Higher Education,Vocational Training and Performance of Firms

Abstract

This article contributes to the debate on the efficacy of traditional forms of education versus vocational training. The effects of technical education (leading to an engineering degree or diploma) and vocational training in engineering on the performance of Indian firms are analysed using regression models based on the Cobb–Douglas production function, enhanced to incorporate education and training. Instrumental variable approach is used to establish the direction of causality. It is found that that when a larger share of workers in a particular sector has a college or university-level technical education or vocational education in technical fields, there is a positive impact on firm performance in those sectors. Further, higher education in a general field seems to consistently benefit the organised manufacturing sector, while some levels of school education appear to benefit the unorganised sectors.

JEL codes: I-23; L-60; M-53

Keywords

Technical Education Vocational Training Education Firms

1. Introduction

Developing countries with limited resources face the challenge of identifying ideal educational or training programmes for their population. The standard approach of returns to education analyses are individualistic in nature, based on Mincerian models that focus on the wages earned by a worker with certain levels of education (Falck et al., 2016; Malamud & Pop-Eleches, 2006). This article attempts to take this analysis a step further by trying to understand the impact of education and skills of the workforce on the performance of firms. It is found that technical training, whether it culminates in a degree, diploma or vocational skill development, leads to improved performance by firms in both the organised and unorganised sectors, while formal higher education in a general field has a consistently positive effect on the performance of firms in the organised sector.

The Indian government has been emphasising on skill-based training programmes.¹ The question is whether these training programmes actually improve employability and productivity in the economy, a question that is relevant both for the demand and supply sides of the market for education. Students face a choice of which level and type of education or training to opt for. Should they complete their school education? After this, should they opt for higher education or vocational training? Meanwhile, policymakers have to take decisions on resource allocation across different forms of learning.

There has been a long-standing debate over the choice between vocational training and general education. Yet, there are limited studies on the connection between vocational education programmes and development, as highlighted by McGrath (2012). Bennell (1996) has carried out a systematic survey of the existing evidence on this issue in developing countries. Opponents to vocational training point out that general education has better labour market outcomes in terms of earnings, employment and flexibility in occupational choice. Acknowledging data limitation, Bennell (1996) finds that the literature indicates that general education has a higher rate of return than vocational training.

Malamud and Pop-Eleches (2006) tries to address this issue by studying educational reform in Romania. They find little difference between a group that opted for general education and another that shifted to vocational training, in terms of employment, family income or wages. They also demonstrate a social preference for general education over vocational training, with the result that individuals with a general education were less likely to remain single. A relatively recent study by Rzepka (2018) uses a German National Educational Panel Study to compare employment outcomes of a group of individuals with only vocational training with another group that followed up vocational training with formal college education. The latter experiences a loss of income for the years spent in college and substantial uncertainty thereafter. The long-run job stability is similar in both groups; however, there are non-monetary benefits of a college education in terms of status in society.

There is relatively sparse literature on the association between education or training and industry-level outcomes, especially in the context of technical skills. Falck et al. (2016) use the framework suggested by Mincer (1974) to estimate returns to information and communication technology (ICT) skills across 19 developed economies. Their results show that an increase in ICT skills leads to significantly higher wages. Using data from Turkey, Atasoy (2011) studies various types of ICT skills and concludes that having at least one ICT skill is associated with 20%–30% higher probability of obtaining employment.

There is limited evidence of the effect of training on firm productivity. Chaibi et al. (2015) investigate causal links between ICT, e-skills and firm performance in Luxemburg. They find a positive impact of ICT usage on the successful implementation of new projects, but no relationship between ICT staff and training on firm performance using firm-level data.

Some studies indicate a positive effect of the use of information technology (IT) on firm performance. For instance, Kim (2004) and Basu et al. (2004) find that IT has a positive influence on firm productivity in Korea and the United States, respectively, but the latter finds that this may not be applicable in the case of the United Kingdom. Motohashi (2001) and Motohashi (2003) use Japanese firm-level data to explore the effect of information network use on firm productivity using cross-sectional and panel data analysis, respectively. Motohashi (2001) tries to assess the impact of IT on firm productivity using a log–log Cobb Douglas production function (Cobb et al., 1928). They find that IT networks play an important role in productivity in certain tasks, like production, sales and inventory control system, and logistics management, but not in other areas like human resource management or management planning systems. In fact, there is a negative impact on tasks like customer relations and financial transactions. Greenan et al. (2001) carry out correlation analysis between various ICT and R&D indicators and measures of firm performance of labour productivity, total factor productivity, and average wage and skill composition, using French manufacturing and services firm data. Under the assumption that the use of IT entails use of IT skills, it may be concluded that there is likely to be a positive relationship between IT skills and productivity of a firm.

In the context of India, Berman et al. (2005) try to assess if there was skill-biased technical change using the Annual Survey of Industries (ASI) data during the 1990s. They point out that some states, like Gujarat, turned favourably towards skill upgradation and others, like West Bengal, lagged behind. More recently, Sharma and Singh (2013) also uses the ASI dataset to show that investment in IT-related capital stock has a positive influence on a firm’s gross value added (GVA). Basant et al. (2006) undertake a firm-level survey of about 500 firms each in India and Brazil and finds that Brazilian firms use ICT more intensively than Indian ones, but that it does influence productivity in both countries.

This article is an attempt to contribute to this literature by presenting a comparative analysis of the role of higher education of three types—graduate and postgraduate and above degrees in a general field; formal engineering or technical education at any level; and vocational training in a technical field—on the performance of firms in the organised and unorganised manufacturing sectors. In this study, vocational training includes disciplines such as electrical and electronic engineering, chemical, mechanical and civil engineering, and computer science. The most popular choice of discipline in vocational training is computer science according to the data from the 68th round of National Sample Survey (NSS) carried out in 2011–2012. A workforce with technical skills not only supports the ICT sector but could improve the productivity of the entire manufacturing sector. Some attention is also paid to the foundations of our education system—school-level education. A by-product of this article is the validation of the notion that school education in India is not good enough, or perhaps insufficient, to effectively contribute to firm performance.

This study uses an innovative strategy to convert information on the educational and vocational training of individuals from the 68th round of the NSS into proxy indicators for the skill levels of manufacturing workers.² This is combined with firm-level data from two sources: ASI and the 73rd round of the NSS for the organised and unorganised manufacturing sectors, respectively.³^,⁴ This approach enables us to generate empirical evidence about the linkages between education and firm performance. The direction of causality is established with the help of a unique instrumental variable based on access to education in the form of number of educational institutions per capita at the state level.

The second and third sections of this article discuss the data and the corresponding empirical strategy used in the analysis. The results are presented in Section 4, followed by a conclusion.

2. Data Description

The organised sector plant-level data has been sourced from the ASI, which is an annual survey of registered plants across the country. We use data from two years, 2011–2012 and 2013–2014. Identification of the sector, industry and products of a firm is based on the NIC 2008 at a 5-digit level of disaggregation. There are more than 1,240 5-digit NIC codes in this dataset. Details about the sampling strategy can be obtained from MOSPI (2016). We use the ASI dataset for information about firm performance and the various factors that may influence firm-level performance, including the labour and capital inputs into a firm’s production process (GoI, 2016).

Unorganised sector firm-level data is obtained from the 73rd round of NSS. We select manufacturing sector data based on the NIC 2008 codes. A typical limitation of the unorganised sector is the lack of record keeping by enterprises. Hence, unlike the ASI dataset which provides information for a financial year, this survey collects information relating to about 30 days prior to the survey in order to minimise recall bias. Information about firm performance and capital and labour inputs for the representative month is used for analysis in this case (GoI, 2017).

A major challenge for this study was obtaining data about education and vocational training of workers in firms in the organised and unorganised sectors. In this article, I calculate this information at the 5-digit level of the NIC code using information from the 68th round of NSS. All the datasets provide 5-digit NIC codes in which a firm operates or an individual is employed and we use this to match data across the datasets (GoI, 2013).

This study extracts four categories of information from the 68th round of NSS. First, data on individuals who have attended college and/or university for graduate or higher level education in a general field. Second, individuals with a technical education in engineering or technology at the undergraduate or higher levels. The third set relates to those with vocational education in the engineering disciplines, including electrical and electronics, computer science, and chemical, mechanical and civil engineering. Finally, there is data on individuals with no form of higher education or training; this includes individuals who have attained schooling at the higher secondary, secondary, primary or middle-school level as well as those who have reported that they are illiterate. Further, there is information on which individuals are employed at a 5-digit NIC code level in the sector. Table 1 provides a summary of the data on educational attainment in this survey. A little over one-fourth of the sample has reported to have no level of education; nearly 45% of the surveyed population have education till the middle school level or below; nearly 12% and 8% of the sample have educational attainment till the secondary and higher secondary levels, respectively. The combined share of those with a technical education and vocational training in the technical fields is fairly small, at less than 2% of the sample. Relatively more individuals, that is nearly 8% of the population, have a graduate or postgraduate degree in general fields (GoI, 2013).

Table 1
Educational Attainment in 68th Round of NSS, 2011–2012

Level of Education Sample Size Share of Sample

Illiterate 117,712 26.07

Below primary, primary or middle school 206,404 45.71

Secondary 53,116 11.76

Higher secondary 36,346 8.05

College or University (undergraduate or postgraduate level degree or diploma in general fields) 35,428 7.85

Technical education (undergraduate or postgraduate level degree or diploma in engineering/technology) 3,491 0.77

Vocational training in technology and engineering fields 4,490 0.99

Others 2,561 0.57

Total sample size 451,567 100.00

Level of Education	Sample Size	Share of Sample
Illiterate	117,712	26.07
Below primary, primary or middle school	206,404	45.71
Secondary	53,116	11.76
Higher secondary	36,346	8.05
College or University (undergraduate or postgraduate level degree or diploma in general fields)	35,428	7.85
Technical education (undergraduate or postgraduate level degree or diploma in engineering/technology)	3,491	0.77
Vocational training in technology and engineering fields	4,490	0.99
Others	2,561	0.57
Total sample size	451,567	100.00

Source: Authors’ calculation using 68^th round of NSS data.

We matched the firm-level data from the ASI (for the organised manufacturing sector) and the 73rd round of the NSS (for the unorganised manufacturing sector) with the training and employment data of the 68th round of the NSS using two methods. The first method uses the NIC codes while the second approach uses a combination of the NIC codes and location at the state level. While India has free movement of labour across the nation in principle, there may still be barriers to migration across state boundaries. The second method of matching is used to carry out robustness checks of the analysis under the assumption that there may be locational rigidities in labour markets.⁵

The first step of the matching process involves calculation of the number of respondents in a particular NIC code (or combination of NIC code and state). In the case of the organised manufacturing sector, for the mapping procedure, we select that segment of the NSS 68th round which pertains to people who work in firms that are legally required to register, that is, those whose primary occupation is working in establishments with 10 or more workers aided by power or 20 or more workers without the aid of power. This eliminates firms not legally required to register, to align it with the sampling strategy used by the ASI. In the case of the unorganised sector, it would be expected that, applying the logic given above, we restrict the matching process to establishments with less than 10 workers aided by power or 20 without the aid of power. However, the unorganised sector also consists of firms with a considerably higher number of workers.⁶ Kanbur (2017) characterises several such unorganised firms as those that do not comply with the laws in order to avoid the transaction cost of registration, taxation, etc. In order to include the entire set of unorganised firms in our sample, we select the entire 68th Round of NSS data for the manufacturing sector for the mapping process. This enables us to calculate the share of those under each level of educational attainment mentioned above in the total number which works in a particular NIC code (or combination of NIC code and state). Hence, while there is no direct information about the level of education of the employees of each firm, proxy indicators of the average level of education for each 5-digit NIC code (or combination of NIC code and state) have been created.⁷

It may be noted that the 68th round of NSS survey was carried out in 2011–2012. This study considers two years of the ASI data: 2012 and 2014. The first case (ASI 2012 henceforth) is a pure cross-section analysis. However, it may take two years from obtaining a degree or diploma or completing vocational education to getting a job and starting to contribute effectively to the productivity of a firm. So the second case, (ASI 2014 henceforth), incorporates this time lag into the analysis. The 73rd round of the NSS was carried out in 2014–2015, hence the third case in this article (NSS 2015 henceforth) involves a time lag of three years.

One limitation of this analysis is the lack of information about the quality of the training programme. It is highly probable that workers who have been educated in high-quality schools, colleges, or technical or vocational training institutes would contribute more effectively towards firm performance, but this factor cannot be incorporated in this analysis. Also, complex patterns of educational progress such as obtaining a university education after vocational training or vice versa are not likely to be captured in this analysis.

3. Econometric Specification

3.1. Baseline Model

The econometric specification is based on a Cobb–Douglas production function. Biddle (2012) describes the historical evolution of this approach to studying individual firm productivity using the Cobb–Douglas production function. Several empirical studies were carried out in the late 1920s and 1930s to develop this method of statistical estimation of the relationship between inputs and output of a firm. One example is Bronfenbrenner and Douglas (1939). Subsequently several studies, such as Motohashi (2001), have used this empirical strategy to study the productivity of a firm.

The basic form of the Cobb–Douglas production function is as follows:

Y = A K^{α} L^{β}

The output of a productive unit depends on the inputs—capital, denoted by K and labour denoted by L. The parameters α and β are the output elasticities of capital and labour. A log-linear version of the Cobb–Douglas production function is used to formulate our estimation equation:

\log G V A_{i j} = c + α {(\log G F A)}_{i j} + β {(\log E m p)}_{i j} + ε_{i j}

(1)

Where, Log GVA is log of the gross value added of a firm, that is, the value of the output less the value of intermediate goods used in the production process. Capital is indicated by the log of gross fixed assets (Log GFA). The variable Log Emp refers to the total amount of labour employed in a firm measured in terms of mandays. The subscript i refers to the firm and j indicates the 5-digit NIC code pertaining to the firm. In the second approach to matching datasets, j indicates a combination of the NIC code and state.

3.2. Human Capital

The model is expanded to incorporate human capital following a rich literature on this theme, including Jones and Hall (1999). The model can be depicted as follows:

\begin{matrix} \log G V A_{i j} = c + α {(\log G F A)}_{i j} + β {(\log E m p)}_{i j} + \\ γ {(H K)}_{j} + φ_{k} + ε_{i j} \end{matrix}

(2)

The term HK reflects the share of individuals with different types of educational attainment in the total number of individuals employed in a particular NIC code (or NIC code and state) among the respondents of the 68th round of the NSS. The different types of educational attainment are: graduate or higher level in a general field; technical education at the graduate or higher level in the field of engineering; vocational training in a technical field; and higher secondary, secondary, or middle school and below.

In addition to the six levels and types of education, we also control for illiteracy in some of the models, and this is measured as the share of illiterate workers in a particular NIC code. Also, state-level fixed effects have been incorporated into the model to account for any idiosyncratic differences that may exist within different states.

3.3. Instrumental Variable Approach

Instrumental variable approach is used to address the possibility of endogeneity in the model. The model presented in Equation (2) is estimated by the 2SLS method using instruments that capture access to education or vocational training.

Models where human capital is captured by school-level education are instrumented by the number of schools per capita in a particular state in 2011–2012. The analysis of the impact of college- or university-level education in either a general field or a technical field are instrumented by the number of colleges and universities per capita at the state level in the same year. It may be noted that separate information on the number of colleges and universities of higher education in the general fields and technical fields is not available at a state level, perhaps because several colleges and universities offer programmes in both general and technical education. The models involving vocational training in technical fields are instrumented by the number of stand-alone polytechnic and technical institutes per capita at the state level. In spite of the diversity of technical institutes in this category, this instrument may be treated as a reasonably close proxy for access to vocational training in a state. All of the instruments are obtained from All India Survey on Higher Education, carried out by the Department of Higher Education, Ministry of Human Resource Development, Government of India (GoI, 2014).

These instruments capture the potential for access to education and are likely to influence our independent human capital variables. At the same time, these instruments are primarily determined by the policies of the Ministry of Human Resource Development on the creation and sustenance of educational institutions in the public and private sectors. They are therefore unlikely to have a direct impact on the performance of firms (the independent variable of our model). Nevertheless, the significant F statistic in the first state of the regression models establishes the strength of these instruments.

4. Results

4.1. Baseline

The results of the baseline model are presented in Table 2 and those relating to the extended model are presented in the subsequent tables. These results pertain to the first method of matching based on the NIC codes only.

The columns 1, 2 and 3 of Table 2 pertain to data from ASI 2014, ASI 2012 and NSS 2015, respectively. The results of the formal sector are in alignment with the findings of similar models in the literature. Bronfenbrenner and Douglas (1939) reports that various cross-section estimates of the Cobb–Douglas model that use firm-level data from the United States and Australia find that the coefficient of the log of labour input ranges from 0.65 to 0.75. In general, for most manufacturing firms, nearly three-quarters of the output of a firm can be attributed to labour. In contrast, nearly the entire output of firms in the informal sector can be attributed to labour. This is not surprising, as the unorganised sector typically faces severe capital constraints. This outcome indicates that our baseline regression model is sound enough for us to proceed with further analyses. Also note that the standard errors are clustered at the 5 digit NIC code level.

4.2. Human Capital

Tables 3–8 extend the baseline model to include various combinations of indicators of educational attainments. The roles of the three levels of school education, that is, up to middle school, secondary school and higher secondary school are explored in Tables 3a–3c. It is found that the coefficients of the shares of individuals with these levels of education in a particular NIC code do not have a positive and significant impact on the GVA of firms in the ASI 2014 or in the ASI 2012 datasets. However, in the NSSO 2015 dataset, which consists of the unorganised manufacturing sector, there is some positive effect from school-level education. When the share of workers with up to middle school education or secondary-school education among the total workers in a particular 5-digit NIC code is higher by 0.1, then firm productivity measured by the GVA is higher by 9% and 14%, respectively. The effect of higher secondary-school education on the performance of firms remains insignificant in this case.

Table 2
Baseline Model

Dep Var: Log GVA ASI 2014 ASI 2012 NSSO 2015

Log GFA 0.324 0.416 0.169

(0.019)* (0.014)* (0.018)*

Log Emp 0.788 0.664 1.152

(0.026)* (0.018)* (0.067)*

Constant 4.227 4.467 2.972

(0.265)* (0.289)* (0.266)***

R ² 0.75 0.75 0.64

N 21,424 18,750 82,836

Dep Var: Log GVA	ASI 2014	ASI 2012	NSSO 2015
Log GFA	0.324	0.416	0.169
	(0.019)***	(0.014)***	(0.018)***
Log Emp	0.788	0.664	1.152
	(0.026)***	(0.018)***	(0.067)***
Constant	4.227	4.467	2.972
	(0.265)***	(0.289)***	(0.266)***
R ²	0.75	0.75	0.64
N	21,424	18,750	82,836

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

Table 3a

School Education—ASI 2014

Dep Var: Log GVA	(1)	(2)	(3)
Log GFA	0.324	0.324	0.324
	(0.019)***	(0.019)***	(0.019)***
Log Emp	0.789	0.789	0.788
	(0.026)***	(0.026)***	(0.027)***
Primary & Middle School	−0.190
	(0.168)
Secondary School		−0.127
		(0.193)
Higher Sec School			−0.101
			(0.140)
Constant	4.243	4.223	4.213
	(0.282)***	(0.281)***	(0.275)***
R ²	0.75	0.75	0.75
N	20,247	20,247	20,247

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

Table 3b

School Education—ASI 2012

Dep Var: Log GVA	(1)	(2)	(3)
Log GFA	0.419	0.420	0.421
	(0.014)***	(0.014)***	(0.014)***
Log Emp	0.658	0.658	0.656
	(0.017)***	(0.018)***	(0.018)***
Primary and Middle School	−0.427
	(0.215)**
Secondary School		−0.342
		(0.272)
Higher Sec School			−0.136
			(0.165)
Constant	4.556	4.505	4.469
	(0.313)***	(0.309)***	(0.301)***
R ²	0.75	0.75	0.75
N	17,859	17,859	17,859

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

Table 3c

School Education—NSSO 2015

Dep Var: Log GVA	(1)	(2)	(3)
Log GFA	0.171	0.166	0.163
	(0.017)***	(0.018)***	(0.017)***
Log Emp	1.151	1.154	1.154
	(0.065)***	(0.064)***	(0.064)***
Primary and Middle School	0.916
	(0.492)*
Secondary School		1.426
		(0.611)**
Higher Sec School			1.050
			(0.760)
Constant	2.723	2.786	2.990
	(0.246)***	(0.248)***	(0.255)***
R ²	0.64	0.64	0.64
N	82,561	82,561	82,561

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

The impact of formal higher education in a college or university on firm performance is presented in Tables 4 and 5. While Table 4 depicts the impact of higher education in the general field, Table 5 relates to the impact of the technical fields. It is found that both these types of education have a positive and significant impact on the GVA of firms in the organised and unorganised sectors. In the organised sector, an increase in the share of workers with higher education (general or technical fields) by 0.1 could lead to an increase in the GVA by 3.5%–5.5%. However, it is interesting to note that the same increase in the workforce with higher education in the unorganised sector can lead to an improved firm performance to the extent to 10%–17%. This indicates that the unorganised sector is substantially skill-starved vis-à-vis the organised sector.

An interesting result is that better access to workers with vocational training in the technical fields is also effective in improving firm performance, as depicted in Table 6. An increase in the share of the workforce with vocational training by 0.1 could increase the GVA of firms in the organised sector to up to 6%–10%. In this case also, the skills-starved unorganised sector experiences a much higher impact on firm performance—almost 30% which is three to five times greater than in the organised sector. These results demonstrate that investment in vocational training programmes is a useful solution for enhancing the demographic dividend of our population.

Table 4

Higher Education in General Field

Dep Var: Log GVA	ASI 2014	ASI 2012	NSSO 2015
Log GFA	0.323	0.417	0.162
	(0.019)***	(0.014)***	(0.018)***
Log Emp	0.789	0.658	1.147
	(0.026)***	(0.017)***	(0.063)***
College and Univ	0.353	0.465	1.076
	(0.127)***	(0.168)***	(0.473)**
Constant	4.157	4.404	3.041
	(0.262)***	(0.279)***	(0.247)***
R ²	0.75	0.76	0.64
N	20,247	17,859	82,561

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

Table 5

Higher Education in Technology/Engineering Field

Dep Var: Log GVA	ASI 2014	ASI 2012	NSSO 2015
Log GFA	0.323	0.419	0.166
	(0.019)***	(0.014)***	(0.018)***
Log Emp	0.790	0.658	1.148
	(0.026)***	(0.017)***	(0.065)***
Technical Edu	0.399	0.559	1.707
	(0.152)***	(0.179)***	(0.460)***
Constant	4.185	4.450	3.020
	(0.270)***	(0.293)***	(0.256)***
R ²	0.75	0.76	0.64
N	20,247	17,859	82,561

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

Table 6

Vocational Training in Technology/Engineering Field

Dep Var: Log GVA	ASI 2014	ASI 2012	NSSO 2015
Log GFA	0.323	0.417	0.166
	(0.019)***	(0.014)***	(0.018)***
Log Emp	0.789	0.658	1.149
	(0.026)***	(0.018)***	(0.065)***
Voc Tech	0.594	0.956	2.963
	(0.291)**	(0.326)***	(0.888)***
Constant	4.191	4.453	3.004
	(0.260)***	(0.281)***	(0.255)***
R ²	0.75	0.76	0.64
N	20,247	17,859	82,561

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

All levels of school education and types of higher education or training are included in the regression models presented in Table 7. The results add credence to the conclusion that access to workers with higher education in a general field, technical field or vocational training has a significant effect on the performance of firms in the organised manufacturing sector, while in the unorganised sector, education or training in the technical disciplines is relevant.

Table 7

All Levels and Types of Education

Dep Var: Log GVA	ASI 2014	ASI 2012	NSSO 2015
Log GFA	0.321	0.414	0.150
	(0.019)***	(0.014)***	(0.012)***
Log Emp	0.791	0.660	1.178
	(0.025)***	(0.016)***	(0.055)***
Illiterate	0.089	0.077	−1.277
	(0.170)	(0.226)	(0.365)***
Primary and Middle School	−0.030	−0.267	0.602
	(0.215)	(0.225)	(0.663)
Secondary School	0.166	0.244	0.414
	(0.233)	(0.302)	(0.588)
Higher Sec School	0.040	−0.022	−0.091
	(0.153)	(0.173)	(0.610)
College and Univ	0.351	0.383	0.159
	(0.147)**	(0.187)**	(0.411)
Technical Edu	0.365	0.419	0.947
	(0.150)**	(0.165)**	(0.402)**
Voc Tech	0.540	0.770	1.551
	(0.299)*	(0.313)**	(0.841)*
Constant	4.086	4.410	3.094
	(0.260)***	(0.261)***	(0.302)***
R ²	0.75	0.76	0.66
N	20,247	17,859	82,561

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. State fixed effects (coefficients suppressed). Brackets: NIC 2008 clustered SE.

4.3. Instrumental Variables Approach

Table 8a–d presents the results of the instrumental variable approach. The organised sector analysis for ASI 2014 and ASI 2012 are depicted in Tables 8a and 8b, respectively, and the results are consistent with each other. While the two types of formal higher education are instrumented by the number of colleges and universities per capita at a state level, vocational training is instrumented by the number of polytechnic institutions per capita at a state level. The coefficients of the share of the workforce with all three types of advanced studies—formal higher education in the general field, formal higher education in the technical field, and vocational training—in a particular NIC code remain positive and significant even after accounting for the possibility of endogeneity. Also, the magnitude of impact of increases in the workforce with vocational training on firms is relatively larger.

Tables 8c and 8d show the results for the unorganised sector, using NSSO 2015 dataset. In Table 8c, we see a departure of the results from the previous section. Here, the share of workforce with formal higher education in the general or technical fields no longer has a positive and significant effect on the GVA of firms. Similarly, at the school level, the shares of workforce with education up to the middle and secondary school levels do not have a positive and significant coefficient as was observed in Table 3c. However, the shares of the workforce with vocational training and higher secondary-school education continue to play a positive and statistically significant role in the performance of firms.

Table 8a
Instrumental Variable Model—ASI 2014

Dep Var: Log GVA 1 2 3

Log GFA 0.300 0.306 0.285

(0.010)* (0.008)* (0.012)*

Log Emp 0.791 0.805 0.794

(0.011)* (0.010)* (0.011)*

College and Univ 6.061

(1.353)*

Technical Edu 5.949

(1.184)*

Voc Tech 14.740

(2.920)*

Constant 3.280 3.473 3.881

(0.152)* (0.107)* (0.094)*

R ² 0.52 0.62 0.47

N 19,649 19,649 19,649

Instrument No. of Colleges and Universities Per capita (state level) No. of Colleges and Universities Per capita (state level) No. of Polytechnic Institutes Per capita (state level)

First Stage F stat 37.41* 64.53* 45.7***

Dep Var: Log GVA	1	2	3
Log GFA	0.300	0.306	0.285
	(0.010)***	(0.008)***	(0.012)***
Log Emp	0.791	0.805	0.794
	(0.011)***	(0.010)***	(0.011)***
College and Univ	6.061
	(1.353)***
Technical Edu		5.949
		(1.184)***
Voc Tech			14.740
			(2.920)***
Constant	3.280	3.473	3.881
	(0.152)***	(0.107)***	(0.094)***
R ²	0.52	0.62	0.47
N	19,649	19,649	19,649
Instrument	No. of Colleges and Universities Per capita (state level)	No. of Colleges and Universities Per capita (state level)	No. of Polytechnic Institutes Per capita (state level)
First Stage F stat	37.41***	64.53***	45.7***

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. Brackets: SE.

Table 8b

Instrumental Variable Model—ASI 2012

Dep Var: Log GVA	1	2	3
Log GFA	0.389	0.397	0.363
	(0.014)***	(0.012)***	(0.016)***
Log Emp	0.660	0.655	0.662
	(0.012)***	(0.012)***	(0.014)***
College and Univ	4.907
	(1.307)***
Technical Edu		6.805
		(1.875)***
Voc Tech			17.979
			(3.452)***
Constant	3.974	4.108	4.446
	(0.117)***	(0.096)***	(0.097)***
R ²	0.62	0.59	0.39
N	17,859	17,859	17,859
Instrument	No. of Colleges and Universities Per capita (state level)	No. of Colleges and Universities Per capita (state level)	No. of Polytechnic Institutes Per capita (state level)
First Stage F stat	33.38***	28.32***	41.12***

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. Brackets: SE.

Table 8c

Instrumental Variable Model—NSSO 2015

Dep Var: Log GVA	1	2	3
Log GFA	0.236	−0.362	−0.173
	(0.013)***	(1.175)	(0.102)*
Log Emp	1.174	0.546	0.848
	(0.011)***	(1.261)	(0.087)***
College and Univ	−7.883
	(2.025)***
Technical Edu		223.692
		(478.400)
Voc Tech			320.754
			(89.933)***
Constant	2.104	8.401	5.690
	(0.112)***	(12.565)	(0.902)***
R ²	0.49	.	.
N	79,896	79,896	79,896
Instrument	No. of Colleges and Universities Per capita (state level)	No. of Colleges and Universities Per capita (state level)	No. of Polytechnic Institutes Per capita (state level)
First Stage F stat	95.27***	21.72***	12.67***

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. Brackets: SE.

Table 8d

Instrumental Variable Model—NSSO 2015

Dep Var: Log GVA	1	2	3
Log GFA	0.128	0.226	0.049
	(0.017)***	(0.010)***	(0.024)**
Log Emp	1.144	1.080	1.197
	(0.011)***	(0.016)***	(0.013)***
Primary and Middle School	−20.853
	(5.760)***
Secondary School		−22.761
		(5.445)***
Higher Sec School			22.593
			(3.921)***
Constant	8.321	5.588	2.685
	(1.602)***	(0.734)***	(0.043)***
R ²	.	.	0.16
N	79,896	79,896	79,896
Instrument	No. of Schools Per capita (state level)	No. of Schools Per capita (state level)	No. of Schools Per capita (state level)
First Stage F stat	17.62***	26.34***	56.67***

Source: The author.

Notes: *p < 0.1; **p < 0.05; ***p < 0.01. Brackets: SE.

4.4. Robustness Checks

Appendix B replicates the analysis using the alternate method of matching based on a combination of NIC codes and states in which the firms are located. The results are identical in almost all cases. The share of the workforce with a college/university education has a positive and significant impact on both the organised and unorganised manufacturing sectors. Further, the magnitude of the impact on the latter is consistently higher. In the case of the instrumental variable analysis, the results are consistent with our earlier findings, with only one additional point to note—a formal technical educated workforce also has a significant effect on the unorganised manufacturing sector data.

The simultaneous inclusion of all the levels and types of education in Table B5 ratifies the result that increases in workers with formal technical education and vocational training in the technical fields consistently improve all firm outcomes, while college and university education only seems to have a significant effect in the formal sector.

5. Conclusion

This article contributes to the debate on the choice of vocational training verses higher education in India in the context of firm performance in the manufacturing sector in India. The results indicate that access to workers with all three forms of advanced training and education considered in this study—that is vocational training, higher education in the general fields, and higher education in the technical fields—leads to better performance of firms in the organised and unorganised manufacturing sectors. This is consistent with the results of Rzepka (2018), which shows that vocational training with and without a college degree have similar employment outcomes. These outcomes remain robust to the use of the instrumental variable approach in the case of the organised manufacturing sector. However, in the case of the unorganised manufacturing sector, vocational training is clearly relevant, but higher education in technical fields also has some positive effect.

It also appears that school education does not influence the performance of firms in the organised sector and has limited effects on firms in the unorganised sector. This does not imply that investment in school education is futile, but this level of education by itself is insufficient for improving the performance of firms.

While this article uses an innovative approach to overcoming the limitations of data in assessing the impact of training and education on the performance of firms, there may be some limitations to this approach. Hence, there is need to develop firm-level datasets with detailed information regarding the educational attainment of the workforce to enable a more nuanced analysis. Nevertheless, these results have crucial implications for policies in the domain of higher education. Clearly, there is a need for policy makers to strengthen higher education in India. A policy dilemma faced in this context is regarding effective allocation of limited resources across different types and formats of education and training. These results support investment in all forms of higher education, including vocational training programmes, in order to improve the productivity of the economy.

Footnotes

Acknowledgements

This article has benefited from comments from Dr Rajesh Chadha, Dr Bornali Bhandari and participants at a seminar held at the National Council of Applied Economic Research.

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

Research on the article was funded by the NCAER JP Morgan NSAWI Program.

Appendix A

Robustness Checks

The results presented here pertain to the second method of matching the datasets based on the NIC code and state.

Notes

References

Atasoy

(2011). ICT skills and employment opportunities [Working Paper #11-24]. http://www.netinst.org/Atasoy_11_24.pdf

Basant

, Commander

, Harrison

, & Menezes-Filho

(2006). ICT adoption and productivity in developing countries: New firm level evidence from Brazil and India. (IZA DP No. 2294). https://www.iza.org/publications/dp/2294/ict-adoption-and-productivity-in-developing-countries-new-firm-level-evidence-from-brazil-and-india

Basu

, Fernald

J. G.

, Oulton

, & Srinivasan

(2004). The case of the missing productivity growth, or does information technology explain why productivity accelerated in the United States but not in the United Kingdom? In Gertler

& Rogoff

(Eds.), NBER Macroeconomics Annual 2003, Volume 18. The MIT Press. http://www.nber.org/chapters/c11441

Bennell

(1996). General versus vocational secondary education in developing countries: A review of the rates of return evidence. The Journal of Development Studies, 33(2), 230–247.

Berman

, Somanathan

, & Tan

H. W.

(2005). Is skill-biased technological change here yet?: Evidence from Indian manufacturing in the 1990 [World Bank Policy Research Working Papers]. http://elibrary.worldbank.org/doi/abs/10.1596/1813-9450-3761

Biddle

(2012). The introduction of the Cobb–Douglas regression. Journal of Economic Perspectives, 26(2), 223–236.

Bronfenbrenner

, & Douglas

P. H.

(1939). Cross-section studies in the Cobb–Douglas function. Journal of Political Economy, 47(6), 761–785.

Chaibi

, Youssef

A. B.

, & Aoun

(2015). E-skills, brains and performance of the firms: ICT and ability of firms to conduct successful projects in Luxembourg. Journal of Applied Business Research, 31(3), 781–794.

Cobb

C. W.

, & Douglas

P. H.

(1928). A theory of production. American Economic Review 18(1), 139–165.

10.

Falck

, Heimisch

, & Wiederhold

(2016). Returns to ICT skills [OECD Education Working Papers, No. 134]. OECD Publishing. https://dx-doi-org.web.bisu.edu.cn/10.1787/5jlzfl2p5rzq-en

11.

GoI (2013). Employment and unemployment: NSS 68th round, Schedule 10, July 2011–June 2012. National Sample Survey Office, Ministry of Statistics and Planning Implementation, Government of India, New Delhi.

12.

GoI (2014). All India survey on higher education—2011–2012. Department of Higher Education, Ministry of Human Resource Development, Government of India, New Delhi.

13.

GoI (2016). Annual survey of industries – 2011-12 & 2013-14. Central Statistics Office (Industrial Statistics Wing), Ministry of Statistics and Planning Implementation, Government of India, New Delhi.

14.

GoI (2017). Unincorporated non-agricultural enterprises (excluding construction)—July 2015– June 2016, 73 round. National Sample Survey Office, Ministry of Statistics and Planning Implementation, Government of India, New Delhi.

15.

Greenan

, Mairesse

, & Topiol-Bensaid

(2001). Information technology and research and development impacts on productivity and skills: Looking for correlations on French firm level data [NBER Working Paper No. 8075]. http://www.nber.org/papers/w8075.pdf

16.

Jones

, & Hall

(1999). Why do some countries produce so much more output per worker than others? Quarterly Journal of Economics, 114, 83–116.

17.

Kanbur

(2017). Informality: Causes, consequences and policy responses. Review of Development Economics, 21(4), 939–961.

18.

Kim

(2004). Information technology and firm performance in Korea. In Ito

& Rose

A. K.

(Eds.), Growth and productivity in East Asia, NBER-East Asia Seminar on Economics, Volume 13. University of Chicago Press. http://www.nber.org/chapters/c10752.pdf

19.

Malamud

, & Pop-Eleches

(2006). General education versus vocational training: Evidence from an economy in transition. http://www.columbia.edu/~cp2124/papers/vocational_latest.pdf

20.

McGrath

(2012). Building new approaches to thinking about vocational education and training and development: policy, theory and evidence. International Journal of Educational Development, 32(5), 619–622.

21.

Mincer

(1974). Schooling, experience, and earnings. Columbia University Press.

22.

MOSPI (2016). Instruction manual annual survey of industries (concepts, definitions and procedures). http://microdata.gov.in/nada43/index.php/catalog/143/download/1554

23.

Motohashi

(2001). Economic analysis of information network use: Organisational and productivity impacts on Japanese firms. http://www.rieti.go.jp/users/it/productivity/docs/ICT+networks+Japan.pdf

24.

Motohashi

(2003). Firm level analysis of information network use and productivity in Japan [RIETI Discussion Paper Series 03-E-021]. http://www.rieti.go.jp/jp/publications/dp/03e021.pdf

25.

Rzepka

(2018). Labor market returns to college education with vocational qualifications. Education Economics, 26(4), 411–431.

26.

Sharma

, & Singh

(2013). Information technology and productivity in Indian manufacturing. India Policy Forum, National Council of Applied Economic Research, 9(1), 189–238.

Higher Education,Vocational Training and Performance of Firms

Abstract

Keywords

1. Introduction

2. Data Description

3.1. Baseline Model

4. Results

4.1. Baseline

4.2. Human Capital

Table 2 Baseline Model Dep Var: Log GVA ASI 2014 ASI 2012 NSSO 2015 Log GFA 0.324 0.416 0.169 (0.019)*** (0.014)*** (0.018)*** Log Emp 0.788 0.664 1.152 (0.026)*** (0.018)*** (0.067)*** Constant 4.227 4.467 2.972 (0.265)*** (0.289)*** (0.266)*** R 2 0.75 0.75 0.64 N 21,424 18,750 82,836

5. Conclusion

Footnotes

Acknowledgements

Funding

Appendix A

Robustness Checks

Notes

References

Table 2
Baseline Model

Dep Var: Log GVA ASI 2014 ASI 2012 NSSO 2015

Log GFA 0.324 0.416 0.169

(0.019)* (0.014)* (0.018)*

Log Emp 0.788 0.664 1.152

(0.026)* (0.018)* (0.067)*

Constant 4.227 4.467 2.972

(0.265)* (0.289)* (0.266)***

R ² 0.75 0.75 0.64

N 21,424 18,750 82,836