• Ешқандай Нәтиже Табылған Жоқ

The impact of training on productivity and wages: firm level evidence


Academic year: 2023

Share "The impact of training on productivity and wages: firm level evidence"


Толық мәтін

1 In fact, with specific training...c it is more efficient if ...ms and workers share the costs and benefits of training. Second, the analysis at the …rm level allows us to control for the endogeneity of training. However, both inputs and the number of trained workers are likely to be related to unobserved productivity elements of a …means that complicates the identification of coefficients.

Frazer (2001) proves that if the …rm is maximizing the pro… 1974) wage equations. This allows us to take a training measure at the …rm level for more than 170,000 Belgian …rms active in the manufacturing and non-manufacturing sectors.

General Results

The exercise is repeated for the entire sample, the manufacturing sector and the service sector. Second, this exercise is repeated, but the sample is now restricted to ... rms included in the productivity estimation sample, where we control for the endogeneity of inputs. By including total factor productivity in the wage equation, we control for these factors, which could be related to the amount of training in each ... rm.

We find that in the entire Belgian private sector, the wages of skilled employees are 16:7% higher than the wages of unskilled workers25. If the workers receiving training are different each year, our estimate of the number of trained workers will be an underestimate. We used the perpetual inventory method to create a measure of the stock of skilled workers and experimented with different depreciation rates that depended on and independent of the number of workers leaving the company.

Our main results are robust to the use of the stock or flow of skilled workers. 2 6We compare the…first column of the wage equation with the…first column of the production function, because in both speci…cations we do not control for the possible endogeneity of training. The same reasoning explains why we compare the second and third specifications of the wage equation with the second and third specifications of the production function, respectively.

In the third specification, we control for the endogeneity of training in both the production function and in the wage equation. The same applies separately to the manufacturing and service sectors, with Chi-square values ​​of 14:1 and 113:0 respectively.

Results Sector Heterogeneity

The fact that we …want the impact of training on productivity to be higher than the impact on wages lends support to the Acemoglu and Pischke (1999a) model that explains why …companies invest in general training of their employees. The results for service sectors are less satisfactory, which is not surprising given the problems with estimating production functions for service sectors. However, we do find …positive and significant effects of training on worker productivity, and for most sectors, the coefficient of training declines when controlling for possible input endogeneity.

The unweighted average of the education coefficient across all service sectors decreases from 0:23 to 0:19 when switching from OLS to the modified Ackerberg et al. Once again we find that the productivity gain from training is slightly larger in the service sectors than in the production sectors. The number of observations refers to the observations used in the first specification, the number of observations used in the second and third specifications is the same as in the productivity tables.

Comparing the impact of job-related training with the impact of general education on wages, a … resembles these in magnitude. In his survey, Card (1999) reports estimates of the impact of a year of education on wages between 5 and 15%, while we estimate the wage premium for trained workers to be 12%. However, note that the average duration of training is only about 2 weeks, meaning much greater returns in a week of training compared to a week of schooling.

3 0 The standard error for the Apparel, Wood Products, and Rubber & Plastics sectors, for example, is much higher than that of the other sectors. They observe training at the sectoral level instead of at the …rm level, so their measure includes possible spillovers of training of workers who move from one employer to another32.

Training as a continuous variable

Most sectors are located on this line which is in accordance with Acemoglu and Pischke34 (1999a). It can be seen that for most sectors, investment in training has a greater impact on a worker's marginal productivity than on his wage. The correlation between the impact on productivity and wages is equal to:76 and is highly significant...

Other Types of Speci…cation

Worker heterogeneity

In Table 12, we report results for estimating equations (A.4) and (A.10), dividing the labor force into trained men, untrained men, trained women, and untrained women to conclude whether there is a differential impact of education on men and women's wages and productivity. For brevity, we only report results for the specification where we control for the endogeneity of training in both the production function and the wage equation. For the production function, we report direct estimates of T; F and F T that measure productivity premiums for the various worker characteristics.

Again, the productivity gap between skilled and untrained workers is greater in the service sector than in the manufacturing sector. However, wages of female workers are approximately 16% lower in the Belgian private sector, which is consistent with previous studies that have shown that female workers earn lower wages than their male counterparts (e.g. Hellerstein and Neumark, 1999). This result can be compared with Booth (1991), who finds that in Britain the wages of male workers increase by 10%, while those of women increase by 16% in response to training.

The average training coefficient in the production function is :14 and :11 in the wage equation when estimating the model for each NACE 2-digit sector separately. Using these data, we calculate the educational level of entry and exit workers and average across all years to obtain an approximation for the educational composition of the labor force of each ... We include the share of workers with higher education as in production function as well as in the wage equation and we estimate both equations controlling for possible endogeneity of inputs (cf.

When there is measurement error in the input variables, ... first or mean difference can exacerbate the bias in the input coe¢sion estimates. The estimate of training impact by sector shows that again for the majority of sectors the productivity premium of trained workers is higher than the wage premium of trained workers as shown in Figure B.1.

Firm speci…c versus general training

However, when we compare the impact of training on productivity and wages, we still see that the productivity premium for skilled workers is substantially higher than the wage premium42, and the difference is statistically significant. In general, you would expect both layoff and layoff rates to be lower in companies that provide a significant amount of training. In summary, we would expect a negative effect of training on dismissal rates under perfect competition and specific training, but also under imperfect competition and general training.

The coefficient on the lagging share of trained employees is not significantly different from zero45. The share of trained employees who stayed behind for two periods even has a positive and significant impact on the resignation rates46. This paper empirically examines the impact of ...rm providing training on both wages and productivity.

After controlling for the possible endogeneity of training, we find that training increases the marginal productivity of an employee more than it increases his wage. Sectors with the greatest effect of training include the Chemical sector and the Rubber and Plastics sector. 34;How to Compete: The Impact of Workplace Practices and Information Technology on Productivity." The Review of Economics and Statistics.

34;The Impact of Training on Productivity and Wages: Evidence from British Panel Data." Oxford Bulletin of Economics and Statistics. 34;Product Differentiation, Multiproduct Firms, and Estimating the Impact of Trade Liberalization on Productivity. " National Bureau of Economic Research Working Paper Series No. European Commission Strengthening Continuing Vocational Training with the Enterprise Initiative." Employment in Europe 2007, O¢ ce for Official Publications of the European Communities, Linking Firms and Workers in Luxembourg: Heterogeneous Work and Returns to Education." mimeo Yale University.

34;Wages, Productivity, and Worker Characteristics: Evidence from Plant-Level Production Functions and Wage Equations.” Journal of Labor Economics.

Table 1: Summary Statistics
Table 1: Summary Statistics


A value of zero for this parameter means that marginal productivity for female workers increases by the same amount as for male workers. Unfortunately, we do not observe the other worker characteristics for skilled versus untrained workers, which forces us to impose some simplifying constraints to find an expression for the labor aggregate that can be estimated. These assumptions are similar to other studies that divide the labor force across different dimensions to reduce the number of parameters to be estimated (e.g. Van Biesebroeck, 2007).

First, we must assume that the relative differences in marginal productivity between two workers who differ in one characteristic are the same regardless of their other characteristics. This means that the relative marginal product of skilled workers compared to unskilled workers is the same for all different types of workers. In addition, we constrain the proportion of one type of worker to be constant in other groups determined by other characteristics.

First, we divide the workers according to their type of work (blue-collar, white-collar or management) and education status. Hence, the unrestricted version of the labor aggregate will consist of six terms, one for each different type of worker. Applying the constraints greatly simplifies the expression for the labor aggregate and is given by.

These relative productivity premiums are the same regardless of a worker's training status. Secondly, we divide the labor force according to the level of education and the qualification status of the workers.



Table 1: Summary Statistics
Table 2: Summary Regressions
Table 4: Impact of Training on Productivity
Table 5: Impact of Training on Wages

Ақпарат көздері


In the world of modern needs of education it is necessary to create such a system of control of knowledge and skills, which not only stated the level of assimilation of the material,