Adjustment Costs of Trade Liberalization: Estimations for the Russian Labor Market Akhmed Akhmedov, Evgenia Bessonova, Ivan Cherkashin, Irina Denisova, Elena Grishina, Denis Nekipelov May 2004 Abstract The paper investigates adjustment costs of trade liberalization in Russia by estimating various labor market elasticities with respect to indicators of trade liberalization in the 90-ies. In particular, the influence of tariff reduction on demand for labor is estimated, inter-sectoral employment flows in recent years and their determinants are studied, as well as determinants of sectoral wage premiums and of wage differentials between skilled and unskilled labor. The estimated elasticities of labor demand and wages show to be of very moderate size implying a modest adjustment cost in the labor market. JEL Classification: J31, J62, F16 Keywords: Labor Market, Trade Liberalization, Labor Demand, Wage Premiums, Employment Flows Introduction Trade reforms, including the liberalization related to WTO accession, having long-run benefits, have at least short-run costs. In particular, the expected resource reallocation is not costless: some transitional unemployment and loss of output could be experienced when some inefficient enterprises are shut down. Moreover, the costs and benefits are unlikely to be uniformly distributed. Hence, in the short-run, there are going to be winners and losers. In the long-run, however, there is evidence that countries which experienced trade-led growth also experienced income growth of the poor which was in line with the average income growth. One of the questions of interest for policy makers is how large the adjustment costs of trade liberalization are, i.e. how strong are the potentially disadvantageous short-run outcomes that result from trade liberalization. A significant portion of potential costs are related to the influence of trade reforms on the labor market. There are several potential channels of influence of trade shocks on the labor market. Free trade expects to change relative prices, and hence redistribution of resources to more efficient use. That would affect output composition, and in turn, demand for labor. Changes in demand for labor transmitted through labor market would shift employment and income distribution between sectors. In addition to this indirect influence, changes in relative prices could affect employment and incomes directly: changes in relative prices of inputs would affect labor demand, while adjustment of relative prices of consumer goods is expected to affect labor supply. Being transmitted trough the labor market this direct effect will also change sectoral distribution of employment and incomes. The total outcome of the resource reallocation and the magnitude of adjustment costs depend both on the characteristics of external shocks and on degree of rigidity and flexibility of internal markets. The degree of flexibility of labor market reflected, among others, by regional and sectoral mobility determines the speed of transition of workers from unemployment to employment or from old jobs to new jobs, thus shaping the size of adjustment costs. The paper attempts to estimate responsiveness of Russian labor market with respect to international trade parameters using the experienced of trade liberalization during the 90-ies. The 90- ies are characterized by the increased openness of the Russian economy resulting in a significant increase in import competition experienced by Russian producers on the one hand, and on enlarged opportunities for exporting sector. The level of tariffs was changing significantly during the period aiming at domestic industry protection. Tariff levels increased in all industries during the period of 1994-1998, except for those in building materials. Wood processing and light industries were those with persistently high tariffs, while chemical industries and fuel and energy industries had relatively low tariffs throughout the period (Diagram 1.1). 2 Tariff dynamics tariffs, % year Source: CEFIR calculations Diagram 1.1. Tariff dynamics in Russian industry, 1994-98 In what follows we look at several channels of the influence of trade shocks on the labor market by estimating labor market elasticities with respect to trade liberalization indicators. In particular, the influence of tariff reduction on demand for labor is estimated in Section 2, inter- sectoral employment flows in recent years and their determinants are studied in Section 3, as well as determinants of sectoral wage premiums (Section 4) and of wage differentials between skilled and unskilled labor (Section 5). We find low magnitudes of responsiveness of the labor demand to trade liberalization, both through the indirect effect of output changes and directly through the influence of tariffs and import penetration. This suggests that the adjustment costs to trade liberalization in the form of changes in industrial labor demand are not high. Moreover, one should take into account the effects of the shift from industrial employment to employment in services which are to dampen the effect of trade shocks. We also find that trade liberalization does not have a significant effect on wages. It is likely that tariff reduction and trade liberalization would lead to only slight increase in the wage differentials between skilled and unskilled labor. It is obtained that there is no significant effect of tariffs on wages and wage premiums. Therefore, no significant evidence for the claim that “workers in more protected industries earn relatively more” is found. The results of the paper suggest that there is a significant negative effect of import on wage premiums while export orientation positively affects wages. 3 2. Estimation of Labor Demand Elasticities In this section we analyze changes in employment due to trade liberalization. Labor demand, and in particular, the elasticity of labor demand with respect to output is the key determinant of employment on the labor market. Trade reform is expected to result in an increased demand for labor by exporting sectors and in a decrease in labor demand in import-competing sectors. We use balance sheets of Russian large and medium enterprises for 1995-2000 to estimate labor demand equation1 and to calculate possible changes in employment due to various shocks in output and tariffs. Table 2.1 reports the obtained labor demand elasticities with respect both to wage and output for the whole economy sample and for each of the nine 2-digit OKONH industry sub-samples. For the entire sample the wage labor demand elasticity is equal to –0.40 implying that a 10% increase in real wage would diminish labor demand for 4%. The labor demand elasticity with respect to output equals 0.22 which means that a 10% increase or decrease in output would cause a 2.2% increase or decrease in labor demanded. These estimates are higher in the absolute value than those reported by Konings and Lehmann (1999) for the Russian enterprises in 1996-1997, but they are still lower than elasticity in Poland, Hungary and Czech Republic during the transition period and than respective elasticity in developed countries. The estimated coefficient for the lagged employment equals 0.24 which is lower than reported for earlier Russia and other transitional economies. This is a sign of decreased inertia of the Russian labor market in the second half of the 1990-ies. To estimate the sensitivity of demand for labor to trade openness indicators, we included (lagged) tariff and import penetration levels in labor demand equations. It turns out that both indicators are statistically significant, with higher import tariffs being associated with higher (lagged) demand for labor and higher import penetration – with lower demand for labor2. Hence, a positive impact of trade barriers and a negative impact of trade liberalization on the number of workers demanded by the Russian industry as a whole is obtained. The magnitude of the influence is not high, however. In addition to the estimates for industry as a whole, labor demands for nine particular industrial sectors were also estimated (Table 2.1). It is clear from the table that coefficients for lagged employment vary across industries. The short-run wage labor demand elasticities are 1 We estimate the following form of labor demand equation: 2000 ln(Lti )  1 * ln(Lt 1,i )   2 * ln(Qt ,i )   3 * ln(Wt ,i )   * X i ,i   t * dt  i  i,t t 1998 where Li,t – is the number of workers employed at the enterprise at period t, Qi,t – sales of enterprise i during year t, and Wi,t – average wage at enterprise i in year t, X – is a set of other variables, dt – time dummies. In our case X contains such regressors as tariffs, import penetration index, unemployment level, Gross Regional Product over Gross Domestic Product in the Russian Federation, industrial output index, real regional average wage and Herfindal-Hershman concentration Index. 2 Herfindal-Hershman index, average wage in the region, GRP per capita over the all-Russia GDP per capita and time dummies were used to control for regional and time differences and turned to be significant. 4 insignificantly different from zero in power, petrochemical, and machinery construction industries, but are as high as –0.58 in wood, woodworking, pulp and paper industry and –0.61 in light industry. The output labor demand elasticities are significantly different from zero in all industries and vary from 0.12 in power and mining industry up to 0.31 in woodworking, pulp and paper industry3. The variation is in line with the basic intuition and the Hicks and Marshall’s labor demand rules. The products of light, food, and construction materials industries are likely to face more competitive markets, i.e. markets characterized by higher product price elasticities. This in turn results in higher own price demand elasticities of inputs. It is worth noticing that in the same industries where low wage labor demand elasticities are reported, high coefficients for lagged employment are observed (around 0.55) indicating significant labor demand inertia in these industries. It is instructive that only weak support to the hypothesis of positive impact of trade barriers, such as higher tariffs rates, on labor demand is found on industry subsamples. In all industries with exception of metallurgy the corresponding coefficients at tariff variable were found to be insignificantly different from zero. The same results hold when using the ratio of import goods to the domestic output as a measure of trade protection. Only in light industry the higher share of imported goods has statistically significant negative impact on the number of employed. The low correlations between tariff level, import penetration rates and labor demand found in our regressions does not mean, however, that trade liberalization does not have impact on the labor demand since trade liberalization affects industrial structure and output in industries, which in turn affects demand for labor. The estimation of labor demand elasticities show that they very not only across industries, but also across regions. Table 2.2 and Diagrams 2.1, 2.2 show the estimates of labor demand equation for each of the eleven economic regions and separately for Kalingradskaya oblast4. All labor demand elasticities with respect to output are significantly different from zero and vary from 0.15 in Povolzhskiy region up to 0.34 in the Northern economic region. The negative and statistically significant impact of real wage changes on labor demand by firms has been found in all 12 analyzed regions, with the Northern, East-Siberian economic regions and Kaliningradskaya oblast experiencing the highest own wage elasticities (around –0.55), while the North-Western economic region having the lowest elasticity of –0.18. Diagram 2.1 makes clear that wage labor demand elasticities are higher in the northeastern parts of the Russian Federation as compared to the western European part, with exception of Kaliningradskaya oblast. 3 Speaking about differences in coefficients between regions and industries we do not mean that they are not statistically identical. Some kind of poolability test is required here. However, the existing first-order autocorrelation of the residuals, which does not cause inconsistency of Arellano-Bond GMM estimator, makes it quite difficult to construct the formal test. This could be the sphere for further research. 4 The estimates of the model also show that enterprises, located in the regions with higher unemployment or smaller economy size, are likely to have lower number of employed. The mixed results were obtained for industry growth. The higher rates of industry growth rates correspond to higher employment in power, metallurgy, machinery construction and food industries, but to lower employment in construction materials and light industry. 5 Brown and Earle (2001) explain interregional differences in gross job flows by the differences in concentration of employers. At the least concentrated markets, i.e. at the markets with higher number of potential employers, the employees have more outside opportunities. This restricts firms to destroy job places. However, in our case we obtain the reverse result, i.e. we find higher wage labor demand elasticities in northeastern parts of Russia. We also conclude that the measure of concentration of employers used – Herfindal-Hershman index - is not significant in the most of the regressions. This implies that other sources of interregional differences could come into play, with distinctions in the industrial structure and variations in the degree of paternalism of regional authorities across regions5 being the candidates. The degree of paternalism in turn could depend on the political orientation of the political leader of the region and on the level of political system development. Turning to the tariff and import penetration variables included in the model to measure effect of trade openness on labor demand, we found positive impact of higher trade barriers on the number of employed in several regions. In all cases, except one, when these variables are significant, the tariff level coefficient is positive and the import penetration coefficient is negative. Summing up, it could be concluded that the Russian labor market is characterized by rather low labor demand elasticities with respect to output and wages. Those are higher though than at the beginning of the transition period implying that on the whole Russian enterprises became more sensitive to the changes in output than they were in 1996-1997. The latter is supported by higher labor demand elasticity with respect to output and wages and lower inertia. The estimated labor demand elasticites vary not only across industries, but also across regions. The direct influence of trade openness on employment could be outlined only in some industries and regions, however. In most of the cases higher protection corresponds to higher number of workers demanded by firms, holding other things constant. With exception of some cases higher industry growth rates and bigger size of the regional economy also lead to higher employment. The found low magnitudes of responsiveness of the labor demand to trade liberalization, both through the indirect effect of output changes and directly through the influence of tariffs and import penetration, suggests that the adjustment costs to trade liberalization in the form of changes in industrial labor demand are not high. Moreover, one should take into account the effects of the shift from industrial employment to employment in services which are to dampen the effect of trade shocks. The next section is to shed some light on the issue of inter-sectoral labor mobility and its determining factors. 5 In this sense our result could be driven by less paternalism of regional authorities in the North-East of Russia as compared with the European part. 6 3. Analysis of Inter-Sectoral Labor Flows Trade liberalization, by changing production structure, induces changes in structure of labor demand, and via interaction with labor supply affects sectoral structure of employment. The latter induces inter-sectoral labor flows, with pace of adjustment to shocks of trade liberalization depending on degree of flexibility of economic system. This section studies inter-sectoral labor flows. A stylized search model of inter-sectoral mobility of labor with a worker choosing one vacancy from several sectors would name the relative wages in different sectors adjusted for costs of retraining and vacancy arrival rates in sectors as the key factors behind the worker’s choice. The model predicts that the probability of movement from one sector into another depends on the magnitude of retraining costs, alternative (non-labor) income, costs of search, job availability (offer arrival rates) and employment opportunities in sectors – mean and variance of wages in comparing sectors. Trade liberalization influences probability to change sector of employment by affecting job arrival rate in various sectors and characteristics of wage distribution across sectors. The study of direction and intensity of labor flow in Russia using vector auto regression model6 and tracing the impact of exogenous variables on probabilities to move across sectors is based on Goskomstat publications7. Based on the time series data on labor flows between 17 and 35 sectors provided by Goskomstat, we integrate the sectors into three and five sectors: manufacturing sector, services sector and unemployment as the three-sector grouping and raw materials sector, processing sector, financial services and management sector, other services sector and unemployment as the five-sector grouping. As the first step we estimated a model describing flows of labor between three basic groups of sectors: manufacturing, service and unemployment. These sectors were consolidated in the model of vector autoregression in the form:  Manufactur ing   Manufactur ing    = Р    Services   Services  .  Unemployment   Unemployment   t   t 1 Estimation results for components of the matrix of probabilities8 are given in Tables 3.1 and 3.2. One can notice that obtained matrices of probabilities of transitions of labor between sectors 6 It is possible to show that probabilities to move from one sector to another comprise a stochastic matrix NxN with elements P = (pij) . With probabilities being stable over time, the dynamics of distribution of employment over sectors is determined by Markov process: xt+1 = P xt , where x – vector of employment distribution over sectors. As a result the dynamics of employment distribution across sectors is described by vector auto regression of order 1. 7 “Social and economic position of Russia, 1998 – 2003” and “Short-term economic indicators of the Russian Federation, 2003”. 8 To check whether the dynamics of inter - sectoral distribution of labor under steady matrix of probabilities of transitions of workers can be described by Markov equation we estimated vector auto regression of the second order. Coefficients for lagged terms of the second order appeared insignificant thus supporting our hypothesis. 7 appear diagonal - diagonal elements of the matrices are the only significant coefficients. The result implies that the main role is given to labor flows within the sectors. To investigate the role of external macroeconomic factors on inter-sectoral flows of labor in Russia vector autoregression with exogenous variables was run9. Three parameters were used: total labor demand as measured by the number of vacancies registered with employment service, industrial sector labor demand measured as a volume of industrial output, and the ratio of average incomes to average wages as the indicator for alternative earnings. The results are presented in Table 3.3. It appears that the probability to stay in the manufacturing sector is higher the higher is the overall demand for labor. At the same time, the higher is the output (and hence labor demand) in manufacturing sector the lower is the probability to remain in manufacturing sector and the higher is the probability of unemployment. Similar analysis could be carried out for five-sectoral division. The appropriate reduced form of the equation of vector auto regression looks like:  Raw material sector   Raw material sector       Processing sector   Processing sector    = Р   .  Finance credit and management   Finance credit and management   Other services   Other services           Unemployment t  Unemployment  t 1 Results of an estimated of model of vector autoregression are given in Table 3.4. One may notice that the net flows of labor from processing sectors to raw material sector, and also to sector offering services, are small (the appropriate coefficients are insignificant). However the interesting fact is that the probability at which labor moves from sectors, offering other services (including transport, communications, public health services and education) to the sectors offering services of financial intermediatin is high enough (17.6%). Adding exogenous parameters shows that the probability to keep work in raw sectors is higher the higher is the number of vacancies registered with employment service. Moreover the probability of transition from processing sectors to raw material sector is lower the higher is the volume of industrial output. The estimates allow generating the so called impulse response functions showing the dynamics of the response of employment in the studied sectors to exogenous shocks. As Diagram 3.1 shows, a positive employment shock in raw materials sector results in a new equilibrium with modest growth of employment in all other sectors except for the processing sector. This implies that an expansion of export creating sector due to trade liberalization generates redistribution of labor from processing industries to services. 9 Nonlinearity was taken into account by including the products of shares of employment in the appropriate sector and values of macroeconomic factors. It provided linearity of dependence of probability of transition of worker from one sector to another from investigated exogenous parameters. Indeed, since the estimated equation is of the form xt=Pxt-1+Zt xt-1, it can be transformed to expression xt =(P + Zt )xt-1, and the matrix of probabilities of transition in such formulation depends on the matrix of exogenous parameters Zt. 8 Diagram 3.1 When there is a positive employment shock in services, however, employment in processing sector declines sharply with employment in raw materials industries slightly increasing (Diagram 3.2). Hence, provided trade liberalization boosts financial services, employment in all the rest of the sectors except for processing industry is likely to increase. Diagram 3.2 Data for regions of the Russian Federation allow further studying changes in matrices of probabilities of inter-sectoral mobility10. To estimate responsiveness of the estimated probabilities of movement across sectors to exogenous variables reflecting retraining costs, alternative income, costs of job search, vacancy arrival rate, and characteristics of wage offer distribution (mean and variance), fixed effects model was estimated. The following variables were used as proxies for the aforementioned parameters: the share of employment in small business to reflect alternative income, 10 The procedure to estimate components of transition matrix is to search a matrix Р on a basis of k observations of realization of distribution of workers in sectors x1, …, xk. Data allow to build k-1 equations to calculate components of matrix Р which has the property that xN=PxN-1=P2xN-2=… PN-1x1, and Р∙1=1. Direct calculations of the matrix of transition probabilities between sector result in probabilities greater than one or smaller zero. This is due to a significant share of errors in initial data and because the process of displacement of workers between sectors can deviate from Markov’s. To estimate the components of the probability matrix the following problem was solved in GAUSS: p i ij  max for all j under condition Р(x1 x2…xk-1)= (x2 x3…xk ) и 0