In this section, I provide three robustness tests for the regression results presented in the previous section. First, I examine the robustness of the results for an alternative way of
measuring misallocation. Secondly, I modify the regression model estimated earlier so that the effect of FDI on misallocation is allowed to depend on other factors, namely R&D and
educational attainment. Finally, I check the robustness of the results to different ways of measuring policy constraints.
Dispersion-based measures of misallocation
The TFP gap which is used as an indicator of misallocation in the previous regressions is derived from a structured monopolistic competition model. Therefore, the measure could be sensitive to the assumptions that underlie the model such as elasticity parameters and forms of aggregation. In this sub-section, I test the robustness of the regression results reported previously to alternative ways of measuring misallocation. As indicated earlier, I measure misallocation using standard deviation and the interquartile range of total factor productivity (TFPR), the marginal product of labor (MRPL) and the marginal product of capital (MRPK). Dispersion-based measures of misallocation have been used in previous studies because they need minimal assumptions and are easy to calculate (Ito and Lechevalier, 2009; Arnold et al., 2008). Higher dispersion of productivity or factor returns indicates the presence of large unrealized efficiency gains from reallocating inputs.
Table 4 reports regression results for the same model estimated earlier, but this time using the dispersion based measures of misallocation as dependent variables. The results reported here are all based on country-level data since in general industry-level data gives identical results. The standard deviations of TFPR, MRPL and MRPK are used in the first three regressions, and the interquartile ranges of the same variables are used in the last three regressions.
--- [Table 4 about here] ---
From Table 4, trade openness enters with negative coefficients, but the results are almost always insignificant. The significance of trade openness thus seems to be sensitive to the way
misallocation is measured. FDI inflow also has negative coefficients, and it appears significant in regression 4. This negative effect of FDI on misallocation is contradictory to the result found in the baseline regressions. The variables measuring entry and exit barriers, financial frictions and the profit rate are insignificant, which is consistent to the finding in the baseline regression.
The cost of firing employees has a positive effect on the standard deviation of TFPR, and both on the standard deviation and interquartile range of MRPL. This result confirms the previous finding that high firing costs induce misallocation by hindering the efficient allocation of labor. However, the cost of firing is insignificant in regression 4, perhaps because of the inability of the interquartile range to capture misallocation at the tails of the distribution. Overall, the results for firing costs are still robust for an alternative way of measuring misallocation, whereas the results for trade openness appear weaker and those of FDI seem contradictory to the previous finding.
FDI and misallocation
FDI has appeared with a positive and significant coefficient in regression 1 of Table 3, but it has entered with a negative and significant coefficient in regression 4 of Table 4. The aim of this robustness test is to provide an explanation that can reconcile these contradictory results.
As indicated earlier, although FDI is expected to improve allocative efficiency by increasing competitive pressure, it could in fact increase productivity heterogeneity since foreign firms are in general more productive than their domestic competitors. Therefore, the effect of increased FDI inflows is not necessarily equalization of marginal products and lower dispersion of productivity. Several studies also reveal that the effect of FDI on the domestic market is not uniform across countries. Among other things, the efficiency of factor allocation between local
and foreign firms depends on the extent to which domestic firms can narrow down their productivity differences with the foreign firms. Studies show that domestic firms can
technologically catch up with more productive multinationals only in countries with a minimum level of technological ‘absorptive capacity’ (Wang and Wong, 2009; Crespo and Fontoura, 2007).
Therefore, local firms in countries with higher stock of knowledge, in the form of R&D expenditure or educational attainment, are more likely to narrow their technological gap with multinationals.
A possible reason for the mixed sign of the coefficients of FDI in the previous regressions is that the regression specifications do not account for the possible interaction between FDI and local absorptive capacity. Therefore, I modify the regression model by including an interaction of FDI inflow with R&D expenditure and with years of schooling. If it is correct that FDI improves allocative efficiency only in countries where the absorptive capacity is higher, the interaction term between FDI inflows and R&D expenditure, or educational attainment, should appear negative and significant. R&D expenditure and years of schooling are chosen because they are frequently used measures of absorptive capacity in the FDI literature (Wang and Wong, 2009).6
Six regression results are reported in Table 5, all of which based on the three measures of TFP gap that are measured at country-level. R&D expenditure and its interaction term with FDI are included in the first three regressions, whereas years of schooling and its interaction term with FDI are included in the last three regressions. There is a substantial fall in the number of observations in the first three regressions because of lack of R&D data, which makes comparison with earlier results difficult. However, the fall in the number of observations is not significant in the regressions results where years of schooling are used. Another notable difference between
these results and those in Table 3 is the considerable improvement in the goodness of fit of the models, signifying the importance the new variables that measure absorptive capacity.
--- [Table 5 about here] ---
The interaction term between FDI and R&D, given in one of the last rows, indeed appears negative and strongly significant, indicating that FDI inflow is more likely to improve allocative efficiency in countries with higher R&D expenditure. It is also noteworthy that the main effect of FDI becomes positive and significant when the interaction term is included. In regression 1, the marginal effect of FDI, evaluated at the mean value of R&D, is not significantly different from zero. Thus the positive effect of FDI on allocative efficiency appears to be realized only when R&D expenditure is reaches a certain minimum threshold. The coefficient of the interaction term is the largest in regression 2 where the dependent variable measures misallocation caused by size-related distortions. This suggests that absorptive capacity is particularly relevant to moderating the size gap between local and multinational firms.
The interaction term of FDI with years of schooling is also negative, as can be seen from the last row of regressions 4-6. The interaction term, however, has low level of significance, possibly because years of schooling do not sufficiently capture domestic absorptive capacity. In general, the results presented in Table 4 provide supportive evidence to the hypothesis that FDI improves allocative efficiency only when absorptive capacity is sufficiently large. Moreover, the last 3 regressions in which the number of observations is relatively high give similar results for the other policy variables as those reported in Table 3. Trade openness and firing costs appear to have strong effect on misallocation whereas entry costs have a weak effect.
Alternative measures of policy constraints
In the previous regressions, policy constraints are measured using indicators which are frequently used in previous studies, and for which data is most widely available. In this sub-section, I check the robustness of the results for alternative ways of measuring policy constraints.
For each policy constraint, I select an alternative indicator than the one used earlier, and re-estimate the previous regression models. Descriptive statistics for all alternative measures of policy constraints is provided in Table A3 in the appendix.
In the baseline regression, trade intensity and net FDI inflows are used to measure openness to trade and foreign investment respectively. While these measures are reliable and widely available, both are outcome variables and do not directly measure the openness of the policy regime. To the extent that we are interested in the policy that drives openness to external competition, and not on the level of external competition as such, a direct indicator of policy with regard to external openness would be preferable. For measuring the openness of country’s trading regime, the tariff rate is thus a more relevant policy variable. Our new measure of trade openness will be the weighted average tariff rate in manufactured products. Although there is no directly available policy variable that measures openness towards foreign investment, previous studies have tried to compile such indicators. I use one frequently used and updated indicator constructed by Chinn and Ito (2002) that measures the extent to which a country’s capital account is open for foreign capital inflows. This indicator is generated by aggregating other sub-indicators of the ease of cross-border financial transactions. Although the indicator is rather broad since it also considers openness to short term capital inflows that do not necessarily affect allocative efficiency, it is still more closely related to the openness of the country’s policy for foreign capital inflows.
Coming to indicators of entry and exit barriers, two alternative indicators are used. Instead of using the cost of entry to measure entry barriers, I use the number of procedures newly entering firms have to complete in order to register officially. And instead of the recovery rate, I measure exit barriers using the cost of insolvency proceedings as a percentage of total estate value. The difference between the two is that the recovery rate is more inclusive since it is calculated using the probability of foreclosing, restructuring or liquidating the insolvent firm, whereas the cost of insolvency proceedings only measures the costs associated with the insolvency process. Although, according to Djankov et al. (2008), the recovery rate is the preferred indicator of the efficiency of bankruptcy proceedings, the measurement of the cost of insolvency is more straightforward and involves fewer assumptions.
With respect to financial frictions, I use the interest rate spread as an alternative measure of financial development. The interest rate spread is calculated as the difference between the lending and borrowing interest rates, and indicates the level of efficiency financial
intermediation. Higher values of interest rate spread indicate low level of efficiency. The advantage of the interest rate spread compared to size-related indicators such as private sector credit is that it is more relevant since it measures the efficiency of financial intermediation.
Finally, the total tax rate is used as a measure of capital market distortion rather than the profit tax rate. Although the profit tax rate is a straightforward indicator of distortions in the cost of capital, small profit tax might also be accompanied by relatively high taxes in other areas.
Therefore, it is useful to check the robustness of the results using the total tax rate which includes profit taxes as well as other mandatory contributions such as social contributions, labor taxes paid by the employer, property taxes etc. Finally, I use a more specific indicator of the cost of firing to measure labor market distortions. Unlike the previous indicator of cost of firing which
includes indirect cost of firing redundant workers, the new measure includes only the direct costs of advance notice requirements and severance payments expressed in salary weeks. The new measure also takes in to consideration differences in the wage of the employees who are to be redundant.
Three regression results based on the new indicators of policy constraints are reported in Table 6. In all of the regressions, the dependent variables are the indicators of the TFP gap in which both types of distortion are removed. The first regression is similar to the baseline
regression 1 in Table 3 except that new explanatory variables are used. The second regression is similar with the extended model reported in Table 5 in which an interaction term between openness to foreign investment and R&D is included. The third regression is similar to the second except that educational attainment is now interacted with foreign openness.
Regression 1 in Table 6 is remarkably similar with regression 1 in Table 3 even though all of the explanatory variables are measured differently in the two regressions. In both regressions, only the indicators of openness for trade and foreign investment, and firing costs are significant.
The coefficient for the tariff rate is particularly high in the first regression of Table 6, indicating that lowering the tariff rate by 1 percentage point will reduce the TFP gap by almost 5%.
According to this result, Botswana, with a tariff rate close to the 75th percentile (10%), could increase its TFP by almost 25% by lowering its tariff rate to the level of Chile’s, which has a tariff rate close to the 25th percentile (5%). The coefficient of firing costs in regression 1 of Table 6 rises almost three-fold compared to the baseline regression. This could reflect an improvement in our estimation because of a more precise measure of the cost of firing. However, this could also be because of the large fall of the number of observations in the present regression. The new results indicate that lowering the cost of firing from the 75th percentile (25 weeks, in
Madagascar) to the 25th percentile (12 weeks, in Lithuania) can increase TFP by 21%, which is very similar to the result from the baseline regression.
--- [Table 6 about here] ---
The interaction terms in regressions 2&3 of Table 6 also confirm the same results reported earlier. The negative coefficients of the interaction terms between financial account openness and the two indicators of absorptive capacity (R&D and educational attainment) reveal that openness to foreign competition increases allocative efficiency in countries with higher absorptive
capacity. In both regressions 2 & 3, the marginal effect of openness to foreign competition evaluated at mean values of R&D and educational attainment is positive and significant, indicating that openness to foreign investment improves allocative efficiency only when absorptive capacity is higher than its level for the average country in our sample.
V. CONCLUSION
The development accounting literature contends that a large part of the cross-country variation in per capita income can be explained by TFP differences (Hsieh and Klenow, 2010;
Caselli, 2005). Whereas the traditional explanation for TFP differences is the presence of barriers to technology adoption, a recent literature emphasizes the role of allocative efficiency. Since firms in the same industry often exhibit substantial productivity heterogeneity, the way resources are allocated among them could have a large impact on aggregate productivity (Restuccia and Rogerson, 2008; Alfaro et al., 2008; Hsieh and Klenow, 2009). A number of recent studies document large misallocation across firms; however, there has been little effort in comparing the the effect of misallocation on aggregate productivity in a cross-country setting. Furthermore, there is little emphasis in empirically identifying the policy constraints that drive misallocation.
This paper tries to contribute to the literature by measuring the magnitude of misallocation for a large number of countries and examining the role of policy differences therein. For this purpose, I use the WBES dataset, which is a unique international dataset with more than 20,000 manufacturing firms across 77 countries. A monopolistic model with heterogeneous producers by Hsieh and Klenow (2009) is used to measure misallocation. The model allows us to measure misallocation caused by two types of distortions: output distortions that affect the size of the firm, and capital distortions that affect the use of capital relative to labor. To calculate the misallocation caused by these distortions, a hypothetical reallocation is conducted by removing the two distortions and calculating the resulting efficient TFP. The results show that for the average country in our dataset the efficient TFP is 120% higher than the actual TFP, indicating that misallocation substantially lowers aggregate TFP. Among the two types of distortions, size-related distortions that inhibit small firms from growing constitute the largest part of total misallocation.
The results also indicate that the magnitude of misallocation varies substantially across countries, suggesting a role for policy differences. Taking advantage of the large country
coverage of the dataset, I conduct cross-country regression analysis to investigate the importance of four types of policy constraints on misallocation, namely barriers to external competition, entry and exit barriers, financial frictions, and labor and capital market distortions. The results show that firing costs, barriers to trade and foreign investment, and high entry costs all contribute positively to misallocation. High cost of firing has a particularly adverse effect on misallocation:
lowering the cost of firing from the 75th to the 25th percentile can raise TFP by 20%. The results also indicate that openness for foreign investment is likely to improve allocative efficiency only when absorptive capacity sufficiently large. This is in line with previous findings that domestic
firms in countries with relatively high absorptive capacity are more likely to narrow their technological gap with foreign competitors.
In general, the findings of this paper highlight the importance of factor allocation across heterogeneous producers. Since large empirical evidence confirms the presence of substantial productivity heterogeneity among firms in the same industry, the scope for reallocation remains great. The presence of favorable policy environment for reallocation is thus essential to capture the potential productivity gains from reallocation in manufacturing. Furthermore, the resulting rise of efficiency and lowering of prices in the manufacturing sector can be beneficial for the performance of related non-manufacturing activities.
REFERENCES
Aghion, P., T. Fally, and S. Scarpetta. "Credit Constraints as a Barrier to the Entry and Post-Entry Growth of Firms." Economic Policy, 22, 2007, 731-779.
Alfaro, L., A. Charlton, and F. Kanczuk. "Plant-Size Distribution and Cross-Country Income Differences," In NBER International Seminar on Macroeconomics 2008. Cambridge, Mass.: NBER, 2009.
Arnold, J. M., G. Nicoletti, and S. Scarpetta. "Regulation, Allocative Efficiency and Productivity In OECD Countries: Industry And Firm-Level Evidence." OECD Economics Department Working Papers with number 616, 2008.
———. "Does Anti-Competitive Regulation Matter for Productivity? Evidence from European Firms." IZA Discussion Papers with number 5511, 2011.
Barseghyan, L. "Entry Costs and Cross-Country Differences in Productivity and Output."
Journal of Economic Growth, 13(2), 2008, 145-67.
Barseghyan, L., and R. DiCecio. "Entry Costs, Misallocation, and Cross-Country Income and TFP Differences." Federal Reserve Bank of St. Louis Working Papers with number 2009-005, 2009.
Bartelsman, E. J., J. C. Haltiwanger, and S. Scarpetta. "Cross-Country Differences in
Productivity: The Role of Allocation and Selection." NBER Working Paper Series with number 15490, 2009.
Bassanini, A., L. Nunziata, and D. Venn. "Job Protection Legislation and Productivity Growth in OECD Countries." Economic Policy, 24, 2009, 349-402.
Bernard, A. B., J. B. Jensen, and P. K. Schott. "Trade Costs, Firms and Productivity." Journal of Monetary Economics, 53(5), 2006, 917-37.
Botero, J., S. Djankov, R. Porta, and F. C. Lopez-De-Silanes. "The Regulation of Labor." The Quarterly Journal of Economics, 119(4), 2004, 1339-82.
Buera, F. J., J. P. Kaboski, and Y. Shin. "Finance and Development: A Tale of Two Sectors."
American Economic Review, 101(5), 2011, 1964-2002.
Buera, F. J., and Y. Shin. "Financial Frictions and the Persistence of History: A Quantitative Exploration." NBER Working Papers with number 16400, 2010.
Caselli, F. “Accounting for Cross-Country Income Differences.” In Handbook of Economic
Caselli, F. “Accounting for Cross-Country Income Differences.” In Handbook of Economic