Is There a Size-Induced Market Failure in Skills Training?

A skilled and educated workforce can support the competitiveness of enterprises of all sizes. However, smaller firms may face greater challenges in developing human capital. We explore differences between smaller and larger firms in offering skills training and in hiring workers with more formal education. Drawing on a dataset of enterprises in five Asian countries, we find major size-based differences in education and training. While smaller firms train less, they also are less inclined to view an inadequately skilled workforce as a major constraint on their operations. It may be that smaller firms are content to occupy niches in a low-skills equilibrium. Our empirical results do offer the possibility, however, that a size-induced market failure in skills training may coexist with a lower regard for skills. The policy implications are not only that governments can reduce the costs for firms to train, but also that micro and small firms need to be sensitized to the benefits of skills upgrading.


INTRODUCTION
Small and medium-sized enterprises (SMEs) face a range of constraints in seeking to establish, sustain, and grow their operations. Access to credit is perceived to be a key constraint but others are also important. Among the nonfinancial constraints is the difficulty of developing a skilled and educated workforce. Skilled workers increase the quality of goods and services produced and raise the productivity and competitiveness of the enterprise.
Developing a skilled workforce can contribute to performance of enterprises of all sizes. However, there may be barriers to hiring and training good workers that are related to enterprise size. These barriers we call "size-induced market failures" in skills and education, and like other market failures they can constrain SMEs from reaching their full potential. Understanding these market failures is important for designing effective public policy interventions. Furthermore, understanding the nature of the market failure is important for designing effective interventions.
Finding ways to improve the survival and performance of SMEs is not a trivial matter. These firms, including microenterprises, are important for the economies of Asia. Although individually small in size, they are massive in number and make a large aggregate contribution to economic value creation and welfare. At least 97% of all Asian enterprises are SMEs and they account for more employment on aggregate than large enterprises. Their contribution to output ranges widely between countries but generally accounts for 30%-50% of gross domestic product. Larger SMEs are engaged in exporting and/or supplying factories that do export. 1 The analysis in the succeeding sections is based on firm-level data from five Asian countries. We divide the enterprises into four size categories and look for trends and possible patterns of association regarding training and education. We also use a linear probability model to test whether firms that find the lack of skilled workers as an enterprise constraint seek to overcome that constraint by providing their own training.

ENTERPRISE SIZE AS A CONSTRAINT
There is considerable evidence that a more skilled and educated workforce raises enterprise productivity and performance (Vandenberg and Trinh, Forthcoming). Furthermore, the impact holds for smaller firms as it does for larger ones. The results suggest that firms should be eager to raise human capital. The reality is that larger firms are much more likely to hire and train skilled workers than smaller firms. This difference may constitute a size-induced market failure.
An enterprise can raise human capital in two ways: hiring and training. Hiring allows an enterprise to secure workers with good skills, either fresh graduates from technical schools, training institutes, colleges, and universities, or experienced workers who are already part of the labor force. Small firms find it difficult to hire both types of workers because skilled workers will opt for jobs with better pay, benefits, and job stability, which tend to be offered by large firms and the public sector. The high "churning" rate among SMEs (high birth but also high death rate) means that SMEs are perceived to offer less long-term job security. Highly skilled workers, knowing the market for their skills, will self-select to secure employment with large firms; small firms will lose out unless they are dedicated to hiring skilled labor and willing to offer a competitive package. Furthermore, small firms have weaker links with training and education institutions, because new hiring is less frequent and at lower numbers than for large firms.
In addition, micro and small firms are often controlled by a single founder/owner/ manager. This entrepreneur manages many aspects of the firm and has less time to dedicate to the recruitment of workers. There is commonly no established human resources department or even a single human resources staff person and there is little time to develop links with colleges and training institutes to recruit good young workers, either as graduates or through internships and apprenticeships.
In developing economies-and also some developed ones-the education system may not be turning out adequately skilled graduates. They may receive training in a specific field, but the skills they obtain are not of adequate quality or relevant to the needs of employers. The problems of inadequate quality and employer relevance result in a "skills mismatch," which is a public policy failure that leaves employers searching for solutions (Martinez-Fernandez and Powell, n.d.). 2 The main solution is for the enterprise itself to train its workers. This allows the enterprise to build on the general education and skills of fresh recruits, or providing firm-specific training to more experienced workers.
Training can be provided either informally through on-the-job mentoring or through formal training programs. Formal programs are offered either on-site (i.e., at the enterprise but away from the production process) or off-site (at a training institution to which the enterprise sends workers).
Many SMEs, especially microenterprises, will resort only to on-the-job training. Ashton et al. (2008) have challenged what they see as the "myths" that micro and small enterprises (MSEs) are not interested and generally do not engage in training. They argue that while these firms may exhibit low levels of formal training, they are involved in considerable nonformal training in which skills are passed from experienced and technically adroit workers to newer and less experienced ones at the firm. 3 Hill (2004, p. 9) also suggests training by SMEs is "more likely to be reactive and informal [and]…also it tends to be short term." Capturing the impact of this type of training on firm performance is difficult. Batra and Tan (2003), who did have data on formal and informal training in their multicountry study, found that the most efficient firms combine both types of training but that informal training by itself was negatively correlated with firm efficiency.
Offering formal training to workers is difficult for small firms. The cost per worker is high in organizing on-site training using a contracted trainer because the number of workers is small. As well, not only is it costly to send workers off-site for training, small firms are more dependent on each worker to keep production going and meet demand. Training is best provided in off-peak times of the day, week, month, or year but external training providers may not want to accommodate these preferences. SMEs are also less likely to use training subsidies from public training funds that they have contributed to through a mandatory training levy (Lee, Forthcoming). Lack of awareness of how to access the subsidy, lack of time to make the application, or disinterest in organizing training may be factors resulting in the low uptake by smaller enterprises. 4 There is also the possibility that small firms are not interested in providing training, meaning that they offer less training not because they find it difficult or costly, but because they do not see the benefits for the firm. Because there is less available evidence linking training to outcomes in small firms than there is for large firms, owners may think there will be no impact (Hill 2004). The owner, who is also normally the manager and in micro firms may also be a member of the production staff, may have more pressing issues to address. As well, the enterprise may be engaged in nontechnical service activities (e.g., retailing) in which training is less important than it is for manufacturing or technical services. Even owners who appreciate the value of skilled workers may be reluctant to train because they fear that trained workers, especially those with certified skills, will be poached by other firms and the training investment would be lost. 5 If this is the case, then low levels of training among SMEs may not be the result of a market failure, and any efforts to make training cheaper or more available may have little impact on raising training levels.

EMPIRICAL ANALYSIS OF THE PROVISION OF SKILLS AND EDUCATION IN ASIA
We analyze the phenomenon of size-induced market failures in skills training and education through an empirical investigation using data from Asia. Statistics, differentiated by enterprise size, are assessed to establish basic patterns and a plausible storyline. We then conduct a linear probability analysis to confirm these trends and explore additional possibilities.
Our empirical analysis is based on data from five Asian countries: the People's Republic of China, Indonesia, Malaysia, Thailand, and Viet Nam. These countries are lower-and upper-middle-income countries. We use data from the Enterprise Surveys of the International Finance Corporation, an agency of the World Bank Group. The surveys use a standard methodology and questionnaire for surveying firms in developing countries and thus data from different countries can be combined without difficulty.
The sample comprises 5,586 enterprises (Table 1). It includes over 1,000 firms each for the PRC and Thailand and somewhat fewer for the other three countries. We impose a uniform size classification on the five countries based on the number of permanent, full-time workers. 6 Microenterprises are defined as having 5-9 workers, small firms have 10-49 workers, medium-sized firms have 50-249 workers, and large firms have 250 or more. The survey includes only formal sector enterprises and excludes enterprises with fewer than 5 workers. This classification is very close to that of the European Union (EU), with the exception that in the EU micro firms are defined as having 0-9 workers. It is also broadly consistent with classifications in Asia, although the range of country definitions is broad. 7 Using this classification, about 11% of the firms are micro, with a very high proportion of 41% for Indonesia (Table 1). Small and medium-sized firms make up most of the sample at 32% and 39%, respectively. About 19% of firms are large. Thus, there is a fairly good distribution across the four size classes.  Table 2 provides firm characteristics related to the share of production workers and skilled workers and the owner's or manager's perception of skills. The share of production workers in the enterprises is fairly consistent across the four size classifications. It ranges from 76% for small firms to 83% for large firms. The share is not the smallest for microenterprises. It may be that nonproduction activities (e.g., service and support functions) are handled by the owner-manager in microenterprises. In contrast, small firms hire workers for the support functions, and because full-time workers are discrete units, they represent a disproportionately larger share of the workforce compared with medium-sized and large enterprises, for which the share of nonproduction workers is smaller. 8 The most striking aspect of the data presented in Table 2 relates to the share of production workers who are skilled. The highest share is among microenterprises (79%) and is nearly twice as high as in the three other size categories (43% to 48%). This suggests that non-micro firms employ a core of skilled workers as well as other workers who are part of the production process but play a nonskilled supporting role. These workers would move supplies and goods into and out of the production line, store goods, and fulfill other production support functions. Micro firms are less likely to employ such workers and instead skilled workers carry out these tasks in the normal course of their work and in conjunction with their skilled activities. Non-micro firms may also employ more standardized machinery that can be operated by unskilled workers. It should also be noted that what constitutes a "skilled worker" is not defined in the questionnaire and therefore is based on the perception of the owner or manager completing the survey, and that perception may differ between micro and non-micro firms.
Micro and small firms are less likely than medium-sized and large firms to consider an inadequately skilled workforce as a key business constraint. Only about a quarter (27%) of microenterprises felt it was a major or severe constraint compared with a majority of small firms and 69% of large firms. There are two ways in which to interpret these figures. One is to take them at face value and interpret with the other data. As such, micro firms-and also, but to a lesser extent, small firms-have a higher share of skilled production workers and see skills as less of a constraint than do medium-sized and large firms. The lower share of skilled workers among the larger firms reflects an inability to secure the skilled workers needed and hence the perception that inadequate skills is a constraint. The second interpretation is derived from the idea that the results are based on perception. Larger firms might see it as a major or severe constraint because they have fewer other main constraints. Smaller firms may have other major concerns that reduce the sense that skills are a key constraint. These interpretations will become clearer as we investigate the data further.

TRAINING AND EDUCATION CHARACTERISTICS
We next investigate the training activities of enterprises and the education level of their workers (Table 3). It is clear that larger enterprises are more likely to offer formal training to their workers. Formal training includes both that which is organized at the firm but taking place away from the normal work of the enterprise and that which is conducted away from the enterprise at the training provider. It excludes on-the-job training by experienced colleagues. Only 7% of microenterprises offer formal training to their workers. The figure rises significantly to 39% for small firms and then to 71% and 79% for medium-sized and large firms, respectively. For these latter groups the effort to provide training is impressive. Furthermore, differences by firm size are also apparent when we consider the share of workers trained in each enterprise. The share rises with firm size: only 7% of the microenterprise workforce receives training, but this figure rises to 39% at small firms and up to 64% at large firms. That is, large firms that do offer training train on average about two-thirds of their workers, and micro firms train less than a tenth.
These basic statistics provide the following story: smaller firms have a higher share of skilled workers and do not see a lack of skilled workers as a major constraint for the enterprise. As a result, they are less likely to offer training to their workers, and when they do offer training they offer it to a fairly small share of their workers. We now turn to the education level of the workforce in these enterprises to add to or modify our story.
In the survey, enterprises are asked to provide the average years of schooling of the workforce. There are several discrete choices; we focus on the ranges of 10-12 years of education and more than 12 years. Workers with 10-12 years of schooling would have completed some or all of secondary school. Workers with more than 12 years of schooling would have completed some postsecondary education, which may be in skills and other technical training or in other areas (e.g., business administration, sciences). More precise measures would be better but are not available from the survey. We also note that the indicator is based on the view of the owner or manager who completes the questionnaire.
The results are presented in Table 4 and are similar to those for training. The average worker education level is lower in smaller enterprises than in larger firms. Only 18% of micro firms have a workforce with an average of at least 10-12 years of education, compared with 47% of both medium and large firms. The pattern carries for education beyond 12 years. Only 4% of micro firms employ a workforce with more than 12 years of education compared with 14% of large firms.
Where does that leave our line of analysis? Smaller firms have a higher share of skilled workers, train less, and have a lower tendency to see skills as a constraint. However, their workers have much fewer years of education. Four-fifths of microenterprises employ a workforce in which the average worker has less than a high school diploma. The share is three-quarters of small firms. But nearly half of medium-sized and large firms boast a workforce with a high school education, on average. The majority of non-micro enterprises see an inadequately trained workforce as a major or severe constraint. We also know that the tendency to provide formal training rises with enterprise size. These are basic tendencies that appear from viewing data from four discrete categories of enterprise size. But are these two trends related? That is, are enterprises that see skills as a constraint likely to try to overcome that constraint by training their workers? The idea is that if skilled workers are not available in the labor market, the enterprise might take on that responsibility itself. For this we estimate the relationship by running a linear probability regression. The dependent variable-what we want to explain-is the tendency for an enterprise to offer formal training. The key explanatory variable is the perception that an inadequately skilled workforce is a major or very severe constraint. We also add other variables, including a size variable to see whether there is a significant size-induced element as well, and we include other control variables. Thus, our model for estimation is the following: T = c + β 1 P + β 2 H + β 3 S + β 4 A +, β 5 T + β 6 L + β 7 C + ε in which: T = whether the enterprise offers formal training or not H = human capital; whether the average education of the workforce is 10 years or more and we include control variables: S = size class of the enterprise (micro, small, or medium-sized; with large as the base case) A = age of the enterprise, in years T = sector L = location (subnational province or state) C = country The constant term is c, the error term is ε, and the coefficients are represented by βs, following standard notation. We estimate a linear probability model, first using the ordinary least squares (OLS) estimator and the second one using the generalized method of moments (GMM) estimator. In the latter case, we instrument our main variable of interest, which is whether the enterprise perceives an inadequately skilled workforce as a major or very severe constraint. We instrument this variable due to the possibility that it may be endogenous. Whereas it is most likely that enterprises train because they feel they do not have access to adequately trained workers, it may be that the causality works in the other direction or that there are unobserved or unobservable characteristics that affect both training and the perception regarding the skills of the workforce. As a result, we instrument the variable using location (i.e., subnational state or province) and subject the instrument to the Hansen test, the weak instrument test, and the under-identification test, all of which it passes. The generated fitted values are then used in the estimation of model 2.
The results confirm our intuition and the trends shown in the analysis of the basic statistics of the two previous tables. Enterprises that feel that the workforce is inadequately skilled seek to overcome this constraint by training themselves. This result is statistically significant and with the expected sign (Table 4). The coefficient for education (10 years or more) is also significant and positive. It is interesting that enterprises with a more educated workforce have a greater tendency to offer training than enterprises with a lower level of education. This may indicate that firms feel that secondary education does not provide the skills needed in the workplace. In addition, firms may feel the need to build on basic education with training in specific, enterpriseor sector-relevant skills.
The coefficients on the enterprise size variables are also interesting. The base case is large enterprises and the results (based on both the coefficients and the negative signs) indicate that micro firms train the least, followed by small firms and then medium-sized firms. Large enterprises train the most. What is particularly important here is that these variables are significant even with the inclusion of the variable on the perception of inadequate skills. In other words, the tendency for smaller enterprises not to train is only partially explained by their perception that they do not need to train (because the workforce skills are not inadequate). This result offers the possibility that micro and small enterprises do face a market failure. That is, they train less partly because they don't feel the need to train and partly because it may be difficult to organize training. More detailed data, which is not available in our dataset, is needed to confirm whether smaller firms do in fact face a market failure of this nature. The significance of the size variables may also be explained by missing explanatory variables.
The model works equally well whether instruments are used or not. The only noticeable difference is that firm age goes from being positive significant in the first estimate to negative insignificant in the second. As the coefficients are small and the variable is not a key variable of interest, this change is of little importance. Standard errors in parentheses.
Reference category for micro, small, and medium-sized enterprises is large enterprises. OLS = ordinary least squares, IV = instrumental variable, GMM = generalized method of moments.
Source: Authors

CONCLUSIONS AND POLICY IMPLICATIONS
SMEs face several distinct market failures based on enterprise size. We have sought to investigate the "size-induced market failures" in skills and education. Using a firm-level dataset that covers five Asian countries, we have broken the sample into micro, small, medium-sized, and large enterprises and investigated differences. The results are interesting, if not conclusive.
The results clearly indicate that the propensity for enterprises to train their workers increases with size. Micro firms train the least and large firms train the most. The share of the workforce at each enterprise that has access to this training also increases with size. On average, only 7% of workers in a micro firm have access to training, with the figure rising to 64% for large firms. A similar trend is evident for education. A medium-sized or large enterprise is 2.5 times more likely than a microenterprise to employ a workforce with an average schooling of 10 years or more.
Our data are not sufficiently detailed to determine whether these differences in hiring and training result from a market failure or instead whether they are the result of explicit decisions by the enterprise. That is, is it that enterprises can't do more training or hire more educated workers or is it that they won't? Our data tells us that micro and small firms are much less likely than medium-sized or large firms to see the lack of skilled workers as a major obstacle to the development of their enterprises. This may suggest that they are consciously deciding against increasing human capital because they do not feel the need. However, our linear probability analysis suggests that a lower level of training among smaller firms is not fully explained by a lack of interest, which leaves open the possibility that part of the explanation derives from market failures.
Our related research, using the same dataset, suggests that the provision of enterprise-based training and hiring better-educated workers results in higher productivity (Vandenberg and Trinh, Forthcoming). However, it is possible that smaller firms do not see the benefit of such investment, or that the costs are perceived to outweigh the benefits. Another explanation is that smaller firms may be content to operate in market segments where the skills requirement is not particularly high and workers with these skills are readily available. They may be content to operate in what has been called a "low-skills equilibrium" (Wilson and Hogarth 2003).
Our findings, although not conclusive, do have implications for policy. For mediumsized and large firms and some small firms that feel the skills level is a major constraint, the government can provide support through enterprise-based and other training options. For micro firms and some small firms, there may be a need to first convince entrepreneurs of the benefits of training and then offer support that reduces barriers to training and ensures that the benefits outweigh the costs.