, general regulations and business promotion services can impact decisively on fum performance. They are important parts of the overall commercial environment, and may individually or collectively either enable or inhibit the pursuit of coIporate goals. 1 This year's smvey shows that, in the year since the :first survey, firms perceive deterioration in the majority of infrastructure services, especially in ports, roads, electricity, freight transport and telephone service. As a consequence of this deterioration, the incidence of self-provision of some of the services increased between the survey rounds, through acquisition of private security and own freight transport, for example, especially by relatively large, modem firms with opportunities for exploiting scale economies. Modest acquisitions of other services were also reported. Most freight acquisitions were by the food sector, which seems to be experiencing infrastructure deteriorations most strongly. There is evidence of further government rationalization of licencing, tariff and tax structures, liberalization of the foreign exchange market, removal of price controls, and relaxation of various business restrictions, all of which are a boon to business pursuits. However, entrepreneurs perceive further deterioration in conuption and in the cost of labour, and there is still limited access to business support services despite the existence of a large number of business support institutions. Future policy reform will have to pay attention to improving the financing and management 1 This is despite confIicIing evidence on the importance of public infrastructure on industrial productivity, as suggested in Lynde & Richmond (1992), Nadiri & Mamuneas (1994), and Holtz-Eakin (1994) among others. 128 " of public infrastructure services. to addressing the rent extraction opportunities on which corruption thrives, and to creating incentives for business support institutions to improve service delivery to the business community. 6.2 Deterioration in Infrastructure Infrastructure services in Kenya continue to deteriorate, due to inadequate maintenance, as a consequence of public sector inefficiency, inadequate government revenue, and conuption. Kenya's road network has been deteriorating for some time now. Road service was the biggest perceived infrastructure problem in last year's suvery. The recent replacement of road toDs with a :fuel-based levy is meant to increase the revenue generated . for road maintenance. 2 Kenya's ports are no longer able to cope with freight movement in either direction, the port of Mombasa being permanently clogged up. Transit users are reportedly exploring alternative routes, as local importers and exporters continue incurring heavy losses related to port delays and wastage, The situation in the airports is equally .unsatisfactorY. More than halfthe firms in the SUIVey reported deterioration in air- and seaports. roads, and electricity, in that order (Table 6.1). AImo~ half complained about deterioration in freight transport and in telephone services. More than 20% perceived deterioration in each of the nine infrastructure services swveyed. 2 · The Road Maintenanee Levy Fwd Bill, which received parliamentary approval at the end of 1993, empovvered the Minister for Public Works and Housing to impose a levy on petroleum fuels to be coUected by petroleum dealers for maintenance of classified roads. FoUowing this bill, levies amounting to Ksbs 1.50 per liter of gasoline and 1.00 per liter of diesel vvere imposed for FY 1994195. 'l'hey vvere expected to generate Ksbs 1.5 billion, which is considered adequate for road maintenanee. Estimates suggest that poor roads are costing Kenya KsbslO billion annuaUy in extra vehicle operating costs (see World Bank, Kenya's Public Expenditure Review, 1994). 3 A cargo handling company at Jomo K.enyatta International Airport was forced to impose a five day embargo on imports of general cargo to unplug the freight clearing system. The cargo terminals at the two main international airports are often congested, and navigational equipment in aU airport pwportcdly in poor state and subject to occasional failures (The Economic Review, February 1995, 'Bad landing: Poor Equipment Shortage of Cargo Space Pose Fresh Problems', Nairobi). 129 Table 6.1 Weighted percentages of sample perceiving changes in infrastructure Improved Deteriorated % % Electricity 8.3 50.7 Water 1.6 24.1 Freight transport 7.7 49.0 Workers'transport 2.1 24.1 Roads 7.1 62.6 Telephones 5.1 46.1 Air- and seaports 1.7 66.5 Waste disposal 11.0 23.6 Security U.8 36.6 Note: For breakdown by formal and informal finn types, see Appendix Table 6A2. Formal firms generally perceived greater deterioration than informal ones. with the exception offreight transport (see Appendix Table 6AI). This may be because formal firms use air- and seaports, roads and telephones more intensively than do informal ones, and thus are more aware of changes. Perceptions about changes in the quality of services should in theory be influenced by the relative use-intensities of firms" so that informal firms with limited access to and use of roads and telephones, air- and seaports for example, would be less likely to perceive changes in the quality of these services. There are also major differences in· perceptions of deterioration in inftastructure services across'the four manufacturing sectors. More food sector firms reported deterioration, notably in air- and seaports, roads. telephones, freight transport, electricity, and waste disposal than did fums in other sectors (Table 6.2). Raw materials and products handled by this sector are more perishable, and the sector's trading outcomes are thus more sensitive to timeliness of delivery. Problems related to communications and transport are thus presumably more worrisome than in other sectors. Wood sector firms are most aware of 130 deterioration in the condition of roads, reflecting the deplorable condition of access roads leading to logging sites. Textile sector firms are most aware of deterioration in air- and seaports, through which these firms' raw materials are procured. and also of deterioration in security. Most metal sector firms reported deterioration in telephone services and roads. Table 6.2 Weighted percentages of sample perceiving deterioration in infrastructure service, by sector Food Wood Textile Metal ALL Electricity 74.7 43.4 27.1 46.2 50.7 Water 18.2 52.8 21.4 15.9 24.1 Freight Transport 75.8 36.9 36.6 19.7 49.0 Workers Transport 22.9 74.7 49.4 10.4 24.1 Roads 86.0 18.2 39.6 49.2 62.6 Telephone 78.0 37.1 23.6 63.8 46.1 Air- and seaports 89.2 34.4 65.0 37.0 66.5 Waste Disposal 37.1 18.3 33.9 23.1 23.6 Security 19.1 40.0 52.4 43.7 36.6 Probit analysis" shows that the probability of firms perceiving deterioration in air- and seaport services increases with output, as well as with size, generaDy. Metal working firms and those located in Nakuru also are more likely to perceive deterioration (Table 6.3). These parameters describe firms likely to use air- and seaports more intensively than others, especially for procurement of raw materials. S For roads, the probability of perceiving deterioration increases generaDy with size (Table 6.4),6 Capacity utilization, however, has 4 'Some of the results presented in this and subsequent sections are based on probit analysis used to explore the determination of perceptions of deterioration in air- and seaports, roads, electrioity, freight transport, telephones, and in regulations. The food sector, Nairobi and micro enterprises are used as benohmarks for other sectors, looations and sizes. The results are restricted to direction rather than magnitude of impacts, as estimation outputs do not direotly yield marginal efIeots, and statistioal pathologies suob as beterosoedasticity oannot be ruled out. 5 It is not clear, however, vomy firms in Nakum should be as ooncerned about deterioration in air- and seaports as are firms in Nairobi, and more so than firms looated elsewhere. Conversely, the growth coeffioient, whioh might be expected to be signi.fioant, is not 6 The growth, output and Natruru ooeffioients are. however. not signifioantly different from zero. 131 ·. an 1Dlexpected negative effect on the probability ofperceiving deterioration in road services. The probability offums perceiving deterioration in the provision of electricity increases with size, but decreases with capacity utilization, possibly as a resuh of greater opportunities for investment in stand-by power generation. This probability is also higher for food firms than for the other sectors, and for those located in Nairobi and Mombasa (Table 6.5).7 Table 6.3 Probit results for factors correlated with perceived deterioration in port services Variable Coefficients t-ratios INTERCPT -1.889 -2.839 OUIPtn' 2.85685E-9 ** 2.111 GROWI'H 0.006 0.459 CAPACITY -0.479 0.841 TEXID -0.354 0.840 METALD 0.609* 1.718 WOOD -0.536 -1.345 ELDD -0.215 0.402 NKRD 0.826 * 1.939 MSAD -0.454 -0.971 SIZE2 0.717 1.290 SIZE3 1.350** 2.409 SIZE4 2.009*** 3.341 SIZE5 9.762 0.0001 Note: Log Likelihood =-60.9 For definitions of variables, see Appendix 6A2. One. two, and three asterisks indicate significance at lo%, 5%, and 1% levels, respectively. Standard errors were not White-corrected for heteroscedasticity. This Dote also applies to the subsequent probit tables. 7 Since Nairobi is used as the benchmark, the high positive estimate for Mombasa suggests that the probability of a firm perceiving deterioration in provision of electricity is higher for Mombasa than for Nairobi firms, so that problems related to provision of eletricity were possibly more acute in Mombasa than in other locations. 132 Table 6.4 hobit results for factors correlated with perceived deterioration in road services Variable Coefficient t-ratios INTERCPf 0.960 2.084 OUTPUT 3.5984E-IO 0.408 GROWTH 0.009 0.774 CAPACITY -1.266··· -2.586 TEXTD -0.433 -1.305 METALD -0.332 -1.036 WOOD 0.130 0.406 ELDD -0.495 -1.269 NKRD 0.341 0.897 MSAD -0.699** -2.477 SIZE2 0.475 1.594 SIZE3 1.112··· 3.410 SIZE4 1.176··* 3.200 SIZE5 0.983 0.952 Note: Log Likelihood -91. 9 Table 6.5 hobit results for factors correlated with perceived deterioration in electricity supply Variable Coefficient t- ratios INTERCPf -0.364 -0.820 OUTPUT 4.6368E-IO 0.760 GROWTH -0.007 -0.876 CAPACITY 0.684 1.520 TEXTD -0.161 -0.519 METALD -0.081 -0.275 WOOD -0.573* -1.908 ELDD -0.882 -1.878 NKRD -0.085 -0.246 MSAQ 0.760··* 2.672 SIZE2' 0.460 1.465 SIZE3 1.152··* 4.736 SIZE4 0.805** 2.284 SIZES 1.311* 1.793 Note: Log Likelihood -103.5 133 Table 6.6 Probit results for factors correlated with perceived deterioration in freight transport Variable Coefficient t-ratios INTERCPT -0.442 -0.849 OUTPUl' 1.56027E-9 1.255 GROW1H -0.012 -1.322 CAPACITY -1.500*** -2.706 TEXTD -0.726· -1.727 METALD 0.052 0.155 WOOD 0.212 0.643 ELDD -0.026 -0.059 NKRD 0.194 0.511 MSAD -0.724* -1.786 SIZE2 0.265 0.654 SIZE3 0.929*· 2.339 SIZE4 LI05** 2.482 SIZES 2.499*· 2.672 Note: Log Likelihood -75.7 Table 6.7 Probit results for factors correlated with received deterioration of telephone services Variable Coefficient t-ratios INTERCPT -1.211 -2.584 OUTPUT 2.77792E-9 ** 2.153 GROWTH 0.006 0.638 CAPACITY -0.973** -2.099 TEXTD 0.377 1.124 METALD 0.241 0.763 WOOD 0.433 1.412 ELDD 0.357 0.849 NKRD 0.426 1.245 MSAP -0.057 -0.815 SIZE2' 0.945*** 2.663 SIZE3 1.677 *** 4.548 SIZE4 1.463**· 3.526 SIZE5 0.850 1.069 Note: Log Likelihood -98.6 The probability offinns reporting deterioration in freight transport increases with firm size 134 but decreases with capacity utilization (Table 6.6). The probability offirms perceiving deterioration in telephone service increases with output, size and capacity utilization (Table 6.7). To summarize, many manufacturers in Kenya report deterioration in the quality of key infrastructure services. Finns' output, growth, capacity utilization, size, sector and location have differential impacts on fumts perceptions of deterioration in different infrastructure services. The relatively larger food sector firms located in Nairobi and Nakuru are most concerned about most of these deteriorations. 6.3 Acquisition of Own Infrastructure Facilities As demonstrated in the first wave report, firms faced with steadily deteriorating infrastructure service may opt for self-provision. This likelihood increases along the size scale, owing to opportunities for exploitation of scale economies, and it increases with location in Nairobi, where public services have been most stretched. Se1f..provision is also limited, of course, to services subject to reasonable exclusivity and appropriability. Between the first and second waves, 36% of the firms, mostly formal ones, either acquired or expanded their security services, and 25% their freight transport services. Although security was not reported to have deteriorated more than other services, more firms acquired private security services, perl1aps due to the presence of a more developed private security market, which in turn may be due to the greater exclusivity and easier appropriabiJity of security service. There was limited acquisition of other services, largely by formal firms with large enough operations to permit exploitation of scale economies (Table 6.8). 135 Table 6.8 Acquisition of own infrastructure by formal and informal flJ"Dls ~L ture acquisitions Fonnal firms Informal firms All firms Generators 3.9 0 2.4 WeUslcisterns 3.2 0 1.9 Walkie-talkieslradios 9.1 0 5.4 Workers'transport 5.1 0 3.1 Roads 1.0 0 0.6 Wute disposal 15.4 6.2 11.8 Air- and seaports labour 1.8 0 1.1 Freight transport 39.9 2.5 25.4 Security 42.1 27.6 36.3 A greater percentage of firms in the textile, metal and wood sectors expanded or acquired their own security services than did food firms (Table 6.9). The majority of these are isolated fonnal finDs, although security services were also aquired by more than one quarter ofthe informal firms making security services by far the largest category of acquisition by infonnal firms. Most acquisitions offreight transport were by food sector firms, where two- thirds offirms reported having acquired freight transport facilities, commensurate with an equally large percentage of firms reporting deterioration in public freight transport services. In this sector, a reliable freight transport system is crucial for timely delivery of industrial raw materials and :final products, both ofwhich are more perishable than inputs and outputs in other sectors. 136 Table 6.9 Acquisition of own infrastructural by weighted percentages of manufacturing fmns, by sector Infrastructure acquisitions Wood Textiles Metal Generators 0.5 2.5 1.8 6.2 WeUslcistems 2.4 0.7 1.0 3.4 WaJkie..talkie/radio 22 1.9 4.4 16.1 Workers' transport 0.3 0 0 2.5 Roads 4.6 4.5 1.1 1.7 Wute disposal 14.8 8.3 10.7 11.2 Air- and seaports labour 0.5 0 0 4.7 Freight transport 66.6 1.3 0.1 8.7 Security 21.7 38.2 47.6 44.8 Deterioration ofinfrastructure services in Kenya continues to force some firms to provide for some of the more critical services in order to survive. The significant increase in the incidence of self provision in the last year is evidence of growing dissatisfaction with publicly supplied infrastructure services. 6.4 Licences, Taxes and General Regulations Fonnal manufacturing :firms in Kenya are subject to a wide range of licences determined by the specific 1ine of business and the municipality of operation. Licence fees are paid to the central and local governments. Some licences are obtained only once, such as registration with the registrar of companies, with the Kenya Bureau of Standards (KBS) and with the National Social Security Fund (NSSF), while others are renewed periodically.s Subsequent 8 For a number of these licences, firms are given the option of renewing for periods greater than one year, up to a maximwn of 36 months, depending of the firm's financial ability. TIns arrangement reduces transactions costs related to licence procurement. 137 payments to the KBS and NSSF are activity based. 9 Between the survey rounds, about 70% of firms renewed their trade licences (Table 6.10). Import licences were of interest to manufacturing firms using imported raw materials, but were irrelevant to the majority of fum&. Only 17% offirms either obtained or renewed general import licences, while a smaller number offirms obtained or renewed a collection of other specific import licences. 10 There was also substantial renewal and acquisition of an assortment of other unspecified licences, mainly by formal firms. Informal firms, as the term suggests, operate largely outside the general regulatory environment. Still, about 37% of the informal firms renewed trade licences, presumably held with the local governments. 46 % of these firms also renewed other miscellaneous licences. Overall, metal-working firms renewed fewer licences than firms in the other sectors, while fewer textile firms renewed trade licences than in other sectors, reflecting the dominance of informal enterprises in textiles and metals. The percentage of firms renewing general import licences was higher for the textile and wood sectors, although licence renewals in general were most frequent in the food sector, where more than 95% of all the firms renewed their trade licences, compared to an average of65% for all firms. Overall, the food sector seems most constrained by licences. 9 Payments to the NSSF depend on total payrolls: 5% of each employee's monthly pay is deducted at source by the finn, which contributes a fwther 100;0 to make 15% remitted to the NSSF. 10 Import licences were removed altogether during the survey period. 138 Table 6.10 Acquisition and renewal of licences in the past year by weighted percentages of sample Licence Type Acquired Renewed Total ~upationalheahh 0.1 17.1 17.2 Trade licences 3.6 64.6 68.2 General import licence 1.6 IS.7 17.3 Speficic import licence 1 2.5 10.6 13.1 Specific import licence 2 0.5 0 0.5 Specific import licence 3 1.0 0.6 1.6 Specific import licence 4 0.8 0 0.8 Other licence 1 8.6 67.8 76.4 Other licence 2 9.7 71.5 81.2 Other licence 3 3.2 66 69.2 Note: For breakdown by sector, see Appendix Table 6A3. Trade licencing and taxation legisJation has been in a state of flux for the last four years. In 1991, diverse trade licences required by the central government were consolidated into only three, namely the manufacturers trade licence, the general trade licence and the small traders (regulated) licence. The Finance Act of 1994 added the wholesale trade licence, catering licence, motor vehicle repair licence, regulated trade licence, distn'bution-of-goods licence, retail-sale-by-manufacturer- of -own-goods licence, and other miscellaneous occupations licences. These licences are payable at different rates for general business areas, urban areas, or rural areas, and are renewable for periods of 12, 24 and 36 months. There is evidence of considerable recent improvement in licence application turn-around times, 'With a mean time lag of 5.5 days. Reported assistance and extraordinary charges related to licence acquisition were insignificant, despite reported increase in corruption. II 11 However, responses to questions related to how much firms paid in assistance and extraordinary charges are likely to have been distorted by reponse bias. Most most respondants were unwilling to openly declare bow much they spent on corruption. We therefore expect more corruption related charges than reported. 139 ·. Mean value offirm output in 1993 was Kshs 68.4 million, sales were 46.5 million, and profits were 5.96 million. On this, fums on average paid Kshs 1.3 million company tax, 2.6 million VAT, 0.253 million import duty on capital equipment, and 1.568 million import duty on raw materials. Legislation pertaining to import tariffs and taxation has been modified under a tax modemi7Jrtion programme put in place in 1988. The 1994/95 Finance Act reduced the top import tariffrate ftom 50% to 45% and reduced the number of effective tariff bands, with 0% rate on goods imported under the Export Promotion Programme Office (EPPO) and Wlder the Essential Goods Production Support Programme (EGPSP).12 The Finance Bill also adjusted VAT rates downwards ftom a previous high of 40%. The current rates are 30%, 18% , 50/0., and 0%. The minimlJm tumover for compuIsary VAT registration was also increased, to Kshs 900,000 for manufacturing, and to Kshs 1,500,000 for traders and service suppliers. 13 Other changes occurred in the' last year. About 70% of firms perceived a worsening of conuption, and 43% reported a rise in labom costs (Table 6. 12a). On the other hand, 45% reported a reduction in difficulties in obtaining business licences, 42% an improvement in opportunities for obtaining investment benefits, 41 % improvements in price controls, and 29% improvement in the tax bmden. There were considerably smaller changes in overall labom regulations, enterprise ownership regulations, and in restrictions on activities. More fonnal than informal fums reported worsening of conuption, and more informal than formal firms reported rising labom costs (Table 6.12b). Many more formal than informal firms reported reduction of difficulties in obtaining licences and improvement in opportunities for obtaining investment benefits. 12 EPPO was introduced in 1990 to gradually replace the Export Compensation Scheme which had been abused, and was eventually abolished in September 1993. Under EPPO, duty remission is given on raw materials used for export processing. EGPSP relates to imports of materials either for processing essential goods for distribution in the domestic market, to the armed forces, or for donors funded projects that require duty-free imports. See also submissions by the Kenya Association of Manufacturers. 13 Changes were also made in income tax effective from beginning of 1995. Included in these were an increase in tax reliefs by 50%, widening of the tax brackets by 300/0, reducing the maximum personal income tax rate from 40% to 30010 and increasing tax deductable contributions to registered pension and provident funds by 100010. 140 More food firms than others reported reduction of difficulties in obtaining licences, improvement in opportunities for obtaining investment benefits and improvement in price controls, whereas more textile firms reported reductions in the tax burden (Table 6.12c). Although ha1f ofthe finns in each sector reported worsening of corruption, the food sector dominated, with more than 80% of the firms reporting a worsening of corruption (Table 6. 12d). Rising labour costs were reported most strongly (70%) in the wood sector. Saw- mill operators, most ofwhich are located in the clash-ridden Rift Valley, had difficulties in retaining workers. Except fur conuption, which was reported as a severe problem by 47% of the firms, most regulations do not seem to pose major problems for large numbers offirms (see Appendix Table 6A4). However, the wood sector seems most affected: 19% of wood firms considered corruption a severe problem, 17% considered restrictions on activities severe, and 16% considered difficulties in obtaining licences severe (Appendix Table 6A5b) Table 6.118 Weighted percentages of sample firms percieving changes in government regulations Regulations Improved Worsened Nocbange Ownership regulations 14.8 0.0 85.5 Taxes 28.6 19.2 52.3 Restrictions on activities 12.1 3.6 84.3 Opportunities for investment benefits 41.6 6.6 51.9 Labour costs 3.5 43.8 52.7 Labour regulations 2.9 5.0 92.9 Diffulties in obtaining licences 44.6 13.3 42.1 Corruption 6.1 69.7 24.2 Price controls 41.3 5.8 52.9 Other 51.0 3.5 45.5 Note: For severity of problems perceived, see Appendix Table 6A4. 141 Table 6.11b Weighted percentages of formal and informal rams perceiving changes in goverment regulations FORMAL INFORMAL I Regulations Improved Worsened Improved Worsened Ownership regulations 11.7 0.0 24.9 0.0 Taxes 34.6 16 10.9 28.5 Restrictions on activties 13.4 2.5 7.3 7.5 Opportunities for investment benefits 50.2 6.2 4.2 8.2 Labour costs 5.6 32.5 0 63.6 Labour regulations 2.4 6.6 4.6 0.0 Difficulties in obtaining licences 58.4 6.6 10.7 29.6 Corruption 8.8 76.3 0.8 56.8 Price controls 50.9 5.9 27.9 5.6 Other 54.4 1.8 49.0 4.5 Table 6.11c Weighted per~entages perceiving improved regulations, by sector lations Food Wood Textile Metal All Ownership Regulations 2.1 1.5 37.4 21.5 14.8 Taxes 15.0 17.6 55.5 26.8 28.6 Restrictions on Activities 3.8 0.9 26.9 24.5 12.1 Gaining Investment Benefit 70.1 1.9 24.1 24.6 41.6 Labour Cost 1.5 1.9 4.7 7.6 3.5 Labour Regulations 0.3 3.6 2.8 9.1 2.9 Difficulties in Obtaining Licences 71.7 35.1 14.4 38.4 44.6 Corruption 5.7 2.3 14.0 1.9 6.1 Price 'Control 62.2 25.0 49.9 24.9 41.3 Other 8.8 34.5 85.9 21.5 51.0 142 Table 6.11d Weighted percentages of sample firms perceiving worsening of regulations, by sectors Regulations Food Wood Textile Metal All Ownership regulations 0.0 0.0 0.0 0.0 0.0 Taxes 10.0 21.5 29.3 22.4 19.2 Restrictions on activities 0.5 11.4 4.8 1.7 3.6 Opportunities for investment benefits 3.7 5.9 10.8 9.0 6.6 Labour costs 24.8 71.3 52.3 41.8 43.8 Labour regulations 1.7 6.5 6.6 9.4 5.0 Difficulties in obtaining licences 6.2 30.0 15.1 9.5 13.3 Corruption 84.1 66.7 65.8 48.6 69.7 Price controls 1.2 5.9 9.5 8.1 5.8 Other 0 16.2 0 2.8 3.5 Note: For severity of problems perceived, see Appendix Tabls 6A5a-b. Many, if not most, formal firms reported improvement in foreign exchange related regulations, such as repatriation ofprofits, foreign exchange for business travel, restrictions on foreign loans, payment of non-resident fees, and payment for technology licences (Appendix Table 6A6). These regulations are of limited relevence to informal firms which have little apparent use for foreign exchange. About one quarter of the formal firms also report improvement in access to domestic finance. NQne of the informal firms, which have limited access to formal credit and thus rely on either personal savings or "angels", reported such improvement. Opportunities for relying on personal savings and informal sources have been drastically eroded by inflation. Relatively few firms reported either positive or negative changes in joint venture restrictions, capital requirements, and restrictions on activities. Similarly, most firms did not report any changes in regulations reJated to reduction in production and firm closure, except for government restrictions against firing workers, which were reported to have improved by 39% and 32% offormal and informal firms, respectively. 143 .. .' These trends appear fiUrly uniform across the four sectors, with a few exceptions: access to domestic finance is reported to have improved by more than half the textile :firms, relative to an average of 20% for all sectors. Similarly, more than 66% of food firms reported improvement in government rules against firing workers, twice the average for all firms. Improvement in this area must be a side effect of the government's civil service retrenchment. While this programme is in force, it will be difficult for the government to maintain lay-off restrictions without appearing to promote a double standard. In any case, these regulation pose large or severe problems only to a smalI percentage of formal firms (Appendix Table 6A7b). More generally, labor-related regulations seem to affect the textile and metal working sectors more than the other sectors. 14 Overall, :firms reported a general relaxation of different aspects of the regulatory environment, particularly as it affects entry and exit. The removal of price controls and import licences and the rationalization of tax tariff regimes are viewed positvely by manufacturers, who are nevertheless concerned about increased corruption and wage costs. 6.5 Business Support Services There is a large assortment of institutions ostenSlDly seeking to support Kenyan industry through training, financial assistance, technology, export assistance, and other business related assistance, including general information. These are either governmental agencies, non-governmental organizations, business organizations, or parts of donor programmes. But, only 30 firms received general business information in the past year, while 21 received export assistance, and 19 some form of training. Even fewer had received other business services, including six who received technology assistance and only two financial assistance. 14 13%-20% of textile and metal firms consider severe the union rules against lay-offs. trade union rules against firing workers, government rules against firing workers. and the cost of firing workers. 150/0 of metal firms consider government rules against laying-off workers severe. 144 This limited use ofbusiness services varies considerably between formal and informal firms. Formal firms have better access than informal ones, with most use of general business information, export assistance, training, and technology information being reported by formal firms. None of the informal firms reported any use of technology or export assistance, while only 3.6% and 2.7% ofthese :firms reported access to training and general business information, respectively (Table 6.18). Table 6.12 Use by formal and informal fIrms of various business support services, by type of service Business support services Formal Informal ALL Training assistance 46.2 3.6 28.4 Financial assistance 0.3 1.1 0.6 Technology assistance 38.1 0.0 22.0 Export assistance 50.9 0.0 29.8 General business information 62.2 2.7 37.7 Most .financial assistance went to informal metal-working firms, through special government programmes targeted to thejua kali sector (Table 6.19). None of the firms from other sectors had access to any financial assistance. Regarding other services, the food sector performed significantly better than the other sectors with 60-70% of its firms receiving training, technology assistance, export assistance and general business information. Less than 20% offirms from the other sectors received assistance in most of these areas. Most of the assistance is concentrated among the large firms, which participate in business organizations intended for helping their members and lobbying the government regarding specific needs. The majority of the recipients of these services found them use:ful. Large firms are better able to successfully seek out business assistance. 145 Table 6.13 Use by manufacturing firms of various business support services, by sector ·· service Food Wood Textiles Metal Training assistance 65.3 6.8 5.1 11.8 Financial assistance 0 0 0 3.2 Technology assistance 59.7 1.4 1.4 0.4 Ex.port assistance 69.9 4.7 7.1 11.8 General business information 61.9 18.5 32.9 17.9 Table 6.14 Weighted percentages offlrms considering business support services as useful, by sector, category, and type of service II .., II Seetor/Cate~OJY 1 t:'. Service 2 Food 12.9 100.0 100.0 Wood 93.3 99.1 - Textiles 100.0 100.0 100.0 Metal 100.0 100.0 24.4 Formal 49.4 99.9 97.1 Informal 100.0 99.9 97.1 All 51.9 99.9 97.1 These results confirm an earlier impression of limited use by Kenyan manufacturers of business support services, and the limited impact of the large number of business support organiz!ltions. The majority of recipients of these services find them useful, but attention needs to be paid to increasing the impact of these organizations. They are in dire need of expansion and strengthening. 146 6.6 Summary and Policy Concerns Significant ground has been gained in improving the regulatory environment with the removal offoreign exchange and price controls, and with rationalization of tariffs and taxes. Restrictions on activities, on Iay-oirs and on fum closure are of little concern to the majority of manufacturers in Kenya. However, infrastructure services continue to deteriorate. Conuption and poor infrastructure therefore remain major obstacles to the development of industry in Kenya. Further attention will have to be paid to improving the financing and management ofpubJic infrastructure, to reducing the rent-extraction opportunities on which corruption thrives, and to strengthening service delivery from the business support institutions. 147 APPENDIX6A Table 6Al Weighted percentages of formal and informal sample rll'ms perceiving changes in infrastructure, by type of service " FORMAL INFORMAL Infrastructure service No No Improved Deteriorated Change Improved Deteriorated Change Electricity 4.2 61.8 34.0 14.8 33.3 52 Water 1.3 27.3 71.4 22 18.4 79.4 Freight transport 21.7 8.0 70.2 7.7 49.0 43.3 Worker's transport 1.0 25.6 73.4 3.1 22.6 74.2 Roads 1.2 80.5 18.4 16.0 35.9 48.1 Telephone 6.4 63.8 29.8 2.9 13.9 83.2 Air- and seaports 2.3 83.7 14.0 0 13.6 86.4 Waste disposal 2.6 23.1 74.3 24.1 24.4 51.4 Security 1.2 23.1 74.3 24.1 24.4 51.4 Note: For combined totals (formal and informal), see Table 6.1. 148 ·. Table 6AZ Dermitions of variables used in Probit estimations OUTPUT - Value of output in units GROW1H - Net increase in labor force CAPACI1Y - Capacity utilization = 100/(100+P1SQ39) TEXID = 1 for textile firms METALD = 1 for metal-working firms WOOD = 1 for wood-working firms ELDD = 1 when firm is located in Eldoret, otherwise O. NKRD = 1 when firm is located to Nakuru, otherwise O. MSAD = 1 when firm is located to Mombasa, otherwise O. SIZE 1 = 1 when number of employees is 1-5, otherwise O. SIZE2 = 1 when number of employees is between 6-20, otherwise O. SIZE3 = 1 when number of employees is between 21-75, otherwise O. SIZE4 = 1 when number of employees is between 76-500, otherwise O. SIZE5 = 1 when number of employees is more than 500, otherwise O. 149 Table 6A3 Acquistion and renewal of licences by percentages of sample, by sector Licence Food Wood Textile Metal ALL l.ObtaiDed: Occupational health 0.4 0 0 0 0.1 Trade licence 0.3 14.2 3.2 0 3.6 General import licence 0.1 2.6 1.0 4.3 1.6 Specific import licence 1 2.3 6.1 1.2 1.6 2.5 Specific import licence 2 1.0 4.0 0 0 0.5 Specific import licence 3 1.3 0.4 1.2 0 1.0 Specific import licence 4 0.3 0.4 1.2 1.7 0.8 Other licence 1 13.6 18.8 3.6 2.3 8.9 Other licence 2 34.1 5.3 0 0 9.7 Other licence 3 14.5 3.7 0 0 3.2 2.Renewed: Occupational health 11.8 7.0 40.8 0 17.1 Trade licence 96.2 60.1 34.6 50.6 64.6 General import 7.5 20.1 22.6 16.4 15.9 Specific import 1 3.8 5.1 15.2 24.2 12.3 Specific import 2 3.2 0 1.4 0.7 2.9 Specific import 3 1.4 0 0 0 0.6 Other licence 1 70.1 67.4 68.2 64.6 67.8 Other licence 2 58.3 62.9 83.8 47.1 71.5 Other licence 3 67.9 26.4 78.9 0 66.0 Note: For combined totals (acquired and renewed), see Table 6.11. 150 Table 6A4 Severity of problems caused by various regulations as perceived by weighted percentages of sample fums Nota Slight Moderate Large Severe Regulations problem problem problem problem problem Ownership 97.9 0.9 0.1 0 1.1 Taxes 44.9 S.9 33.6 14.2 1.4 Restrictions on activities 83.4 1.3 2.3 9.1 3.9 Investment benefits 80.1 1.8 25 13.7 1.9 Labor costs 30.9 40.2 14.5 13.7 0.7 Labor regulations 67.2 4.3 15.6 11.4 1.6 Obtaining licences 79.8 6.3 3.8 5.5 4.7 Corruption 27.1 11.4 14.2 40.7 6.6 Price controls 92.4 1.5 1.9 3.2 1.0 Other regulations 54.4 39.1 4.5 1.2 0.8 151 Table 6ASa Weighted percentage of sample fll'ms perceiving slight or moderate problems with various regulations, by sector Large problem Severe problem Regulations Food Wood Textile Metal Food Wood Textile Metal Ownership 0.1 0.0 0.4 4.9 0.2 0.0 0.3 0.0 Taxes 10.4 3.0 0.5 0.1 69.0 1.5 6.9 16.6 Restrictions on activities 1.7 0.6 0.5 2.6 0.0 3.9 0.3 10.7 Investment benefits 0.3 1.2 3.2 4.5 0.2 0.8 1.5 12.6 Labor costs 67.5 5.0 53.7 8.8 7.1 20.6 14.5 21.9 Labor regulations 2.1 5.0 6.5 5.6 1.6 23.7 36.1 10.2 Obtaining licenses 9.2 2.5 5.1 4.8 2.6 4.3 3.1 7.6 Corruption 5.5 17.2 20.3 4.6 11.1 9.9 8.7 33.4 Price controls 0.5 0.0 0.0 6.8 0.6 5.6 0.7 0.0 152 Table 6A5b Weighted percentages of sample fmns perceiving large or severe problems with various regulations, by sector Large problem Severe problem Regulations Food Wood Textile Metal Food Wood Textile Metal Ownership 0 0 0 0 0 0 3.9 0 Taxes 4.7 19.5 23.8 17.5 0.4 3.3 0.2 4.5 Restrictions on activities 0 2.0 30.9 1.5 0.1 16.8 3.8 0 Investment benefits 0 23.9 34.1 7.5 1.2 2.0 4.2 0 Labor costs 4.8 27.9 12.9 17.6 0.4 1.7 0.5 0.4 Labor regulations 1.7 0.5 30.6 15.5 0.2 1.6 4.8 0 Obtaining licenses 3.1 9.9 7.5 2.9 0.1 15.9 4.5 4.1 Corruption 63.6 27.1 36.0 13.7 2.6 19.2 1.9 8.3 Price controls 8.1 0 0.4 3.1 1.2 0.4 0 2.8 153 Table 6A6 Weighted percentages of formal sample rll'ms percepting changes in various regulations, by sector Improved Worsened General Regulations Food Wood Textiles Metal Food Wood Textiles Metal Restrictions on activities 5.3 8.9 19.2 29.6 0.3 3.0 3.1 4.8 Capital requirements 11.6 4.9 3.7 12.5 0.2 0.0 1.4 21.0 Joint venture restrictions 4.0 5.8 2.0 24.4 1.3 0.0 0.0 2.7 Access to domestic finance 11.7 7.7 50.7 19.7 0.0 0.0 13.8 9.5 Repatriation of profits 40.5 35.0 72.2 39.9 0.0 3.4 0.0 0.0 Foreign excbange for 87.8 44.3 75.5 52.7 0.0 1.6 0.0 6.0 business travel Restrictions on foreign loans 26.6 5.5 65.8 49.6 0.0 QO 0.0 0.0 Payment of non-resident fees 85.6 5.4 63.4 46.2 0.0 0.0 0.0 0.0 Payment of technology licences 33.4 27.9 67.1 33.1 5.2 0.0 0.0 2.0 Regulations Restricting Production Reduction: Trade union lay-off rules 1.1 2.8 4.8 1.7 0.5 14.0 5.8 5.8 Other lay-off rules 0.1 9.7 4.2 2.8 0.8 7.8 5.3 9.3 Government lay-off rules 0.0 19.9 0.0 0.0 4.0 9.1 9.1 14.2 Cost oflay-offs 0.3 0.0 0.0 3.3 0.3 0.0 8.8 0.0 Closure Related Rules: Selling of enterprise restrictioos 1.0 0.0 6.3 5.0 1.6 0.1 0.3 1.7 Bankruptcy process 0.0 0.2 0.0 4.5 4.0 6.9 2.1 0.0 Government rules against firing 66.2 2.4 2.4 0.0 4.3 6.8 6.8 9.7 Trade union rules agains 0.3 0.0 0.8 0.0 4.9 11.1 8.7 12.0 firing Cost of firing workers 0.0 9.1 0.0 0.0 5.4 17.8 14.0 24.8 154 Table 6A7a Weighted percentages of formal sample firms perceiving slight or moderate problems with various regulations, by sector Slight Moderate General Regulations Food Wood Textiles Metal Food Wood Textiles Metal Restrictions on activities 0.4 0.0 6.5 2.2 1.0 1.2 0.5 4.1 Capital requirements 0.0 0.0 0.0 2.2 1.1 1.1 2.4 0.0 Joint venture restrictions 0.0 0.0 3.7 0.0 0.0 0.0 0.0 0.0 Access to domestic finance 0.1 0.0 3.7 2.4 0.1 0.1 5.6 0.0 Repatriation of profits 0.0 0.0 4.0 1.8 0.0 0.0 0.0 0.0 Foreign exchange for business travel 0.0 1.6 3.5 8.1 0.0 0.0 0.0 0.0 Restrictions on foreign loans 0.0 0.0 3.9 0.0 0.0 0.0 0.0 0.0 Payment of non-resident fees 0.0 0.0 3.7 0.0 0.0 0.0 0.0 0.0 Payment of technology licences 0.0 0.0 4.0 0.0 0.0 0.0 0.0 1.7 Regulations Restricting Production Reduction: Trade union lay-off rules 2.7 10.1 5.1 16.8 0.8 8.1 37.2 9.8 Other lay-off rules 1.8 13.6 4.6 3.1 0.9 3.3 32.3 20.5 Government lay-off rules 70.1 12.6 5.1 1.9 5.4 6.5 11.2 9.5 Cost oilay-offs 0.0 11.6 48.6 6.2 4.8 1.8 0.0 4.1 Closure Related Rules: Selling of enterprise 0.0 6.4 37.9 2.4 70.2 0.8 0.0 9.4 restrictions Bankruptcy process 0.2 6.6 37.8 3.1 4.0 8.0 1.6 9.2 Gove~ent rules against 0.2 8.5 36.9 8.8 0.5 7.6 0.3 20.9 firing Trade union rules agains firing 69.5 8.2 36.7 1.3 1.5 10.9 2.2 21.0 Cost offirin~ workers 0.0 15.2 37.4 0.8 1.7 2.5 4.3 23.2 155 .' Table 6A7b Weighted percentages of formal sample rlJ'ms perceiving large or severe problems with various regulations, by sector Large Severe General Regulations Food Wood Textiles Metal Food Wood Textiles Metal Restrictions on activities 0.0 1.0 32.6 0.0 0.1 0.4 0.0 0.0 Capital requirements 0.2 0.0 0.0 0.0 0.0 0.0 0.0 5.5 Joint venture restrictions 0.0 0.0 0.0 1.6 0.0 0.0 0.0 0.0 Access to domestic finance 0.4 3.9 1.9 2.4 0.0 0.0 10.1 1.4 Repatriation of profits 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Foreign exchange for business travel 0.0 0.0 4.1 0.0 0.0 0.0 0.0 0.0 Restrictions on foreign loans 0.0 0.0 0.0 5.1 0.0 0.0 4.6 0.0 Payment of non-resident fees 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Payment oftechno]ogy licences 0.0 0.0 0.0 1.7 5.0 0.0 0.0 0.0 Regulations Restricting Production Reduction: Trade union lay-off rules 1.4 5.6 26 3.2 0.0 4.8 17.8 19.5 Other lay-off rules 1.5 7.1 6.0 0.0 0.1 0.4 9.3 14.6 Government Jay-off rules 1.4 3.4 2.5 5.7 0.0 0.7 5.8 7.5 Cost oflay-offs 0.0 0.0 0.0 12.4 0.0 0.0 6.1 0.0 Closure Related Rules: Selling of enterprise restrictions 0.0 0.3 3.9 0.0 0.0 0.0 0.0 8.2 Bankruptcy process 1.5 1.1 4.6 1.9 0.2 3.3 0.3 0.7 Govel'lllI!ent rules against firing 5.0 2.2 5.1 0.0 0.2 7.1 14.3 14.3 Trade union rules agains firing 5.1 1.7 4.4 7.5 0.0 7.0 14.9 14.4 Cost of firing workers 5.4 2.9 5.4 7.4 0.2 8.4 15.7 13.1 156 7. IMPACf OF ADJUSTMENT 7.1 Introduction After setbacks related to the elections in late 1992, the process of macroeconomic stabilization and structural adjustment was resumed in 1994. Inflation had gotten out of hand, and the main problem for the authorities in 1993·94 was to mop up the large excess liquidity through sales of treasury bills (Levin, Ndung'u,1994). Foreign reserves were essentially depleted at the beginning of 1993, and a lot of effort also had to be devoted to building these up again. The mopping up operation, and the use of treasury bills to finance the budget deficit, caused interest rates to be very high for an extended period of time. It was not until April 1994 that the money supply growth started to decline. Meanwhile, high interest rates attracted more capital from abroad than expected, which caused the shilling to appreciate. The cost of redeeming treasury bills now add to the pressure on the government in tIying to meet its fiscal obligations. There is a risk that the government will start to print money to cover these obligations, but this would definitely undermine the credibility that the government has been able to build up in the last year. There has been talk about using donor funds to redeem debt, but it seems more appropriate to use funds from the privatization program for this purpose. OtheJWise, there is a risk that the government will relax fiscal discipline, especially as the 1997 elections get closer. All these pOSSIbilities and risks have to be taken into consideration by Kenyan firms, when they make investment decisions and decide on their future business strategy. This chapter analyzes the firms' perceptions and experiences of the adjustment policies pursued. We asked about structural changes in the economy during the last year, and about the severity of problems in various areas. These included export opportunities, access to foreign exchange, cost offoreign exchange, competition from imports and from local firms, demand, utility prices, infrastructure and so on. Firms were also asked to rank the severity ofthe problem they experienced with each business factor. First, we discuss the results for each factor, and then we conclude with a comparison among them 157 7.2 Export Opportunities Fmns were asked whether export opportunities had changed. and whether they considered them to be a problem (see Table 7.1). The majority of:firms indicated that there was some improvement compared to the previous year, and only 3% indicated that export opportunities had worsened. Informal:firms did not report much improvement in export opportunities, but they did not report any problem in that regard either, presumably because they do not export. The vast majority offonnal :firms reported large improvement in export opportunities during the year, and most (85%) did not find this area to be a problem, but, as we noted earlier, veIY few ofthem actually export, either. The answers can therefore not be taken to mean that Kenya has had a breakthrough in exports, but they could at least indicate that the firms believe that obstacles to exports have been reduced. The opportunity is now there, if the :firms are competitive. Table 7.1 Export opportunities ~ Wood Textiles Metal FOIDlal Informal ALL No change 6.6 34.2 39.9 43.2 7.0 80.5 25.6 Improved 91.6 63.7 59.9 46.0 89.1 19.5 71.5 Worsened 1.7 2.1 0.3 10.7 3.9 - 2.9 Not a problem 93.1 88.1 94.8 68.9 85.2 100.0 88.8 Slight problem 4.1 6.6 2.5 10.3 6.7 - 5.0 Moderate problem 1.3 2.2 1.3 13.7 4.6 - 3.5 Large problem _ ,. ... 0.2 2.9 - 0.6 0.8 - 0.6 i ~"v"',... 1 ':I 0' 1 '" ~'" "'7 - ') 1 It is striking that so few textile firms found that export opportunities were a problem This is highly swprising, given that the lack of progress on exports is one of the most serious problems for Kenyan industry, especially textiles. This may mean that most textile :firms do not even consider export markets as an option for them 158 7:3 Access to and Cost of Foreign Exchange Foreign exchange availability had been a problem in Kenya for some time, most critically between 1989 and 1993. To discover whether policies introduced since early 1993 might have improved the situation, :firms were asked whether availability of foreign exchange had changed and they were asked to rank the severity of the problem (Table 7.2). 72% of firms reported that access to foreign exchange had improved dramatically; 89% indicated that foreign exchange availability was no longer a problem Still, 27% of firms in the textile sector, as well as a scattering of firms in the other sectors, did indicate that access to foreign exchange remained a problem The overall impression is that the foreign exchange situation has been drastically improved, and that firms were generally finding little problem in this regard. Table 7.2 Access to foreign exchange Food Wood Textiles Metal Formal Informal ALL Nocbange 7.2 26.9 41.8 41.6 8.2 75.8 25.3 Improved 92.5 67.8 58.2 49.1 88.6 24.2 72.3 Worsened 0.2 5.3 - 9.3 3.2 - 2.4 Not a problem 98.5 90.4 72.6 89.6 92.9 75.2 88.8 Slight problem 0.2 2.4 23.3 9.1 3.3 24.8 8.2 Moderate problem - 2.9 4.1 1.3 2.3 - 1.7 Large problem - 1.3 - - 0.2 - 0.2 Severe problem 1.3 3.0 - - 1.3 - 1.0 Most informal firms indicated no change in the availability of foreign exchange, but it was not a problem, as very few deal with foreign exchange. In genera~ it was the formal firms actively dealing with foreign exchange, and thus previously hampered by the foreign exchange squeeze, which indicated an improvement. Thus, availability of foreign exchange is not now perceived as a factor inhibiting enterprise activities or growth. 159 Fmns were also asked to assess the cost offoreign exchange (Table 7.3). The exchange rate was highly unstable at this time and the responses reflect this. Roughly half the firms respondents reported that the cost of foreign exchange had not changed during the year. Relatively few feh that the cost of foreign exchange was a problem. Complaints were mostly confined to the fonnal sector, as we would expect. Since the time of the interviews, the shilling has appreciated another 20%, making imports even cheaper. Table 7.3 Cost of foreign exchange Food Wood Textiles Metal Formal Informal ALL No change 71.9 30.9 37.1 428 44.6 74.3 51.7 Improved 17.7 19.7 16.6 38.3 25.5 6.5 20.9 Worsened 10.4 49.4 46.3 19.0 30.0 19.2 27.4 Not a problem 85.8 89.8 47.8 50.4 63.5 91.7 70.3 Slight problem 4.6 3.5 . 9.0 5.2 . 3.9 Moderate problem - 1.3 .13.1 26.5 10.7 - 8.1 Large problem 8.2 4.7 35.0 12.0 17.9 8.3 15.6 Severe problem 1.3 0.7 4.1 2.1 2.8 - 2.1 7.4 Competition As we noted earlier, Kenyan firms have been highly insulated from foreign competion, and they are still highly protected. But with the continuation of the h'beralization process, and with appreciation of the CUITeIlcy, one would assume that many firms would face increasing compe,tition from abroad. Many firms in fact complained about increased competition from imports during the last year (Table 7.4). The problem is considered severe by quite a number offirms, especially in the textile sector. Informal firms, of comse, are less affected by foreign competition than are formal ones. 160 Table 7.4 Competition from imports Food Wood Textiles Metal Formal Informal ALL Nocbange 54.6 72.6 37.1 55.5 31.6 82.9 50.9 Improved 2.5 1.8 5.2 8.4 6.9 1.2 4.8 Worsened 42.9 25.5 57.7 36.1 61.5 15.9 44.4 Not a problem 58.4 92.7 41.3 59.6 37.2 93.2 58.3 Slight problem 2.1 0.7 1.1 11.8 5.7 - 3.6 Moderate problem 30.8 4.0 13.3 7.0 21.3 - 13.3 Large problem 8.7 1.3 36.7 10.4 27.2 5.7 19.1 Severe problem - 1.3 7.6 11.3 8.6 1.1 5.8 As the playing field is leveled with hoeralization of the economy,· we might also expect greater competition among local firms, so we asked the firms about this (Table 7.5). A majority of all firms, and over 60% of formal firms, indicated that competition had increased. However, over 70% of the firms indicated that local competition was either no problem or on1y a slight problem. The general impression from the previous round thus remains: Competitive pressure in Kenya is low. Table 7.5 Competition from local firms I Food Wood Textiles Metal Fonnal Informal ALL Nocbange 17.5 36.4 28.3 68.3 26.9 44.4 33.1 Improved 6.4 3.0 28.2 10.7 U.2 15.1 12.6 Worsened 76.1 60.6 43.5 21.0 61.8 40.4 54.3 Not a problem 11.9 51.1 75.5 50.7 39.7 48.2 42.7 Slight problem 69.9 0.8 4.4 9.3 44.5 2.0 29.4 Moderate problem 6.1 23.0 3.4 20.7 4.3 23.3 11.0 Large problem 8.5 10.1 10.6 13.8 4.0 21.8 10.4 Severe problem 3.6 14.9 6.2 5.4 7.5 4.6 6.5 161 7.5 Access to and Cost of Imported Raw Materials Fnms were asked 'Whether there was any improvement in access to imported raw materials, and whether they still considered this a problem, as many did previously, before the liberalization offoreign exchange transactions. On average, 71 % of the firms indicated that there was no change, while 24% reported that the situation had improved (Table 7.6). The vast majority (80%) offirms indicated that access to imported raw materials was not a problem It was mainly the textile sector that complained about access, but no textile firm considered the problem severe. Differences between re~onses of formal and informal firms were rather small, perhaps indicating that informal firms also depend on imported inputs. Table 7.6 Access to imported raw materials Food extiles Metal Fonnal Informal ALL No change 81.5 67.4 70.8 46.5 68.7 75.3 70.6 Improved 15.1 31.6 20.1 44.8 26.7 17.3 24.0 Worsened 3.4 1.0 9.1 8.7 4.6 7.5 5.4 Not a problem 95.4 78.9 55.8 85.1 80.5 80.2 80.4 Slight problem 0.4 15.5 1.3 3.2 1.5 8.8 3.7 Moderate problem 3.6 4.7 37.7 8.0 17.1 4.9 13.5 Large problem 0.2 0.9 5.2 3.7 0.6 6.0 2.2 Severe roblem 0.4 0.2 0.2 With the sudden depreciation ofthe shilling in 1993, the cost of imported raw materials had risen steadily. To get an idea 'Whether firms perceived it as a hindrance to enteIprise develo~ent or operations, firms were asked to report 'Whether costs had changed, and the severity of the problem About 40% of firms reported that the cost had increased. It was particularly the textile sector again that complained about the problem Again, informal firms seemed to be affected almost as much as the formal firms. 162 Table 7.7 Cost of imported raw materials Food Wood Textiles Meta1 Formal Informal ALL No change 49.4 54.6 42.7 34.4 31.1 61.3 44.8 Improved 29.3 5.8 7.1 24.6 22.6 15.4 Worsened 21.3 39.7 50.1 41.0 40.3 38.7 39.8 Not a problem 67.8 81.4 29.7 52.8 48.2 61.3 53.2 Slight problem 4.8 0.7 2.6 5.9 4.3 1.8 3.4 Moderate problem 8.2 62.1 21.5 33.3 23.1 29.4 Large problem 19.2 9.0 2.8 14.4 12.2 6.2 9.9 Severe problem - 8.9 2.7 5.4 2.0 7.6 4.1 7.6 Opportunities to Buy Foreign Machinery and Equipment Firms were asked whether it had become easier to buy foreign machinery and equipment which was unavailable before the reform program. The situation has apparently improved, particularly for the food sector (Table 7.8). The informal sector saw less improvement, however. The majority of firms indicated that accessto foreign machinery it was not a problem. Only 11% offirms feh that it was a large or severe problem, more so in the wood and textile sectors. Table 7.8 Opportunities to buy foreign machinery and equipment ood Wood Textiles Metal Formal Informal ALL Nocbange 19.0 51.4 53.0 47.9 23.0 78.9 31.8 Improv~ 19.0 34.8 46.1 51.7 14.5 15.4 58.9 Worsened 2.0 13.8 0.9 0.4 2.5 5.7 3.4 Not a problem 94.0 78.5 76.5 85.4 90.6 71.4 85.5 Slight problem 1.5 0.8 - 11.2 3.4 - 2.5 Moderate problem 0.3 2.0 1.2 2.6 1.6 - 1.2 Large problem 4.2 16.8 22.3 0.8 4.0 28.6 10.5 Severe--problem . 2.0 - - 0.4 - 0.3 163 7.7 Business Support Services In view ofthe presumed importance ofbusiness support services in enterprise development, the sample firms were asked whether these services had improved over the adjustment period, and whether they considered lack of these services a problem for their activities. About halfthe firms indicated that there was no change in business support services, though 60% of formal firms, especially in the food sector, indicated that there had been an improvement during the year (Table 7.9). Infonnal firms complained somewhat more about the lack ofbusiness support services than did formal firms. But most offirms reported that the lack ofbusiness support services was not a problem (81 % offormal firms). Perhaps this reflects the fact that firms provide their own services when they cannot get service from other sources. This may partly answer the apparent contradiction as to why the majority offirms indicated that business support services had not changed in the adjustment period, yet the majority also indicated that there was not a problem Table 7.9 Business support services Food Wood Textiles Metal Formal Informal ALL No change 26.3 71.7 52.0 65.7 35.0 71.5 48.4 Improved 67.7 20.5 44.6 21.4 59.4 19.5 44.8 Worsened 6.0 7.8 3.4 12.9 5.5 9.0 6.8 Not a problem 82.4 56.8 85.4 58.1 81.2 61.3 73.9 Slight problem 9.0 6.7 1.3 13.4 7.5 7.1 7.3 Moderate problem 5.7 12.6 11.1 19.2 10.4 11.4 10.8 Large problem 2.9 17.0 0.1 5.6 0.8 13.5 5.5 Severe problem - 6.8 2.2 3.7 0.1 6.8 2.5 " 7.8 Infrastructure With the budget squeeze, the government has not been able to maintain the infrastructure already in place, Dnlch less to extend it further. The deterioration of roads, telephones, and 164 other infrastructure &ciIiti.es was evident everywhere the study team visited. The adjustment program has brought about some beneficial changes in prices and regulations, but infrastructure is dilapidated, worse than ever. We asked the firms what changes they saw and how the state of infrastructure affected their operations (Table 7.10). The results confirm that infrastructure had worsened. About 91% of the formal firms and 50% ofthe informal firms reported that infrastructure services had worsened. All sectors considered the problem large, but again especially the food sector, in which over 70% of fums said that the problem was severe, probably due to the perishable nature of their raw materials and products. The wood sector also experienced considerable problems, due to poor access to logging areas. Table 7.10 Infrastructure Food Wood Textiles Metal Formal Informal ALL Nocbange 11.1 20.3 34.6 21.2 5.8 45.9 20.9 Improved ~ - 2.9 13.7 3.0 3.7 3.3 Worsened 88.9 79.7 62.4 65.1 91.2 50.4 75.8 Not a problem 4.3 24.6 30.1 24.0 4.4 41.8 18.6 Slight problem 10.3 3.9 12.6 17.8 7.6 16.8 11.1 Moderate problem 6.7 19.0 31.4 16.5 19.1 14.5 17.3 Large problem 8.0 26.8 19.9 32.1 18.3 20.5 19.2 Severe problem 70.7 25.7 5.9 9.6 50.7 6.4 33.8 7.9 . Utility Prices ' Another consequence of the fiscal squeeze is that utility prices have gone up; there are now widespread complaints about higher utility prices, on a scale comparable with those about infrastructure (Table 7.11). 90% of firms think that utility prices have worsened, and 90% report that the high cost is a problem for their operations. The worst hit sector was food, 165 where only 3% report that utility prices are no problem Unfortunately, higher prices have not yet led to better service, and the combination of higher prices and poor service is undoubtedly a hindrance to growth. Table 7.11 Utility prices II II "' .... _. Wood Textiles Metal Formal Informal ALL Nocbange 0.6 9.2 15.2 15.7 1.7 21.3 8.8 Improved 2.4 0.7 - - 0.2 2.4 1.0 Worsened 97.0 90.1 84.8 84.3 98.1 76.3 90.2 Not a problem 3.0 21.3 7.2 20.0 1.2 26.8 10.7 Slight problem 5.6 6.9 28.6 18.1 4.9 30.1 14.3 Moderate problem 5.0 11.7 6.8 25.8 9.9 12.4 10.8 Large problem 74.2 35.3 49.7 26.2 69.4 21.0 51.4 Severe problem 12.2 24.8 7.7 9.9 14.6 9.7 12.8 7.10 Product Demand Finally, firms were asked whether there was any improvement in demand for their products in the adjustment period (7.12). A majority of the firms (58%) indicated that product demand had faDen, especiaDy food and wood firms. The metal sector seemed least affected. Three-quarters of the formal firms also indicated that lack of demand was a problem, ranging from a slight problem for 52%, to a moderate problem for 6%, a large problem for 6% and a severe problem for 12%. Many more informal firms than formal ones reported it to be.a moderate, large, or severe problem, though a much greater proportion also said it was not a problem at all. The low level of demand revealed by the snrvey is consistent with the national perception of recession at the time. The adjustment program may thus have been hannful to growth in the short run. · 166 Table 7.12 Product demand Food Wood Textiles Metal Formal Informal ALL Nocbange 9.3 18.5 21.0 28.3 11.3 27.9 17.8 Improved 20.7 14.2 28.7 31.8 20.4 29.1 23.8 Worsened 70.0 67.3 50.3 39.9 68.3 43.0 58.4 Not a problem 21.4 34.8 40.3 35.8 23.3 44.9 31.6 Slight problem 66.3 0.6 29.9 7.7 52.1 2.4 33.1 Moderate problem 5.5 10.2 3.1 21.3 6.3 13.0 8.9 Large problem 2.1 36.1 8.6 15.6 6.4 23.3 12.9 Severe problem 4.6 18.3 18.1 19.6 11.8 16.4 13.6 7.11 Overall Impact of Adjustment Table 7.13 summarizes all the results above. Infrastructure is the area rated as the most severe problem by the most firms (34%), followed by product demand and utility prices. Most ofthe other areas had improved, or the firms were more able to do something about them internally, so they were not ranked as major problems. But the restrictive policies have clearly caused a worsening of some external problems, such as infrastructure, demand, and utility prices. ... 167 · Table 7.13 Summary of responses PROBLEM No Nota Slight Moderate Large Severe change Improved Worsened problem problem problem problem problem Export opportunities 25.6 71.5 2.9 88.8 5.0 3.5 0.6 2.1 Access to foreign exchange 25.3 72.3 2.4 88.8 8.2 1.7 0.2 1.0 Cost of foreign exchange 51.7 20.9 27.4 70.3 3.9 8.1 15.6 2.1 Competition from imports 50.9 4.8 44.4 58.3 3.6 13.3 19.1 5.8 Competition from local fIrms 33.1 12.6 54.3 42.7 29.4 11.0 10.4 6.5 Access to imported raw materials 70.6 24.0 5.4 80.4 3.7 13.5 2.2 0.2 Cost of imported raw materials 44.8 15.4 39.8 53.2 3.4 29.4 9.9 4.1 Opportunities to buy foreign machinery & equipment 37.8 58.9 3.4 85.5 2.5 1.2 10.5 0.3 Business support services Infrastructure 4S.4 44.8 6.8 73.9 7.3 10.S 5.5 2.5 Utility prices 20.9 3.3 75.S IS.6 11.1 17.3 19.2 33.8 Product demand 8.8 1.0 90.2 to. 7 14.3 10.8 51.4 12.8 17.8 23.8 5S.4 31.6 33.1 8.9 12.9 13.6 168 : .. We also asked the firms to rank their three biggest problems last year (Table 9: 14). 20% ranked lack of credit as the number one problem, while lack of demand was mentioned by 12%, and inadequate infrastructure by 5%. Competition from imports and lack of skilled labor were each mentioned by 3%. Inadequate infrastructure tops the list of secondary problems at 30%, followed by lack of credit at 11 %, and inadequate demand for products at 7%. As the third problem, high utility prices were mentioned by 32%, while infrastructure, credit, and demand were again mentioned by significant numbers of firms. The firms seem most concerned with basic conditions of production, that is, with credit, infrastructure, utility prices and demand. These resuhs tally with the summary of responses given above. While basic infrastructure was crumbling, utility corporations were hiking prices to cover their inefficient activities, and the whole economy was undergoing a recession. The industrial sector was getting squeezed. · 169 ·" Table 7.14 Biggest problems last year Food Wood Textiles Metal Formal Informal ALL Ownership regulations 0.8 - - - 0.5 - 0.3 Inadequate infrastructure 2.4 4.8 0.2 14.8 4.9 4.2 4.6 Utility prices 1.9 - - - 0.2 1.3 0.7 Lack of credit 7.7 14.9 32.1 28.4 7.9 36.9 19.7 Inadequate demand 1.8 30.4 12.5 13.1 4.6 23.1 12.1 Access to foreign exchange - - - 0.5 0.2 - 0.1 Competition from imports 4.0 0.8 3.8 2.2 4.2 1.2 3.0 Competition from local firms 5.0 0.2 5.6 - 3.9 2.4 3.3 Uncertainty about govt. policies Lack of skilled labor - 1.1 - 7.4 2.8 - 1.7 4.5 4.9 - 2.7 1.8 4.7 3.0 Taxes Government restrictions on 4.9 2.0 - 0.9 3.8 - 2.3 activities - - 0.3 2.9 0.1 1.4 0.6 Labor regulations - 1.0 - - 0.3 - 0.2 Corruption 0.2 0.1 0.8 - 0.5 - 0.3 Price controls Lack of business support 2.5 - - - 0.1 2.0 0.9 services - 3.5 4.0 - 1.8 1.6 1.7 Other 64.3 36.4 40.8 27.0 62.3 21.1 45.5 The lastlssue in the adjustment questionnaire was investor confidence for the future (Table 7.15). We asked the firms about their expectations of future sales and about how they expected access to and the cost of credit and foreign exchange to change over the next year and the next three years. They expected things to improve! Most firms expected sales to be higher, access to credit to be easier, cost of credit to be lower, access to foreign exchange to be easier, and cost of foreign exchange to be lower in the future. This may be understandable, because the recession and the impact of the short-run reallocation of · 170 resources should eventually ease. But it is reassuring to know that the business comnumity has confidence that structural adjustment policies will work and that the economy will be stimulated, thus creating a better environment for enteIprise growth and development. Table 7.1S Investor confidence Lower Same Higher N/A Expectations of sales next year 26.3 12.2 61.5 - - three years to come 21.9 9.2 68.9 - Expectations of access to credit next year 23.0 27.2 49.7 - - three years to come 23.1 22.5 54.4 - Expectations of cost of credit next year 41.5 16.5 42.0 - - three years to come 47.9 13.8 38.3 - Expectations of access to Forex next year 19.2 17.9 53.0 9.9 - three years to come 23.1 15.6 51.0 10.2 Expectations of cost ofForex next year - three years to come . 48.0 16.2 27.7 8.1 47.1 13.8 29.7 9.4 · 171 8. FIRM GROWTH 8.1 Introduction In this chapter we conclude the analysis of this yeats smvey data by looking at the employment growth of our sample firms from 1990 to 1994. Employment growth can be seen as a summary measure of the impacts on the firms of all the factors discussed in this report. As descnbed earlier, this has been a turbulent period both economically and politically, with some stabilization towards the end. Our analysis is, of course, based on surviving:firms, which means that we cannot generalize to the sector as a whole. Still, the simple analysis undertaken can provide some clues about changes in the Kenyan manufacturing sector. We undertake two types of data analysis. First, we investigate how growth differs for different types offirms, and secondly, we estimate simple growth equations. The analyses we undertake are similar to those undertaken last year (for the period 1981-1992), but this time we focus on the most recent period. 8.2 Firm Employment Growth We start by looking at the growth ofall firms in our sample (Table 8.1). We asked about the levels of employment in 1990, 1991, 1992, and at the time of the interview, that is, mid- 1994. This means that the last period is almost two times longer than the other periods. Even allowing for this, we note that there seems to be an acceleration of employment growth in our firms. This may be due to some extent to underestimation of employment levels in the earlier years, but the increase is so marked that we may assume that there is an actual increase in the rate of employment growth. We must note, however, that there is a high degree ofvariability in the data. There is obviously no generalized expansion of firms, which is not what one would expect, either. In a process of structural adjustment, one · 172 would expect some :firms to do very wen. while others shrink and may be forced out of business. Table 8.1 Employment growth 1990-1994 Period N Mean Std C.Y. 1990-91 164 7.36 109.49 1488.3 1991-92 153 11.34 114.56 1009.8 1992-94 158 35.75 170.74 477.6 1990-94 151 12.49 66.28 530.5 Like our analysis of last yeats data, we again compared growth rates according to a few different classifications, namely: size of:firm, informal vs formal:firm, domestic vs foreign ownership, and etbnicity of owner. The most important comparison concerns growth rate by firm size (Table 8.2). We applied significance tests to check whether growth in one categOlY was significantly different from growth in the other categories combined (Prob>ltl>. The right-hand colunm shows a test of whether the variances are equal or not. In all cases except one, we find the variances to be unequal, and therefore we apply this assumption when testing. In last year's analysis of growth rates for 1981-1992, we found an inverted V-shape, that is, growth was highest in the intermediate size classes, while :firms in the smallest size category were actually sbrinking somewhat. We did not identify such a pattern for the period 199()"94. During this period there is no obvious systematic relationship between firm size and employment growth. We ariued last year that there was a tendency for informal firms to grow by proliferation, while each firm tended to remain small. The explanation advanced was that there were constraints against expansion, as well as incentives to remain outside the formal economic framework. We cannot confirm this result on the basis of the data for 1990-94. · 173 Table 8.2 Employment growth by fInD size 1990-1994 N Mean Std C.V. t df Prob>ltl F Prob>F Growth 1990-1991 Total 164 7.36 109.49 1488.3 1-5 34 1.89 180.72 9554.3 -0.521 36 0.6058 5.178 0.0000 6-20 46 19.59 124.90 637.4 0.845 68 0.4013 1.538 0.0334 21-75 45 13.86 45.04 325.1 0.496 161 0.6206 7.785 0.0000 76- 500 35 0.83 22.05 2657.8 -0.577 152 0.5649 31.111 0.0000 501 + 4 4.78 11.35 237.2 -0.247 31 0.8063 95.437 0.0002 Growth 1991-1992 Total 167 11.33 109.72 968.4 1· 5 35 10.51 201.15 1914.1 -0.073 36 0.9418 8.531 0.0000 6·20 46 13.03 76.. 24 585.3 0.153 127 0.8790 2.488 0.0003 21· 75 45 -0.90 7.30 ·811.0 -1.066 123 0.2886 308.784 0.0000 76- 500 37 32.11 98.40 306.5 1.110 65 0.2713 1.316 0.1678 501+ 4 12.60 12.27 97.4 0.120 25 0.9055 81.867 0.0003 Growtb 1991-1994 Total 174 35.60 162.82 457.4 1-5 38 32.12 229.44 714.2 -0.290 44 0.7730 2.725 0.0000 6-20 47 47.10 231.01 490.5 0.446 56 0.6575 3.240 0.0000 21-75 47 8.17 24.92 305.0 -1.617 137 0.1081 58.059 0.0000 76- 500 38 18.38 47.75 259.8 -0.998 171 0.3196 14.610 0.0000 501 + 4 9.43 8.39 89.0 -1.973 123 0.0507 385.026 0.0000 Growtb'I990-1994 Total 164 12.45 63.63 511.0 1- 5 34 9.87 101.47 1028.3 -0.436 37 0.6650 4.338 0.0000 6-20 46 19.14 79.24 414.1 0.746 62 0.4583 2.053 0.0010 21· 75 45 4.15 16.18 389.9 -1.173 143 0.2429 20.957 0.0000 76 - 500 46 4.39 12.80 291.7 -1.228 152 0.2215 31.179 0.0000 '\01 + .4 II <~ ~ .:10 40Q .0 7~1 114 0.:11'>1'>1 ~11/111 ')7 00000 174 Employment in the informal firms in our sample actually grew faster than employment in formal firms, but we failed to come up with significant estimates of this difference due to the high variability in the data. Domestic firms on average grew faster than firms with some foreign ownership, mainly during 1992-94, and African-owned firms grew faster than the Asian-owned firms during an three periods, but neither ofthese differences was statistically significant, either. The last result tends to undermine the view that Asian-owned firms have a growth advantage relative to Aftican-owned firms, however. In the simple estimates we have just described, we did not control for the influence of other variables. We therefore proceed with regressions of a few simple employment growth equations. In Table 8.3 we estimate a basic growth model with only age offirm and size of firm as explanatory variables. Again, these estimates confirm our observation above that, in the period considered here, there is no systematic relationship oetween firm size and growth. On the other hand, age of firm parameters tum out to be significant, though the parameters are small They give us a U-shaped profile, that is, younger and older firms grew faster than middle-aged ones. These resuhs for age and size are confirmed in our extended regression of a growth model with contro1s (Table 8.4). We introduced sectoral dummies for the food, textiles, and metal industries, dropping wood as the reference sector. We a1so introduced dummies for ethnicity of owner, informal vs formal, and domestic vs foreign ownership. The regression shows that the textile sector grew significantly slower than the wood sector during the period, which confirms that the textile industry has suffered during the adjustment phase, partly by the importation of second-hand clothes. The food sector a1so grew more slowly than the wood sector over the period; the explanation for this is less obvious. As with the statistical analysis descnbed above, the effects of ethnicity of owner, informal vs formal firm, and domestic vs foreign ownership are not very significant in the regressions. · 175 Table 8.3 Basic growth model. Dependent variable: Log(l+r), where r is employment · growth rate 1990-1991 1991-1992 ·.. Parameter TforHo: ......... :::n Prob>JTI Parameter w,'" ... TforHo: · ....... :::n Prob>JTI Intercept 0.100692 1.631 0.1048 0.164127 3.195 0.0017 Age of the firm ·0.013285 -1.632 0.1047 -0.015167 -2.273 0.0244 Age squared 0.000270 1.309 0.1925 0.000304 1.792 0.0750 Employment -0.000073 -0.092 0.9267 0.000177 0.265 0.7914 Employment sq. 6. 689E-8 0.147 0.8830 -6. 113E-8 -0.160 0.8728 Adj. R-square -0.0073 0.0091 F-test 0.705 0.5898 1.381 0.2428 1992-1994 1990-1994 Parameter TforHo: Prob>JT1 Parameter T for Ho: Prob>ITI Estimate Param.=O Estimate Param.=O Intercept 0.429916 6.765 0.0001 0.231745 6.425 0.0001 Age of the firm -0.028569 -3.380 0.0001 -0.019293 -4.055 0.0001 Age squared 0.000510 2.347 0.0201 0.000366 3.033 0.0028 Employment 0.000053 0.062 0.9508 0.000050 0.108 0.9138 Employment sq. -4.435E-8 -0.090 0.9282 -1.078E-8 -0.041 0.9616 Adj. R-square 0.0543 0.0810 F-test 3.485 0.0092 4.592 0.0016 · 176 Table 8.4 Growth model with Controls. Dependent variable: Log(J+r), where r is employment growth 1990-1991 1991-1992 Parameter T for Ho: Prob>jTI Parameter TforHo: Prob>IT/ ~ .. D ......... ~:a{l 11'. · .Pal"am~:a{l Intercept 0.521395 1.187 0.2372 0.078644 0.204 0.8385 Age of the finn -0.011919 -1.449 0.1497 -0.013083 -1.813 0.0720 Age squared 0.000274 1.364 0.1748 0.000300 1.666 0.0980 Employment -0.000588 ·0.366 0.7149 0.000298 0.201 0.8407 Employment sq. 0.000001 0.455 0.6500 -0.000000 -0.110 0.9129 Food industries -0.247365 -2.466 0.0149 ·0.013250 -0.147 0.8830 Textile industries -0.374587 -5.113 0.0001 -0.228542 -3.455 0.0007 Metal industries -0.018261 -0.213 0.8315 -0.042238 -0.567 0.5714 Asian -0.146771 -0.589 0.5571 0.151166 0.674 0.5014 European -0.353298 -0.393 0.6947 0.063477 0.078 0.9378 African -0.031487 -0.151 0.8800 0.217330 1.156 0.2495 Informal sector 0.006207 -0.151 0.9696 0.057856 0.396 0.6926 Domestic finn -0.200278 -0.545 0.5868 ·0.092086 -0.291 0.7718 Adj. R-square 0.1922 0.0847 F-test 3.975 0.0001 2.172 0.0161 (to be continued) . 177 1992-1994 1990-1994 Intercept -0.010148 -0.020 0.9843 0.200090 0.752 0.4532 I Age of the firm -0.035178 -3.747 0.0003 -0.020665 -4.148 0.0001 Age squared 0.000660 2763 0.0065 0.000416 3.425 0.0008 Employment 0.001699 0.870 0.3856 0.000382 0.392 0.6953 Employment sq. -0.000001 -0.724 0.4704 -0.000000 -0.203 0.8393 Food mdustries -0.207417 -1.725 0.0867 -0.161191 -2.653 0.0089 Textile industries -0.034668 -0.392 0.6958 -0.216098 -4.870 0.0001 Metal industries -0.148310 -1.507 0.1341 -0.074637 -1.439 0.1525 Asian 0.259968 0.864 0.3892 0.090762 0.601 0.5488 European 0.369279 0.338 0.7356 0.028137 0.052 0.9588 African 0.421928 1.670 0.0971 0.205969 1.634 0.1046 Infonnal sector 0.101329 0.531 0.5962 0.044528 0.453 0.6514 Domestic firm 0.049088 0.115 0.9083 -0.069715 -0.313 0.7547 Adj. R-square 0.0656 0.2077 F_t....t 1 Q1Q om(\"\ 417R 00001 8.3 Concluding Remarks The results on firm growth reported in this chapter suggest that there has been some recovery of the manufacturing sector in the last few years. Though firms complain about the impact of contractionary macroeconomic measures, they have benefitted from liberalization. We have not been able to identify any specific category of firm with regard to size or type of ownership which has suffered disproportionately. It seems, however, that the textile sector is suffering particularly from increased competition., and the food sector also seems to have problems. Still, structural change is an integral part of the reform process. The manufacturing sector as a whole has gradually improved its performance, at least with regard to employment generation., and this seems to suggest that policy reforms are beginning to have positive growth effects. · 178 9. CONCLUDING REMARKS The situation for Kenyan manufacturing firms was still difficult in mid-1994, when the I second round ofinterviews with our panel offums was undertaken. There were some initial signs ofmacroeconomic stabilization, but not enough to have had a very significant impact on the immediate fortunes of our firms. On the other hand, progress had been made with regard to economic liberalization and the advantages were already being appreciated by the firms. Kenyan manufacturing firms show a large variation in technologies and productive efficiency. We cannot discem any general productivity advance between 1992 and 1993. Some firms had a large increase in productivity, but even more showed productivity declines. Since there is an extremely large spread between the best practice technology and that applied by the less efficient firms, there is a vast potential for rationalization measures to move firms closer to best practice. Ahhough some benefits can be brought about by increasing the scale of production, a much greater potential for productivity improvement can be had by bringing firms up to best-practice technology at observed output levels. Increased competitive pressure, from intensified international competition, for example, could help to reduce the spread in the efficiency distn'bution, and could also help to move the frontier out. Improved productivity requires better technologies, among other things. Technology acquisition in Kenya has largely depended on investment in new machinery, while licensing, for example, has been rather uncommon. Technology transfer has been limited, and the general technological level ofKenyaiJ. firms remains modest. There was no change in these respects between the two rounds of interviews. The use of foreigners as managers and technicians also remained rather limited. Another crucial factor in the business environment is financing. We note that different categories offirms have very different access to credit. For example, many large firms can use trade credit from suppliers to finance their activities, but small firms are instead often 179 . forced to provide extensive trade credits to their customers. Small firms essentiaDy lack · access to the formal credit market, largely because these firms are unable to put up assets for collateral Thus, small firms have to rely on informal loans, mainly from friends and relatives. ~ough these are usually given without collateral, and often without any interest · charges, their availability is necessarily quite limited. Other informal credit does not seem to be particularly important in Kenya, so informal loans do not substitute effectively for formal loans. Investments are the basis for further expansion of manufacturing, and they have to be financed. Our probit analyses show that access both to formal loans and to trade credits is strongly related to firms' abilities to put up collateral, so that trade credits also do not act as substitute for formal loans. We also note that the collateral required is usually several times the value of the loan. These high collateral requirements put a severe constraint on bank-financing, and thus severely limit investment activity. The regulatory environment for Kenyan firms has improved. Foreign exchange and price controls have been removed, and tariffs and taxes have been rationalized. Government controls are therefore less of a concern to entrepreneurs than they used to be. However, many still complain about conuption, and there is a universal complaint about the deterioration of infrastructure, especially with regard to ports, roads, electricity, freight transport, and telephones. As a consequence, there has been some increase of self-provision of infrastructure services, mainly of security and transport. Some changes brought about by adjustment policies are definitely appreciated by the firms. However, the effects ofthe contractionarypolicies instituted to stabilize the macro economy were bemg felt by the firms. We noted that the main complaints of the firms concerned lack of credit, low demand and poor and expensive infrastructure. To some extent, these problems are part ofthe adjustment process, considered necessary for the long-term revival of the economy. And the firms are optimistic about their longer-term prospects, which suggests that they have some confidence in the adjustment programme. · 180 The main pwpose ofthis project is to understand manUfacturing firm growth. In this report we have analyzed the impact on firms of a wide range of factors, both intemal and extemal to the firms. We have considered the role of labor, capital and technology, as well as the impact of the economic environment. We have reported on the firms' evaluations of the adjustment measures undertaken by the government and their impact, and note that most firms took a relatively opt:imi&tic vie\1V ofthe future. Still, long-term expansion requires that investments take off: and so far this has not happened. Here we deal with the intricate problems of confidence and the credibility of the reform process. The establishment of such confidence depends on the quality of government, and here much remains to be done. · 181 REFERENCES Alinge, S. (1987), Acquisition of Capabilities Through International Technology Transfer, The Case of Small Scale Industrialization in Tanzania, ChaJmers University of Technology, GOteborg. Amemiya. T. (1981), "Qualitative Response Models", Journal of Economic Literature, VoLl9, December, pp. 1483-1536. Atuahene-Gima, K (1992), "Relative importance offirm and managerial influences on international technological licensing behaviour", International Marketing Review, VoLlO, No.2. Berg, S.A, F.R. Forsund and E.S. Jansen (1991), "MaJmquist Productivity Growth During the Deregulation of Norwegian Banking 1980-1989", Working Paper, Bank of Norway. BhaDa, AS. (ed.) (1975), Technology and Employment in Industry. Intemationl Labour Organization, Geneva. Bigsten A, W. Masai, and H Semboja (1994), ''The Acquisition of Foreign Technology in the Kenyan Manufiu::turing", Paper presented at the International Conference On Afiican Economic Issues, held in Arusha, Tanzania, October, 11-14. Caves, D.W., LR. Christensen and W.E. Diewert (1982), ''The Economic Theory of Index Numbers and the Measurement ofInput, Output, and Productivity", Econometrica 50, pp. 1393-1414. Central Bureau of Statistics, (CBS), (various issues), Economic Survey, Nairobi Chames, A, W.W. Cooper (1985), ''Preface to Topics in Data Envelopment Analysis", Annals of Operations Research 2, pp. 59-94. Chames, A, W.W. Cooper and E. Rhodes (1978), "Measuring the Efficiency of Decision . Making Units", European Journal of Operational Research, Vol2, No.6, pp. 429-444. Coughlin, P. and K Gemshon lkiara, (1991), Kenya's Industrialization Dilemma, Heinemann, Nairobi Dahlman, C. 1. and M.Cortes (1984), "Exports of Technology by Newly-Industrializing Countries: Mexico", World Development, Vo1.l2, No 516. 182 Davidson, Rand IG. Mackinnon (1993), Estimation and Inferences in Econometrics, · Oxford University Press, Oxford. Departments of Economics, Universities of Gothenburg and Nairobi (1993a), Economic , . Development and the Manufacturing Sector in Kenya, RPED, World Bank, Washington D.C. Departments of Economics, Universities of Gothenburg and Nairobi (1993b), Regional Programme on Enterprise Development, First Report on the Kenyan Survey, World Bank, Washington D.C. Departments of Economics, Universities of Gothenburg and Nairobi (1994), Limitations and Rewards in Kenya's Manufacturing Sector: A Study of Enterprise Development, RPED, World Bank, Washington D.C. Departments of Economics, Universities of Gothenburg and Nairobi (1995), Regional Programme on Enterprise Development. First Report on Round 1 of the Kenyan Survey, World Bank, Washington D.C. Economist Intelligence Unit (EIU) (1994), Kenya, London. Forsund, F.R. and L. Hjalmarsson (1974), "On the Measurement of Productive Efficiency", Swedish Journal of Economics 76, pp. 141-154. Forsund, F.R and L. Hjalmarsson (1979), ''Frontier Production Function and Technical Progress: A Study of General Milk Processing in Swedish Daily Plants", Econometrica, Vo147, pp. 883-900. Forsund, F.R and L. Hjalmarsson (l987), Analyses of Industrial Structure: A Putty- Clay Approach, The Industrial Institute for Economic and Social Research, Almqvist & Wiksen International, Stockholm. Forsund, F.R. (1991), '''The Malmquist Productivity Index", Memorandum, Department of Economics, University ofOsIo, No.28. Piire, R', S. Grosskop.t; B. Lindgren and P. Roos (1989), "Productivity Developments in Swedish Hospitals", Mimeo. Fare, R. and S. Grosskopf (1990), "Theory and Calculation of Productivity Indexes: Revisited", Working Paper, Carbondale, USA Greene, W.K (1990), Econometric Analysis, Macmillan Publishing Company, New York. 183 Greene, W. H. (1992), LlMDEP, User's Manual and Reference Guide, Version 6.0 BeDport, NY. Helbert-Copley, B. (1990), 4'Technical Change in Latin America Manufacturing Finns: A Review and Synthesis", World Development, VollS, No.l1. Hohz-Eakin D. 1994. "Public Sector Capital and the Productivity Pu.zzle", Review of Economics and Statistics, VolLXXV1, No.1 pp. 12-21. Hussain, Q. (1995), Capital Structures in Middle-Income Countries: A Study of Indonesia and Malaysia, PhD dissertation, Berkeley. Katz, J., (1983), ''Technological Change in the Latin American Metal Working Industry". CEPAL Review, April Kennedy, P. (1994), A Guide to Econometrics, Blackwen Publishers, Oxford. Kenya, Government of(1994), Economic Survey 1994, Nairobi. Kenya Assoaciation of Manufacturers, KAM, (1988), Price Controls, Experience and Opportunities in Kenya's Manufacturing, Nairobi, Kenya. Lan, S. (1984), "Export of Technology by Newly-industrialized Countries", World Development, Vol 12, No.5/6. LaD. S. (1987), I.arning to Industrialize, The Acquisition ofTecbnological Capability by India, Macmillan, London. Levin, J. and N. Ndungu (1994), Striving for Credibility: Kenya 1993-1994, Macro- economic Studies, SIDA, Stockholm. Lynde, C. and J. Richmond (1992), "The Role of Public Capital on Production", Review of Economics and Statistics, Vol LXXIV, No.1 pp.37-44. Maddala, G.S. (l983), limited-Dependent and Qualitative Variables in Econometrics, Cambridge University Press. Nadiri, M. I. and P.T. Mamuneas (1994), ''The Effects of Public Infrastructure and R&D Capital on the Cost Structure and Performance of US Manufacturing Industries", Review of Eeonomics and Statistics, VolLXXVI, No.1 pp. 22-37. Noll, RG. (1990), "Economic Perspectives on the Politics of Regulation", in Schalensee Rand R Willig (eds.), Handbook of Industrial Organization, North-Holland. Amsterdam. · 184 Romano, AC. (1990), "IdentifYing Factors Which Influence a Product Innovation: A Case Study Approach", Journal of Management Studies, Vol.27, No.1. Seyoum, B. (1990), Technology Licensing in Easten Africa: A Critical Exposition and Analysis, Avebmy, Gower Publishing Company, Aldershot, England. · Shah, A (1992), ''Dynamics of Public Infrastructure, Industrial Productivity and Profitability", Review of Economics and Statistics, Vol. LXXIV, No.1 pp. 28-36. Teitel, S., (1993a), Industrial and Technological Development, Inter-American Development Bank, Washington, D.C. Teitel, S, (1993b), Technology Acquisition, Operation and Development in Selected Kenyan Manufacturing Establishments, RPED, The World Bank Report. Tostenson, A and J.G. Scott (1987), Kenya. Country Study and Norwegian Aid Review, The Christian Michelsen Institute, Bergen. White, H (1980), ''Nonlinear Regression on Cross-Section Data", Econometrica, 48(3). World Bank (1994) Kenya-Public Expenditure Review, Washington D.C. 185 p ·