Jan 22, 2025 Skills Enterprise Note No. 43 Educated Workers and Managers in the EU-27* Mohammad Amin T is Brief highlights issues related to the education and skill level of workers and top managers in rms in 27 European Union countries (the EU-27), using the World Bank Enterprise Surveys (WBES). e exercise is an important step toward understanding the use of skilled and adequately educated workers and top managers by a rm and its likely e ects. e Brief identi es several factors at the NUTS2 region level and rm level that are correlated with the di culty rms face in obtaining adequately educated workers as well as the skill level and education level of the workers and top managers. Somewhat surprisingly, income per inhabitant in the NUTS2 regions is not a strong predictor of the use of skilled and educated workers and top managers or rms’ reported di culty in nding adequately educated workers. Several rm performance measures—such as labor productivity, employment growth, exporting, research and development (R&D), and management quality—are found to be correlated with the use of skilled and educated workers and top managers. Some of these correlations di er sharply between low and high levels of the outcome variables. ere is evidence that training provided to workers by the rms is associated with less dispersion of labor productivity between rms, and greater use of skilled workers is associated with less dispersion of wage rates across rms. Overall, the Brief nds that starting at low-income levels in EU regions, policy focus needs to shift more toward ensuring the availability of adequately educated workers than on reducing other obstacles as the economy develops. is shifting of policy focus can stabilize after the economy is su ciently developed. Possible causes of an inadequately educated ere are sharp di erences between NUTS2 regions workforce at the region level and firm level within a country in the incidence of rms citing inadequately educated workforce as the top obstacle ( gure 2). us, it is More than any other constraint, rms in the EU-27 important to also consider regional or NUTS2-level factors to countries1 rank inadequately educated workers as their top understand the problem of inadequately educated workers obstacle (see the rst Brief in this series). For a typical area in faced by rms. the European Union with between about 800,000 and 3 Figure 2. e share of rms that report inadequately million inhabitants (NUTS2 regions),2 27 percent of rms educated workforce as the top obstacle varies substantially report “an inadequately educated workforce” as the top obstacle between NUTS2 regions to their operations. In more than half of the 186 NUTS2 Understanding regional characteristics that are correlated groupings analyzed in the series, this obstacle is the most with the likelihood of rms reporting inadequately educated frequently cited. Such rms abound among rms of di erent workforce as the top obstacle is a good starting point for sizes, sectors, income groups, and ages ( gure 1). However, identifying possible causes of an inadequately educated their proportion is signi cantly higher among medium and workforce and its likely e ects, the type of policies required to large rms compared to small rms; manufacturing rms address the problem, and which types of policies should be compared to services sector rms; most developed NUTS2 targeted. regions followed by transition regions and then the least Economic development. e most natural determinant of developed regions; and older rms (more than 10 years) the availability of adequately educated workers is the level of compared to younger rms. us, these groups of rms may be economic development (see Lange et al. 2018). Macro-level targeted by policy makers on a priority basis. studies have shown that richer countries have a much higher *A liations: World Bank, Development Economics, Enterprise Analysis. For correspondence: mamin@worldbank.org. Acknowledgments: is Brief is a part of a series focusing on issues of regional disparities and growth opportunities in the EU-27 area. e series is a product of the World Bank’s Enterprise Analysis team (DECEA) and has bene tted from generous support from the EU DG REGIO directorate. e team would also like to thank Norman V. Loayza and Jorge Rodriguez Meza for comments and guiding the publication process. Nancy Morrison provided excellent editorial assistance. Objective and disclaimer: e ndings in this series of Briefs do not necessarily represent the views of the World Bank Group, its Executive Directors, or the governments they represent. All Briefs in the series can be accessed via: https://www.worldbank.org/en/research/brief/global-indicators-briefs-series. Enterprise Note No. 43 Figure 1 Inadequately educated workers is the biggest obstacle for several types of firms in EU-27 countries Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: EU-27 = the 27 member countries of the European Union (EU) in the euro area. level of human capital than the poorer countries (see Lange, e data reveal that more rms rank inadequately educated Wodon, and Carey 2018). Is the problem of inadequately workers as the top obstacle as income per inhabitant increases educated workers less severe compared to other obstacles in the ( gure 3). However, this increase tapers o and becomes more developed NUTS2 regions? Do policy makers need to insigni cant above a certain threshold level of income. ere focus less on education of workers and more on other obstacles are two implications for policy. First, compared to poorer as income level of the NUTS2 regions rises? NUTS2 regions, richer regions need to focus more on ensuring The share of firms that report inadequately educated workforce as the top obstacle varies Figure 2 substantially between NUTS2 regions Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. 2 Enterprise Note No. 43 Inadequately educated workers is cited more often as the top obstacle as the income level of Figure 3 NUTS2 regions increases, but then tapers off Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. the availability, relative to demand, of adequately educated mathematics, and engineering skills is signi cantly higher in workers than on other obstacles. Second, starting at the richer NUTS2 regions. low-income levels, policy focus needs to shift more toward Fourth, the remaining eight indicators show no signi cant ensuring the availability, relative to demand, of adequately correlation with the income level. ese indicators are the educated workers than on reducing other obstacles as the proportion of rms that face di culty in nding workers with economy develops. is shifting of policy focus can stabilize appropriate interpersonal and communication skills, problem after the economy is su ciently developed. solving or critical thinking skills, managerial and leadership Other measures. Several other measures are available to skills, computer or general IT skills; the proportion of rms capture the use and availability of skilled and adequately that report inadequately educated workers as the top obstacle; educated workers and top managers. Some are based on rms’ the percentage of workers with a secondary education in a rm perceptions and others are objective measures. e analysis that on average; the percentage of workers with a university degree follows focuses on 15 indicators of the use of skilled and in a rm on average; and the percentage of rms with the top educated workers and top managers and rms’ reported manager having a bachelor’s or higher degree. di culty in nding them when averaged at the NUTS2 level. For a couple of variables, income matters at su ciently ese indicators potentially capture both the demand and high levels but not otherwise. at is, above a certain level of supply of skilled and educated workers and managers. e income, but not below, higher income is associated with a analysis examines their relationship with income per inhabitant signi cantly higher proportion of workers with a university as of 2019. e results are mixed, education ( gure 5) and a signi cantly higher proportion of First, as expected, higher income is associated with a rms with the top manager having a bachelor’s or higher signi cantly higher share of skilled workers among production degree. Summing up, while economic development may workers in the manufacturing sectors, and a lower share of somewhat improve the availability of skilled and educated semi-skilled and low-skilled workers ( gure 4). workers and top managers relative to demand, it is unlikely to Second, three other indicators show signi cantly better solve the problem of inadequately skilled and educated workers skills availability in the richer NUTS2 regions. ese indicators and top managers in a major way or completely. are the proportion of rms that provide training to their workers; the proportion of rms that face di culty in nding Differentials in training, education, and skills workers with foreign language skills; and the proportion of and their effects rms that face di culty in nding workers with technical skills (other than in information technology, IT), vocational skills, or Training job-speci c skills. Firms in EU-27 countries often provide training to their ird, three more indicators show that the proportion of workers. In a typical NUTS2 region, 43 percent of the rms rms that face di culty nding workers with natural sciences, provide such training. As mentioned, the proportion of rms 3 Enterprise Note No. 43 Figure 4 Firms in richer NUTS2 regions employ more skilled workers and are more likely to provide training 60 49 49 50 40 35 Percent of firms 29 30 20 10 0 Skilled production workers % of firms that provide (%) training Least developed LTransition and Most developedeast developed Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. that provide training increases with the income level of the Gittlemann, and Joyce 2000). A few studies have also shown NUTS2 regions. However, this relationship is largely driven by that younger rms are more likely to train workers. In the NUTS2 regions at the low end of the income distribution. EU-27 countries, there is no signi cant relationship between Above a critical level of income, increases in income show no the likelihood that a rm provides training and the age of the further increase in the proportion of rms that provide rm. However, training is signi cantly more common among training. e provision of training may be especially attractive large rms than small and medium enterprises (SMEs). About for large rms due to the xed costs involved (see Frazis, 41 percent of SMEs compared to 70 percent of large rms The share of workers with a university degree decreases with higher income level at initial levels of Figure 5 income but increases at higher levels of income Source: Original calculations for this Brief based on World Bank Enterprise Surveys. 4 Enterprise Note No. 43 provide training. At the NUTS2 and rm-level, training is Education and skill levels more likely among rms that report that inadequately educated On average across the ten EU-27 countries for which data workers is a more severe obstacle (on 0–4 scale) for their are available, about one in ten workers in the EU-27 has a operations. us, it seems that training is in part aimed at university degree. e distribution of university-educated resolving the shortage of skilled workers. workers across rms is skewed. Only 30 percent of rms One concern with training is that it may be a mere employ university-educated workers at all. Large rms, substitute for education acquired outside the rm. is is “train exporters, and foreign-owned rms are more likely to employ drain.” If true, it implies that a greater availability of higher university-educated workers and, in turn, also employ higher educated workers outside the rm may lead to less training proportions of university-educated workers (table 1, columns 1 provided by rms. As a result, training by rms may not and 2). Firms that spend on R&D also have proportionately increase the total stock of human capital in the country. By more university-educated workers, but this relationship is contrast, if higher education and training are complements—as mainly because rms that spend on R&D happen to be large would be the case if newly hired graduates also received rms, which tend to employ more university-educated workers. additional, on-the-job training—the overall stock of human A key concern for policy makers is why only one-quarter of capital will increase due to training. In the case of the EU-27 SMEs employ university-educated workers and their share countries, at the NUTS2 level, there is no evidence of “train averages less than 9 percent of all workers. Do SMEs nd these drain” for either university-educated or secondary-educated workers too costly or are they lacking in the kinds of skills workers. In fact, there is a signi cant positive relationship useful to SMEs? between training and the share of university-educated workers Most rms in the EU-27 countries use skilled production in a rm, suggesting that training and university education are workers. Nearly 80 percent of rms employ skilled production complements ( gure 6). workers, and the average share of skilled workers among all Training seems to alter the distribution of labor production workers in a rm is 44 percent. In contrast to productivity across rms. Average labor productivity is higher university-educated workers, the share of skilled workers by about 48 percent for rms that provide training compared to among production workers declines signi cantly with rm size rms that do not. e di erence is highly signi cant. What is (table 1, column 3), while the share of the semi-skilled and more, there is evidence that training leads to a much a larger low-skilled workers increases. Skilled workers and improvement in labor productivity of the less productive rms secondary-educated workers seem to complement each other. than the more productive rms.3 us, there is the possibility at is, the share of secondary-educated workers is signi cantly that training may allow the less productive rms to catch up higher for rms that have a higher proportion of skilled workers with the more productive rms (box 1). (table 1, column 4) and a lower proportion of semi- and Figure 6 Firms across NUTS2 regions seem to provide training to university-educated workers Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. 5 Enterprise Note No. 43 Box 1: Can less productive firms catch up with the more productive firms through training? e possibility of “catch up” can be tested using the methodology of Combes et al. (2012). is methodology tests for di erences in the distribution of a variable between two groups. e comparison is summarized in three key parameters—shift, dilation, and truncation. ese parameters refer to how much the rst distribution needs to be altered to best approximate the second distribution. e parameters are (1) a rightward shift of the rst distribution (Shift); (2) what constant factor each of the observations in the rst distribution need to be divided by to match the second distribution (Dilation); and (3) what share of the observations in the rst distribution need to be excluded from its left tail (Truncation). Intuitively, the Shift parameter captures the di erence in the mean value of labor productivity, Dilation captures if one distribution is more homogenous than the other. Truncation re ects possible selection e ects whereby rms with very low values of the variable under consideration are more likely to survive in one group than the other. Table B1.1 provides the estimates of the three parameters for the distributions of labor productivity of rms that provide training versus those that do not (column 1) and for the bottom half versus the top half of the NUTS2 regions in terms of the percentage of rms that provide training (column 2). e statistical signi cance shown is for the following null hypothesis: Shift = 0, Dilation = 1, Truncation = 0, which basically benchmarks the case that the distributions are the same. Consider column 1 rst. As may be expected, Shift > 0, implying that labor productivity is higher for rms that provide training. e Dilation factor is less than 1 and statistically signi cantly so (at the 1 percent level). is implies that the distribution of labor productivity is more homogeneous among rms that provide training than those that do not. In other words, labor productivity is more dispersed and heterogenous among rms that do not provide training. Figure B1.1 illustrates the point graphically. e results are qualitatively similar when comparing the distribution of labor productivity in the bottom half versus top half of the NUTS2 regions in terms of the percentage of rms that provide training (column 2). To summarize, training provided by rms to their workers bene ts the relatively less productive rms more and thereby narrows the dispersion of labor productivity. As a result, training allows the less productive rms to catch up with the more productive rms. is can have important e ects on the possible (mis)allocation of resources, with consequent e ects on the overall productivity of the regions and countries (see Hsieh and Klenow 2009; Heise and Porzio 2022). Table B1.1. How training a ects the distribution of (log of ) Figure B1.1. Distribution of labor productivity is more labor productivity of rms and NUTS2 regions homogenous among rms that provide training (1) (2) Shift 1.973*** 2.668*** (0.249) (0.290) Dilation 0.866*** 0.817*** (0.021) (0.024) Truncation -0.002 .0004 (0.006) (0.010) R-squared 0.986 0.984 Observations 17,236 17,292 Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. For Dilation, the signi cance level is for the deviation from 1. Bootstrapped standard errors with 500 replications shown in parentheses. NUTS = Nomenclature of Territorial Units for Statistics. Source: Original calculations for this Brief based on World Bank *** p<0.01 Enterprise Surveys. 6 Enterprise Note No. 43 Table 1 Relationship between the use of skilled and highly educated workers and jobs growth Dependent Share of Firm employs Share of Share of Employment growth variable: university- university- skilled among secondary- rate, (%, annual) educated educated production educated workers (%) workers Y:1 N:0 workers workers (Marginal effects) (%) (%) (1) (2) (3) (4) (5) (6) Exporter Y:1 6.787*** 0.085*** 0.170 -3.132* 1.810** 0.981 N:0 (1.673) (0.025) (2.174) (1.845) (0.802) (1.037) Foreign 10.351*** 0.072** 0.148 -0.769 2.438** 1.254 ownership Y:1 N:0 (2.588) (0.029) (2.978) (2.661) (1.169) (1.888) Number of 1.607** 0.166*** -6.865*** -0.163 workers (logs) (0.748) (0.012) (0.886) (0.815) Share of skilled 0.074** -0.026** among production (0.029) (0.012) workers (%) Share of semi- -0.020* skilled among production (0.011) workers (%) Share of university 0.024 educated workers (%) (0.038) Multi establishment 1.556 0.011 1.715 -1.052 2.078** 3.607** rm Y:1 N:0 (1.877) (0.030) (2.829) (3.070) (0.834) (1.690) Age of rm -0.438 0.003 2.447* -0.088 -2.352*** -3.212*** (logs, years) (0.669) (0.012) (1.318) (1.182) (0.455) (0.775) Number of workers -1.972*** -2.517*** 3 scal years ago (logs) (0.327) (0.452) Industry dummies Yes Yes Yes Yes Yes Yes (ISIC, 2 digit) Constant 4.005* 56.437*** 65.030*** 15.779*** 17.678*** (2.181) (4.453) (4.637) (2.056) (2.770) Number of observations 3,915 3,913 9,233 8,344 8,873 3,779 R-squared 0.306 0.082 0.048 0.076 0.127 Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: Huber-White robust standard errors clustered on NUTS2 level in brackets. Logit (marginal effects) estimation in column 2 and ordinary least squares (OLS) in all the other columns. NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. *** p<0.01, ** p<0.05, * p<0.1 low-skilled workers. e share of skilled production workers is growth, which may lead to more jobs overall. also lower for rms that use manual production processes. e empirical evidence on the issue is mixed in general However, this relationship becomes weak and statistically (see, for example, Balsmeier and Martin 2019; Jung et al. 2017) insigni cant after accounting for rm size. and in the EU-27 countries, in particular. First, controlling for convergence or the initial level of employment at the rm, the Employment growth growth rate of employment over the last three scal years One concern is that greater use of skilled workers is labor signi cantly declines as the share of skilled and semi-skilled saving and therefore it may hinder jobs growth. Skilled labor is workers rises (table 1, column 5). is result is driven by often accompanied by greater use of computers, robots, and di erences between rms within NUTS2 regions rather than other labor-saving technologies. It also embodies greater across regions. us, macro-level studies that explore human capital than low- or unskilled workers, which may di erences across regions but not within regions may not detect reduce the need for additional workers. However, it is also a lower growth rate of employment associated with greater use possible that skilled workers may boost rm productivity and of skilled and semi-skilled workers. Second, there is no 7 Enterprise Note No. 43 signi cant relationship between the growth rate of employment regions. is could be because of diminishing returns to and the share of workers that have a university degree (table 1, education, given that poorer regions typically have fewer column 6) or secondary education. Overall, the evidence on the university-educated workers. Another reason could be more relationship between the use of skilled and highly educated imitation and innovation possibilities in the poorer regions that workers and jobs growth in the EU-27 countries is complement university-educated workers. inconclusive. At the rm level, labor productivity is higher for rms that have a higher proportion of skilled workers among production Labor productivity workers, but this relationship is not statistically signi cant. A higher share of university- and secondary-educated However, across NUTS2 regions, there is a strong and workers is associated with higher labor productivity. However, signi cant positive relationship between the two. Likewise, this relationship is weak and statistically insigni cant at lower higher shares of semi-skilled and low-skilled workers across quantiles of labor productivity, and large and signi cant at NTUS2 regions is signi cantly and negatively correlated with higher quantiles in EU-27 countries. For instance, a one labor productivity. ere is sharp di erence in these standard deviation increase in the share of university-educated relationships at low versus high levels of labor productivity. workers is associated with an increase in labor productivity by 2 at is, at the NUTS2 level and the rm level, the relationship percent of the initial level (insigni cant at the 10 percent level) between labor productivity and the share of skilled workers is at the 20th percentile of labor productivity and by 24.6 percent positive and signi cant at lower quantiles of labor productivity. (signi cant at the 1 percent level) at the 80th percentile. us, It is also much larger than at the higher quantiles. In fact, at the it is the more productive rms that take advantage of more rm level, there is no signi cant relationship between the two educated workers, while the less productive rms are at the relatively high quantiles of labor productivity. ese completely deprived of any bene ts. Assuming that there are results for low- versus high-labor productivity quantiles are the substantial gains to be reaped by improving productivity at the opposite of what was found for university-educated workers. low end, policy makers should try to increase educated workers’ e implication is that skilled workers and university-educated usefulness or use other policy tools for the less productive rms. workers are not always mere substitutes. At the NUTS2 regions level, a higher share of university-educated workers is positively correlated with labor Wages productivity at relatively low levels of income (below the Does greater use of skilled workers lead to more unequal median), but there is no signi cant correlation between the two distribution of wage per worker? Unfortunately, data on the at high levels of income. us, poorer regions bene t more wage rate by workers are not available. Hence, this analysis from an increase in university-educated workers than the richer compares the distribution of the log of the total labor cost per Map 1 Low-skilled and high-skilled NUTS2 regions Low Skilled High Skilled No data Data from the World Bank Group (WBES) Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics. 8 Enterprise Note No. 43 worker (a proxy for the wage rate) at the rm level across groups educated managers) are more likely in NUTS2 regions that of rms and across NUTS2 regions depending on their use of have a higher share of adults (25–64 years) with tertiary skilled production workers. education. e probability that a rm has a highly educated top For this, rms are divided into two groups: those in which manager does not show any signi cant correlation with income skilled production workers make up less than 50 percent of the per inhabitant across the NUTS2 regions. However, it increases workforce, and the rest. Similarly, NUTS2 regions are divided signi cantly with the share of adults in the NUTS2 regions into two groups: those where the share of skilled production who have completed tertiary education. With all the baseline workers in a typical rm is more than 50 percent (high-skilled controls included, a one standard deviation increases in the NUTS2 regions), and the rest (low-skilled NUTS2 regions). share of the tertiary-educated adult population is associated Map 1 shows the high- and low-skilled NUTS2 regions. Box 2 with an increase in the probability of a rm having a highly compares the wage distributions in the two rm-level groups educated top manager by 5.5 percentage points. is is an and the two NUTS2 regions level groups, using the economically meaningful increase, given that on average 51 methodology of Combes et al. (2012) as described in box 1. percent of the rms have a highly educated top manager in the e analysis suggests that greater use of skilled workers does not EU-27 countries. lead to higher wage inequality across rms. High manager education is associated with several rm characteristics. As may be expected, larger rms, exporters, Manager education rms with foreign ownership, and rms with proportionately Top managers with a bachelor’s or higher degree (highly more university- educated and secondary-educated workers are Box 2: Does greater use of skilled workers lead to higher wage inequality across firms? Table B2.1 presents the results of an analysis using the Combes methodology. e statistical signi cance shown in the table is for the following null hypothesis: Shift = 0, Dilation = 1, Truncation = 0. Column 1 provides the results for the group of rms with a higher share (above 50 percent) and lower share (below 50 percent) of skilled workers in their workforce. Column 2 provides the results for the high-skilled NUTS2 regions and the low-skilled NUTS2 regions. Because the results in columns 1 and 2 are similar, it su ces to focus on the results in column 2. e Shift parameter is positive, implying a higher average wage rate in the high-skilled NUTS2 regions than in the low skilled ones. e Dilation factor is less than 1, implying that the distribution of the wage rate is less homogenous in the low-skilled NUTS2 regions. e truncation factor is negative and signi cant, implying that very low-paying rms are more likely to survive in high-skilled NUTS2 regions than in low-skilled NUTS2 regions. Overall, these results indicate that the dispersion in the wage rate across rms declines as rms employ more skilled production workers. Of course, the issue of how within- rm wage inequality responds to greater use of skilled workers remains to be explored. Table B2.1. Comparing the distribution of wage rate of rms and NUTS2 regions with a high versus low share of skilled workers (1) (2) Shift 4.897*** 5.138*** (1.384) (1.125) Dilation 0.563*** 0.547*** (0.129) (0.107) Truncation -0.236*** -0.124*** (0.076) (0.047) Observations 8,188 8,188 Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: NUTS2 regions have between about 800,000 and 3 million inhabitants. For Dilation, the signi cance level is for the deviation from 1. Bootstrapped standard errors with 500 replications shown in brackets. NUTS = Nomenclature of Territorial Units for Statistics. *** p<0.01 9 Enterprise Note No. 43 more likely to have a highly educated top manager. However, Last, rms with a larger share owned by a single family are there is no signi cant relationship between the share of skilled, signi cantly less likely to have a highly educated top manager. semi-skilled, or low-skilled workers and the likelihood of e same holds for rms with a larger share of management having a highly educated top manager. e likelihood of a rm positions occupied by a single family. having a highly educated top manager declines with the age of Manager’s education matters for the quality of the rm. management. e overall quality of management is On average, women top managers are more likely to be signi cantly higher for rms that have highly educated highly educated than men top managers. is result is entirely managers (table 2, column 1). However, this relationship varies driven by NUTS2 regions that have a relatively high share along two important dimensions. First, quantile regressions (above median) of the tertiary-educated adult population. One reveal that while rms at all quantiles of the management score possibility here could be a selection e ect, which is that relative bene t from a highly educated manager, the impact is much to men, women face glass ceilings in their career progression bigger at intermediate quantiles than at low and high quantiles. and therefore only the highly educated ones aspire to become us, policies that aim to improve management practice via top managers. ere is some evidence that higher education better education of the top managers are likely to be more may help women compete better with men top managers. at e ective when directed to rms at intermediate level of is, labor productivity of rms with a woman top manager is management quality. Second, rm size matters, with the signi cantly lower than of rms with a man top manager. is positive relationship between management quality and higher gender-based gap exists only among rms that do not have a education of the top manager being driven by small and highly educated top manager. us, higher education seems to medium rms. For large rms, there is no signi cant eliminate the disadvantage that women top managers have relationship between management quality and higher relative to men top managers. education of the top manager. Table 2 Relationship with the top manager's education level Dependent Management Labor Direct Employment Firm spent on Tax rates is a variable: Quality productivity exports growth rate R&D Y:1 N:0 major obstacle (logs) (% of sales (%, annual) (Marginal Y:1 N:0 effects) (Marginal effects) (1) (2) (3) (4) (5) (6) Top manager has 0.059*** 0.133*** 2.511** 1.160* 0.055*** 0.016 bachelor’s or higher degree Y:1 (0.015) (0.046) (1.007) (0.613) (0.016) (0.021) N:0 Exporter Y:1 N:0 0.028* 0.314*** 1.449** 0.125*** -0.012 (0.014) (0.068) (0.641) (0.018) (0.021) Foreign ownership Y:1 0.054*** 0.279*** 13.087*** 0.060 -0.023 -0.062* N:0 (0.015) (0.070) (2.652) (0.965) (0.028) (0.035) Multi establishment 0.078*** 0.248*** -0.037 2.897*** 0.040** -0.014 rm Y:1 N:0 (0.014) (0.058) (1.107) (0.931) (0.016) (0.029) Age of rm (logs, -0.018** 0.030 0.706* -2.664*** -0.013 0.005 years) (0.007) (0.024) (0.421) (0.482) (0.009) (0.011) Number of workers 0.061*** -0.020 1.954*** 0.054*** -0.016 (logs) (0.006) (0.020) (0.391) (0.016) (0.016) Number of workers 3 -2.362*** scal years ago (logs) (0.297) Industry dummies Yes Yes Yes Yes Yes Yes (ISIC 2 digit) Constant 0.271*** 11.777*** 0.904 15.167*** (0.026) (0.087) (1.712) (1.841) N ( rms) 4,992 7,503 7,819 7,544 7,794 7,727 R-squared 0.249 0.303 0.252 0.099 Source: Original calculations for this Brief based on World Bank Enterprise Surveys. Note: Huber-White robust standard errors clustered on NUTS2 regions in brackets. Columns 5 and 6 contain marginal effects from logit estimation evaluated at the mean value of the explanatory variables. All other columns are based on ordinary least squares (OLS) regression. NUTS2 regions have between about 800,000 and 3 million inhabitants. NUTS = Nomenclature of Territorial Units for Statistics; R&D = research and development. *** p<0.01, ** p<0.05, * p<0.1 10 Enterprise Note No. 43 Labor productivity is signi cantly higher by about 14 managers will be more e ective in improving rm performance percent for a rm with a highly educated top manager than for if they are properly targeted to the sorts of rms identi ed in a rm with a less educated top manager (table 2, column 2). this Brief that tend to bene t more from such policies: less While manager’s higher education is positively and signi cantly productive rms, rms that need greater training for workers, related to labor productivity at most quantiles of the labor rms in poorer regions, rms that need greater take-up of productivity distribution, the relationship is much stronger university-educated workers to increase productivity, and rms (more positive) at the relatively high quantiles. For instance, the at the intermediate level of management quality. labor productivity of a rm with highly educated manager is higher by about 5.1 percent at the 20th percentile of labor productivity, and by a much larger 16.2 percent at the 80th Notes percentile value. In short, higher-labor productivity rms seem 1 e sample used is described in detail in the rst Brief in this series: to bene t more from a highly educated top manager than lower-productivity rms. https:/ documents1.worldbank.org/en/publication/documents-reports/documentdetail/099253311142421263 https://documents1.worldbank.org/en/publication/documents-re ports/documentdetail/099253311142421263 ere are other potential bene ts of having a highly 2 is analysis focuses in particular on subnational regions varying educated manager. Having a highly educated top manager is between about 800,000 and 3 million inhabitants, known in the EU geographical classi cation system, the Nomenclature of associated with more exports (table 2, column 3), more job Territorial Units for Statistics (NUTS), as NUTS2 regions. creation (table 2, column 4), and higher likelihood of R&D 3 at is, the marginal relationship between training and labor activity (table 2, column 5). e channels through which more productivity is stronger (more positive) for the least productive educated managers improve rm performance is di cult to rms. Results are based on a quantile regression approach. While the impact of training is signi cant at most points of the labor ascertain from the data. For one, there is no consistent and productivity distribution, it is greater at lower quantiles than at robust evidence that more educated managers are successful in higher quantiles. For instance, labor productivity is higher for a lowering the obstacles faced by a rm (see column 6 for an rm that provides training by 65 percent at the 25th percentile example). One possibility is that rms that face greater value of labor productivity and by a much smaller 21 percent at the 75th percentile value. obstacles are more likely to hire highly educated managers. is makes it di cult to identify the true impact of highly educated managers on the level of severity of the obstacles. References Concluding remarks Balsmeier, B., and M. Woerter. 2019. “Is is Time Di erent? How Digitalization In uences Job Creation and Destruction.” Research Some of the main ndings can be summarized as follows. Policy 48 (8): 103765. As discussed in the rst Brief in this series and reiterated here, Combes, P-P, G. Duranton, L. Gobillon, D. Puga, and S. Roux. 2012. “ e Productivity Advantages of Large Cities: Distinguishing more than any other obstacle, rms regard inadequately Agglomeration from Firm Selection.” Econometrica 80 (6): 2543–94. educated workers as the biggest obstacle to their operations. Frazis, H., M. Gittlemman, and M. Joyce. 2000. “Correlates of e correlation between use of or the di culty in nding Training: An Analysis Using Both Employer and Employee skilled or adequately educated workers and top managers and Characteristics.” Industrial & Labor Relations Review 53 (3): 443–62. Heise, S., and T. Porzio. 2022. “Labor Misallocation across Firms and the income per inhabitant in the NUTS2 regions is weak. us, Regions.” NBER Working Paper 30298, National Bureau of economic development is unlikely to automatically solve the Economic Research, Cambridge, MA. problem of inadequately skilled and educated workers and Hsieh, C.-T., and P. J. Klenow. 2009. “Misallocation and managers in the EU-27 countries. Training provided by rms is Manufacturing TFP in China and India.” Quarterly Journal of Economics 124 (4): 1403–48. an e ective channel for improving the skills of workers because Jung, S., J-D Lee, W-S Hwang, and Y. Yeo. 2017. “Growth versus it boosts labor productivity and does not substitute for Equity: A CGE Analysis for E ects of Factor-biased Technical education or skills acquired outside the rm. High education of Progress on Economic Growth and Employment.” Economic the top managers is important because it improves Modelling 60 (C): 424–38. Lange, G-M, Q. Wodon, and K. Carey. 2018. e Changing Wealth management quality and labor productivity. Policies that aim of Nations 2018: Building a Sustainable Future. Washington, DC: to boost education and skills among workers and/or top World Bank. e Enterprise Note Series presents short research reports to encourage the exchange of ideas on business environment issues. e notes present evidence on the relationship between government policies and the ability of businesses to create wealth. e notes carry the names of the authors and should be cited accordingly. e ndings, interpretations, and conclusions expressed in this note are entirely those of the authors. ey do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its a liated organizations, or those of the Executive Directors of the World Bank or the governments they represent.