Abstract

This paper investigates the link between trade liberalization and the job-matching process in India by estimating an aggregate matching function by incorporating trade openness as a proxy for trade liberalization. Monthly data are drawn from the National Employment Service's Employment Exchange, India's only public employment service. The results show that trade liberalization leads to a decline in the number of new hires. This implies the exacerbation of matching efficiency, described by an outward shift of the Beveridge curve. This finding is in accordance with a widely held public view that trade liberalization increases unemployment. Therefore, the Indian government should continue carrying out structural reforms for the Indian economy to deal with the inefficiencies in the labor market due to external liberalization.

1.  Introduction

Before the last decade of the twentieth century, the Indian economy was relatively closed. Average tariff rates exceeded 80 percent, the proportion of India's tradable goods protected by quantitative restrictions was greater than 90 percent, and foreign direct investment (FDI) was limited (Alessandrini et al. 2011). In late 1991, however, India accepted the IMF's bailout program in order to overcome its balance of payments crisis, which also set off unexpected trade reforms. As a result, India's trade openness has more than trebled since then.

Many scholars have extensively studied the effects of trade liberalization on India's economic outcomes. When focusing on the labor market, the various topics investigated include the relationship between trade liberalization and employment (Banga 2005; Mitra 2009; Goldar and Aggarwal 2010; Sahoo 2010), employment structure (Goldar and Aggarwal 2010), unemployment (Hasan et al. 2012), income (Hasan, Mitra, and Ramaswamy 2007, Topalova 2007), income inequality (Kumar and Mishra 2008), labor demand (Berman, Somanathan, and Tan 2005; Chamarbagwala 2006; Sahoo 2010), labor demand elasticity (Hasan, Mitra, and Ramaswamy 2007), poverty (Topalova 2007; Mehta and Hasan 2012), and wages (Chamarbagwala 2006; Banga 2005).

The aim of this paper is to examine the link between trade liberalization and the job-matching process in India. More specifically, this study investigates how trade liberalization affects the job-matching process through matching efficiency. The matching function is simple, but is also the most representative of all methods used for showing labor-market frictions in an economic model, and for explaining why unemployment and job vacancies coexist. The matching function is one of the most important elements in search-matching models and has been widely used in macroeconomics and labor economics. The inputs of the matching function consist of the volumes of job seekers and job vacancies; and output is new hires. The matching function is closely related to the Beveridge curve that depicts the negative association between job vacancies and unemployment. Shifts in the Beveridge curve (i.e., changes in job-matching efficiency) can be examined by using empirical matching functions. To the best of my knowledge, the matching function of the Indian labor market has not been estimated. The research most closely related to this paper may be that of Hasan et al. (2012), who investigate the effect of trade liberalization on unemployment. The indirect link between my paper and Hasan et al. (2012) is the negative relationship between new hires and unemployment.

The data are drawn from various publications of Employment Exchange Statistics published by the Directorate General of Employment and Training (DGE&T) in the Ministry of Labor and Employment in India. The DGE&T gathers data from almost 1,000 employment exchanges across states in India. The frequency of the data used in this paper is monthly and covers the period from April 1990 to March 2012.

With the data from the Indian employment exchanges, the aggregate matching functions are estimated by incorporating trade openness. The degree of trade openness is adopted as a proxy for trade liberalization, and is determined by the sum of imports and exports divided by total GDP. Trade openness incorporates the effects of different trade liberalization and non-trade related policies, such as macroeconomic shocks and policies, geographical attributes, and other factors (Dutt, Mitra, and Ranjan 2009, 38).

The results from the empirical matching functions show that there is a negative correlation between trade openness and new job hires, which implies a positive relationship between trade liberalization and unemployment. This indicates an exacerbation of the matching efficiency that is described by an outward shift of the Beveridge curve. This finding is in accordance with a widely held public view that trade liberalization increases unemployment. This differs from the findings of Hasan et al. (2012), however, who show that overall there is no link between trade liberalization and unemployment. Furthermore, the finding of this paper contradicts that of Dutt et al. (2009), who argue in a cross-country analysis that unemployment and trade openness are negatively related. In contrast to these previous studies, I recommend that the Indian government continue carrying out the necessary structural reforms that address the existing labor market inefficiencies, and allow the Indian economy to respond positively to external trade liberalization.

This paper is organized as follows. Section 2 introduces an overview of the matching function and the Beveridge curve in relation to trade liberalization. Section 3 presents the empirical strategies, including the empirical model, the data, and the econometric issues. Section 4 reports the results of the matching-function estimations, and Section 5 concludes.

2.  Matching function, Beveridge curve, and trade liberalization

2.1  Matching function

The matching function summarizes the job-matching process, which is characterized by trade friction, incomplete information, and heterogeneities between job seekers and firms. It plays a key role in describing the labor market dynamics and efficiencies in search-matching models. It relates the joint movement of job seekers and vacancies to new hires and is generally given by , where H denotes the number of new hires in a given interval, S is the stock of job seekers, and V is the stock of vacant jobs. These variables can be time-series, cross-sectional, or include both dimensions.

The following properties are natural and testable assumptions of the matching function: ; ; and , which indicate that new hires increase with respect to both arguments, that at least one job seeker and one vacancy are required to generate a new hire, and that the matching function is concave with respect to both arguments. Furthermore, for a unique and stable unemployment equilibrium in search-matching models, the matching functions are assumed to be constant returns to scale (Pissarides 2000, 6).

This paper uses a specification of the matching function shown in equation (1).
formula
1
Equation (1) is a Cobb-Douglas specification and is standard in the empirical matching literature (Petrongolo and Pissarides 2001). Other functional forms can also be adopted, such as the translog or the constant elasticity of substitution, but the empirical evidence supports the Cobb-Douglas specification (Pissarides 2000, 6). All the variables are indexed by t because this paper uses monthly time-series data. Taking the logs, equation (1) is transformed to become linear in the parameters in equation (2).
formula
2
α is the elasticity of the matching function with respect to job seekers and is the elasticity with respect to job vacancies. Intuitively, α is the percentage change in H with respect to a 1 percent increase in S. t is the time trend and its coefficient δ can be interpreted as additional effects in the efficiency of the matching process over time.

The degree of returns to scale for the matching function is obtained by the sum of α and β and can be tested under the null hypothesis: α + β = 1 in equation (1) or (2). α and β are also considered in terms of the relative importance (a contribution of matching) of job seekers and vacancies in the job-matching process. In addition, these parameters are often interpreted as the matching shares of job seekers and vacancies in creating new job matches. For example, the matching function with a small α and a large β implies relative low labor demand. This indicates that an additional vacant job leads to a new hire with a high probability, while an additional job seeker makes almost no contribution to the number of new hires (Fahr and Sunde 2004, 411). In this situation, for new job creation, government policies to promote labor demand are recommended.

The matching function has been extensively studied in the macro labor literature. Generally, there are two approaches here: the economy-wide aggregate matching functions that use time-series data, and the disaggregated matching functions that use panel data across regions or industries. Petrongolo and Pissarides (2001) provide an extensive survey of empirical matching functions. Since then, there have been a considerable number of studies on matching functions with various objectives across countries, including studies on the matching function itself, functional forms, the degree of returns to scale, data issues of matching function variables, matching efficiencies across industries or regions, and other topics.

Earlier studies focused on estimating matching-function parameters and the degree of returns to scale of the matching function in European countries and the United States. Recent works tend to investigate particular issues as well as estimation of the matching function itself. Moreover, the empirical matching function and its application are performed not only for advanced nations but also for emerging markets. In Asia's case, Liu (2011) estimates the matching functions for three different groups of job seekers in China: the unemployed, on-the-job searchers, and job seekers who migrate from rural to urban areas. Liu's main finding is that the congestion effects from the latter two groups are significant, especially the effect of on-the-job searches on the unemployed. Kano and Ohta (2004, 2005) investigate matching functions in the Japanese labor market. Kano and Ohta (2004) focus on the long-run relationship between the main variables in the matching function and find that these variables are co-integrated and that the standard within-groups estimator in panel regressions is significantly biased. Kano and Ohta's (2005) main finding is that more urbanized regions reveal lower matching efficiencies, which implies that matching efficiency is negatively related to population density and per capita income.1

2.2  Beveridge curve

The matching function and the Beveridge curve are closely related. A Beveridge curve is a graphical representation of the negative relationship between job seekers (unemployment) and job vacancies. As the production function is described in an isoquant curve, a Beveridge curve depicts a matching function. Along the curve, all possible combinations of job seekers and job vacancies are given the same amount of new hires, .

Figure 1 shows job seekers (S) on the horizontal axis and job vacancies (V) on the vertical axis. In general, the position at the top left of the curve indicates low unemployment and high job vacancies in a boom. The position at the bottom right of the curve describes high unemployment and low job vacancies in a recession. Thus a Beveridge curve summarizes the state of the labor market in the business cycle.

Figure 1.

The Beveridge curve

Figure 1.

The Beveridge curve

Changes in labor market (matching) efficiency cause shifts in the Beveridge curve. Figure 2 shows an outward shift from the origin, which implies a deterioration of the matching process: More job seekers and job vacancies are required to create the same number of new matches. In contrast, an inward shift indicates an increase in the efficiency of the labor market or the matching efficiency.

Figure 2.

Decrease in the efficiency of the labor market

Figure 2.

Decrease in the efficiency of the labor market

A number of factors can cause shifts in the Beveridge curve, such as frictional unemployment (e.g., the length of time it takes for a job seeker to find a job that fits with their training or skill set), structural unemployment (occupational, industrial, and skills mismatch), economic uncertainty (e.g., shortening the hiring process because of an optimistic forecast of future economic condition), the duration of unemployment, a change in the size of the labor force, and so forth.2

Looking at this relationship, δt in equation (2) is interpreted as changes in the efficiency of the matching process over time. This is described by shifts in the Beveridge curve.3 For instance, a negative estimate for δ implies an outward shift of the Beveridge curve.

2.3  Trade liberalization and the matching process

Most international trade models, such as the Ricardian, Heckcher-Ohlin, and even the Melitz-type models, assume perfect competition, a condition that does not allow for unemployment. Recently, scholars (e.g., Felbermayr, Prat, and Schmerer 2011) have introduced economic models that incorporate the relationship between trade liberalization (either opening an economy or reducing [eliminating] tariffs) and unemployment.

In all search-matching related models, the channel through which trade liberalization changes the value of the marginal product of labor leads to an increase or a decrease in the number of job vacancies that firms open. In turn, these changes in vacancies influence new hires, which finally affect equilibrium unemployment.

Because the overall effect of liberalization relies on the share of sectors having a comparative advantage with a higher value of the marginal product of labor, the effect of trade liberalization on the job-matching process, that is on new hires, is an empirical question.

In theory, the effect of trade liberalization is linked to job vacancies through the value of marginal product of labor (e.g., Hasan et al. 2012). In practice, however, there is a possibility that the indirect effect of trade liberalization also plays an important role in the job-matching process. In the case where imported goods replace domestic goods, a number of workers flow into the labor force as job seekers. This may influence the job-matching process, including its efficiency. Graphically, this is expressed by an upward shift in the Beveridge curve. Intuitively, more job vacancies would be needed to produce new hires, because of the additional inflow of job seekers from the firms that disappeared as a result of trade liberalization. In a regression analysis, changes in the matching efficiency, due to trade liberalization, are captured by the correlation between new hires and trade liberalization, after controlling for the volumes of job seekers, job vacancies, and other factors.

In the following sections, I examine the link between trade liberalization and the job-matching process by estimating the matching functions. More importantly, I investigate how trade liberalization affects the job-matching process through matching efficiency.4

3.  Empirical strategies

3.1  Specifications of a matching function with trade liberalization

There are many measures related to trade liberalization, such as tariffs, import quotas, antidumping regulations, export duties, and so forth. Choosing a single measure of trade liberalization is almost impossible. This paper uses trade openness as a proxy for trade liberalization. Trade openness is a useful proxy because its data and matching-function data are tabulated monthly, which means that the impact of trade openness (liberalization) on the matching function can be assessed.5

Trade openness is the ratio of the sum of exports and imports to GDP. It incorporates the effects of many different trade liberalization as well as non-trade related policies, such as macroeconomic shocks and related policies, geographical attributes, and other factors (Dutt et al. 2009, 38). For robustness checks, the import penetration rate (imports/GDP) and the ratio of exports to GDP (exports/GDP) are also utilized in the estimation. To incorporate trade openness in the matching function, in equation (1) is modified to
formula
and equation (3) is the revised econometric model for the estimation. ln A equals c and ln Openness is the natural logarithm of trade openness.
formula
3

3.2  Data description

The data for the matching function variables are drawn from India's employment exchanges (EEIs). These data are published in Employment Exchange Statistics, published by the Directorate General of Employment and Training (DGE&T) in the Ministry of Labor and Employment in India. The EEIs are the only public employment service that helps job seekers across India (DGE&T 2012). These exchanges serve 28 states and 7 union territories, with 966 offices, and provide not only job-matching services between job seekers and businesses, but also vocational guidance and carrier counseling. They also gather labor market information, such as employment and unemployment data published by DGE&T (DGE&T 2012, 1). EEI data include entire areas, occupations, and industries, in India. These data are also used to estimate the state-level unemployment rates, which demonstrate the data's usability and reliability.6

The data used in this paper are time-series with a frequency of one month. The range covered is from April 1990 to March 2012 and, hence, the total range of observations is 264 months. Although more informative state-level data are available, time-series aggregate data are utilized because the data for the 1990s are not available at the state level but only at the aggregate time-series level. This study looks at the 1990s because this was an important period of progressive economic reform, which implies that this is the most important period for the purpose of this investigation.

This paper uses the data for registrations, vacancy notified, and placements in the employment exchange Statistics. These variables correspond to the three key variables for the matching function, which are the number of new hires (H), job seekers (S), and vacancies (V). Trade openness is constructed by using the relevant data from the UN Comtrade and CEIC databases. Table 1 shows the descriptive statistics for trade openness and the variables in the matching function.

Table 1. 
Summary statistics of the variables
VariableNo. ObsMeanStd. Dev.MinMax
H 264 19.68 8.79 3.9 71.9 
S 264 486.96 189.71 191 1654 
V 264 35.12 16.55 10.6 143.7 
Openness 264 0.29 0.11 0.13 0.58 
VariableNo. ObsMeanStd. Dev.MinMax
H 264 19.68 8.79 3.9 71.9 
S 264 486.96 189.71 191 1654 
V 264 35.12 16.55 10.6 143.7 
Openness 264 0.29 0.11 0.13 0.58 

Source: DGE&T (2008, 2012), CEIC database, UN Comtrade database.

Note: Unit of H, S, V is thousand. Openness is the ratio of the sum of exports and imports to GDP. All variables in the time frequency are month.

Figure 3 illustrates the annual trends in the number of job seekers, in both flow and stock values. The “s” here represents the stock value of job seekers at the end of each year and “f” denotes the flow value between January 1 and December 31 in any year. The figure shows that the stock value of job seekers increased from approximately 30 million in 1988 to 40 million in 2011. This implies that the stock of job seekers grew by an average of 1.4 percent per annum during this period. The annual growth rate of the stock value was about 3.1 percent between 1988 and 2001, and it was almost 7 percent per annum before the economic reforms of the 1990s, which implies a sharp increase in unemployment before economic reform. The stock value shows a declining trend after 2001, with the exception of 2006, when it grew by 5.3 percent. The 2008 financial crisis spurred on another upward trend in the number of job seekers.

Figure 3.

Annual trends of job seekers (flow and stock values), 1988 to 2011

Figure 3.

Annual trends of job seekers (flow and stock values), 1988 to 2011

The flow value of job seekers can be considered as the annual inflow of job seekers. The average annual number of persons who were in a “job search” between 1988 and 2011 was 5.8 million, except for 2006, when there were 7.3 million job seekers. The new inflow of job seekers each year during this period was 6 to 7 million. In particular, since the start of the Great Recession in 2008, the flow value shows an increasing trend, which also indicates an increase in unemployment.

Figure 4 presents the flow values of job seekers, vacancies, and new hires. The vacancies and new hires show a close co-movement over time. This co-movement reveals the importance of vacancies on new hires, which implies that new hires are vacancy-dependent. The figure shows that the numbers of vacancies and new hires tended to decline until 2003, at which point they began to increase.

Figure 4.

Annual trends of job seekers, vacancies, and new hires, 1988 to 2011

Figure 4.

Annual trends of job seekers, vacancies, and new hires, 1988 to 2011

Figure 5 illustrates the trend in trade openness in the 1990s and the 2000s. It shows an upward trend, and a particularly sharp increase from 2003 when the Indian economy began to boom. In 1990, India's trade volume was approximately 15 percent of GDP; it increased substantially up to 52 percent in the middle of 2008, plummeting to 35 percent in 2009, but then recovering to over 50 percent of GDP in 2012.

Figure 5.

Trend in trade openness, April 1990 to March 2012

Figure 5.

Trend in trade openness, April 1990 to March 2012

3.3  Econometric issues

As shown in equation (3), this paper uses time-series analyses for its estimations. First, the stationarity of each variable is tested, with the results presented in Table 2. All of the variables are in natural logarithm form, following the regression specification in equation (3). The results of the augmented Dickey-Fuller test indicate that all variables are either stationary with drift or with trend: ln H, ln S, and ln V are stationary, while ln Openness is trend-stationary.

Table 2. 
Unit root test of the variables
Augmented Dickey-Fuller test statistic
SeriesDriftDrift and trend
ln Ht −8.157*** −8.142*** 
ln Vt −6.101*** −6.076*** 
ln St −10.712*** −10.769*** 
ln Openness −1.178 −4.207*** 
Augmented Dickey-Fuller test statistic
SeriesDriftDrift and trend
ln Ht −8.157*** −8.142*** 
ln Vt −6.101*** −6.076*** 
ln St −10.712*** −10.769*** 
ln Openness −1.178 −4.207*** 

Source: Author's estimation.

Note: H is the number of new hires, V is job vacancies, S is job seekers, Openness is the ratio of the sum of exports and imports to GDP. All series are in natural log. Augmented Dickey-Fuller tests are performed for stationarity with no lagged difference. The null hypothesis for the augmented Dickey-Fuller test is a unit root of the series (the Augmented Dickey-Fuller test examines whether a time series follows a unit root process). In the Augmented Dickey-Fuller test, one can also include the differences of the lagged values. The data covers April 1990 to March 2012 (total observations is 264). ***Statistically significant at the 1 percent level; **statistically significant at the 5 percent level; *statistically significant at the 10 percent level.

In this study's regression analysis, the dependent and independent variables are flow variables. In the estimation of the production function or the matching function, theoretically, the dependent variable is a flow value and the explanatory variables are stock values, which causes an endogeneity problem. For example, in equation (2), H2011 is the number of new hires during 2011 (a flow value), whereas S2011 and V2011 are cumulative values at the end of 2011 (31 December 2011, a stock value). In this case, the depletion of H in a year leads to decreases in S and V, which cause the endogeneity problem, due to reverse causality. Thus, the lagged values of S and V are generally used for instrumental variables. Instrumental variables are not used in this study, however, because flow values are used for S and V, as proxies for stock values. In the regression, what is important is how variations of S and V explain variation of H. Changes in the stock values of S and V are determined by the flow values of S and V and, thus, the flow values can be used as proxy variables for the stock values of S and V. More importantly, the stock values of job vacancies are not available and thus, it is inevitable that we must use the flow values of job seekers and vacancies.

Several estimation methods are used in the regression analysis, including ordinary least squares (OLS) with a Newey-West standard error; a feasible generalized least squares (FGLS); and the autoregressive (AR) model.7 These methods are used to correct for any serial correlation of the disturbances, which can cause biased estimates. In addition, the first lag of the dependent variable is controlled to deal with lagged effects of covariates.8

The problem of endogeneity also rises with trade liberalization. In a recession, the number of new hires may decrease and, as a result, unemployment increases. In this situation, there is a possibility that the government would change its external policy toward one of protectionism. This may cause a decrease in trade openness, which is endogenous because of reverse causality. Granger causality tests are thus performed to deal with this issue. The results (in Table 3) show that there is no reverse causality. Therefore, the instruments for trade openness are not used in a regression analysis.

Table 3. 
Granger causality test
Null hypothesisChi-squaredfp
ln Openness does not Granger-cause ln H 9.266 0.010 
ln H does not Granger-cause ln Openness 0.466 0.792 
ln EX does not Granger-cause ln H 8.766 0.012 
ln H does not Granger-cause ln EX 0.166 0.920 
ln IM does not Granger-cause ln H 7.030 0.030 
ln H does not Granger-cause ln IM 0.397 0.820 
Null hypothesisChi-squaredfp
ln Openness does not Granger-cause ln H 9.266 0.010 
ln H does not Granger-cause ln Openness 0.466 0.792 
ln EX does not Granger-cause ln H 8.766 0.012 
ln H does not Granger-cause ln EX 0.166 0.920 
ln IM does not Granger-cause ln H 7.030 0.030 
ln H does not Granger-cause ln IM 0.397 0.820 

Source: Author's estimation.

4.  Results

Table 4 shows the results that use OLS with the Newey-West estimator.9 Table 5 provides the results from the FGLS, and Table 6 presents the estimates from the AR model. In Tables 46, column (1) presents the estimated coefficients in equation (3) without ln Openness; column (2) presents the estimated coefficients with ln Openness; columns (3) and (4) show the results with the first lag of the dependent variable. The differential between columns (3) and (4) is that ln Openness is not controlled in column (3) but it is added in the estimation in column (4). In all cases, month dummies are controlled for seasonality.

Table 4. 
Results using OLS with Newey-West standard error
variables(1)(2)(3)(4)
Constant −0.478 −1.233* −0.595 −1.380* 
 (0.417) (0.637) (0.494) (0.735) 
ln H(1)   0.117 0.122* 
   (0.071) (0.069) 
ln S 0.122* 0.124* 0.129* 0.131* 
 (0.068) (0.067) (0.069) (0.068) 
ln V 0.740*** 0.782*** 0.671*** 0.711*** 
 (0.052) (0.060) (0.059) (0.063) 
ln Openness  0.292**  0.302** 
  (0.134)  (0.139) 
Trend 0.000 0.001** 0.000 0.001** 
 (0.000) (0.001) (0.000) (0.001) 
Observations 264 264 263 263 
R2 0.705 0.712 0.712 0.720 
Constant returns to scale No Yes No Yes 
DW statistic 1.770 1.824 2.051 2.109 
Serial correlation (p-value) 0.069 0.164 0.572 0.243 
variables(1)(2)(3)(4)
Constant −0.478 −1.233* −0.595 −1.380* 
 (0.417) (0.637) (0.494) (0.735) 
ln H(1)   0.117 0.122* 
   (0.071) (0.069) 
ln S 0.122* 0.124* 0.129* 0.131* 
 (0.068) (0.067) (0.069) (0.068) 
ln V 0.740*** 0.782*** 0.671*** 0.711*** 
 (0.052) (0.060) (0.059) (0.063) 
ln Openness  0.292**  0.302** 
  (0.134)  (0.139) 
Trend 0.000 0.001** 0.000 0.001** 
 (0.000) (0.001) (0.000) (0.001) 
Observations 264 264 263 263 
R2 0.705 0.712 0.712 0.720 
Constant returns to scale No Yes No Yes 
DW statistic 1.770 1.824 2.051 2.109 
Serial correlation (p-value) 0.069 0.164 0.572 0.243 

Source: Author's estimation.

Note: The dependent variable is the natural log of the number of new hires (ln H). The last row presents the p-values of the Durbin's alternative test for serial correlation. Its null hypothesis is no serial correlation. Rejecting the null indicates serial correlation of disturbances. The data range from April 1990 to March 2012. ***Statistically significant at the 1 percent level; **statistically significant at the 5 percent level; *statistically significant at the 10 percent level.

Table 5. 
Results using the FGLS
variables(1)(2)(3)(4)
constant −0.236 −0.955** −0.612* −1.442*** 
 (0.374) (0.478) (0.358) (0.432) 
ln H(1)   0.183*** 0.221*** 
   (0.047) (0.046) 
ln S 0.106* 0.110* 0.140** 0.150** 
 (0.064) (0.063) (0.060) (0.058) 
ln V 0.734*** 0.774*** 0.628*** 0.648*** 
 (0.039) (0.042) (0.045) (0.046) 
ln Openness  0.281**  0.302*** 
  (0.123)  (0.101) 
Trend 0.000 0.001** 0.000 0.001*** 
 (0.000) (0.001) (0.000) (0.000) 
Observations 263 263 262 262 
(Adjusted) R2 0.655 0.669 0.727 0.753 
Constant returns to scale No Yes No No 
Rho 0.112 0.087 −0.100 −0.161 
DW d-stat (original) 1.770 1.824 2.051 2.109 
DW d-stat (transformed) 2.030 2.018 1.975 1.976 
variables(1)(2)(3)(4)
constant −0.236 −0.955** −0.612* −1.442*** 
 (0.374) (0.478) (0.358) (0.432) 
ln H(1)   0.183*** 0.221*** 
   (0.047) (0.046) 
ln S 0.106* 0.110* 0.140** 0.150** 
 (0.064) (0.063) (0.060) (0.058) 
ln V 0.734*** 0.774*** 0.628*** 0.648*** 
 (0.039) (0.042) (0.045) (0.046) 
ln Openness  0.281**  0.302*** 
  (0.123)  (0.101) 
Trend 0.000 0.001** 0.000 0.001*** 
 (0.000) (0.001) (0.000) (0.000) 
Observations 263 263 262 262 
(Adjusted) R2 0.655 0.669 0.727 0.753 
Constant returns to scale No Yes No No 
Rho 0.112 0.087 −0.100 −0.161 
DW d-stat (original) 1.770 1.824 2.051 2.109 
DW d-stat (transformed) 2.030 2.018 1.975 1.976 

Source: Author's estimation.

Note: The dependent variable is the natural log of the number of new hires (ln H). The FGLS uses a Prais-Winsten transformation and a Cochrane-Orcutt Iterative Procedure. The Rho is the estimate for the parameter in the AR(1) process of the error term. The DW d-stat (original) is a Durbin-Watson d statistic of the untransformed regression. A DW d-stat (transformed) is a Durbin-Watson d statistic for the transformed regression. The data range from April 1990 to March 2012. ***Statistically significant at the 1 percent level; **statistically significant at the 5 percent level; *statistically significant at the 10 percent level.

Table 6. 
Results using the AR model
variables(1)(2)(3)(4)
constant −0.284 −0.860* −0.603* −1.436*** 
 (0.314) (0.456) (0.360) (0.435) 
ln H(1)   0.108* 0.15** 
   (0.064) (0.061) 
ln S 0.116* 0.118* 0.138** 0.149** 
 (0.061) (0.062) (0.062) (0.063) 
ln V 0.691*** 0.719*** 0.652*** 0.666*** 
 (0.041) (0.044) (0.047) (0.047) 
ln Openness  0.225  0.274** 
  (0.146)  (0.117) 
Trend 0.000 0.001 0.000 0.001* 
 (0.000) (0.001) (0.000) (0.001) 
Observations 264 264 263 263 
Constant returns to scale No No No No 
Autoregressive disturbance     
L1 0.104* 0.091 −0.009 −0.080 
 (0.055) (0.056) (0.089) (0.090) 
L2 0.230*** 0.212*** 0.217*** 0.176** 
 (0.063) (0.066) (0.068) (0.074) 
Sigma 0.199*** 0.198*** 0.198*** 0.196*** 
 (0.008) (0.008) (0.008) (0.008) 
variables(1)(2)(3)(4)
constant −0.284 −0.860* −0.603* −1.436*** 
 (0.314) (0.456) (0.360) (0.435) 
ln H(1)   0.108* 0.15** 
   (0.064) (0.061) 
ln S 0.116* 0.118* 0.138** 0.149** 
 (0.061) (0.062) (0.062) (0.063) 
ln V 0.691*** 0.719*** 0.652*** 0.666*** 
 (0.041) (0.044) (0.047) (0.047) 
ln Openness  0.225  0.274** 
  (0.146)  (0.117) 
Trend 0.000 0.001 0.000 0.001* 
 (0.000) (0.001) (0.000) (0.001) 
Observations 264 264 263 263 
Constant returns to scale No No No No 
Autoregressive disturbance     
L1 0.104* 0.091 −0.009 −0.080 
 (0.055) (0.056) (0.089) (0.090) 
L2 0.230*** 0.212*** 0.217*** 0.176** 
 (0.063) (0.066) (0.068) (0.074) 
Sigma 0.199*** 0.198*** 0.198*** 0.196*** 
 (0.008) (0.008) (0.008) (0.008) 

Source: Author's estimation.

Note: The dependent variable is the natural log of the number of new hires (ln H). Various cases of autoregressive disturbances are applied and the AR(2) is reported because more than two lags of the disturbances are not significant. L1 and L2 stand for the coefficients of the AR(2) process of the error term. Sigma represents the standard deviation of the white-noise error. The data range from April 1990 to March 2012. ***Statistically significant at the 1 percent level; **statistically significant at the 5 percent level; *statistically significant at the 10 percent level.

In all estimations, the elasticity of the matching function with respect to vacancies is much higher than the elasticity for job seekers. The estimates for job vacancies range from 0.63 to 0.78 and are statistically significant, and the estimates for job seekers are approximately 0.11 to 0.15 and several estimates reveal a weak statistical significance at 10 percent. The results confirm the strong co-movement of vacancies and new hires in Figure 4, which implies a shortage of labor demand, where an additional vacancy leads to a new hire, with a high probability, but where an additional job seeker creates almost no new hires.

In most cases, the effect of trade openness is negative and statistically significant, as shown in columns (2) and (4) in Tables 46.10 This finding indicates that the number of new hires declines as the degree of trade openness increases. This implies a negative effect of trade liberalization on new job creation in India, which also indicates a negative impact on unemployment. The results of the negative relationship between trade liberalization and new hires are consistent with a widely held public view that trade liberalization increases unemployment through a deterioration of the matching efficiency. This negative effect of trade openness, in estimation of the matching function, implies an outward shift of the Beveridge curve.

Our findings contradict previous studies, such as Dutt et al. (2009) and Hasan et al. (2012). Dutt et al. find that a robust negative relationship exists between trade openness and unemployment in cross-country analyses. Hasan et al., focusing on India, show that on average there is no correlation between tariff reductions and unemployment. In contrast to these studies, our results suggest that structural reforms, for improving the matching efficiency in the Indian labor market, are inevitable, and hence should be continued in response to external trade liberalization.

5.  Conclusions

In this paper, the relationship between trade liberalization and the job-matching process is empirically examined using data from India's employment exchange, the only public employment service in the country. The findings show that the job-matching process in India is vacancy-dependent. This means that the contribution of job vacancies to the creation of new hires is much larger than that of job seekers. The key results show that there is a negative correlation between trade liberalization and new hires. This finding implies that, as the Indian economy is externally liberalized, the efficiency of the job-matching process deteriorates, and this is shown by an outward shift of the Beveridge curve. Therefore, it is recommended that the Indian government continue to carry out structural reforms for the labor market to improve the matching efficiency. One way is to create more public employment agencies (i.e., offices of employment exchanges in India) to correct imperfect information between job seekers and businesses. Policies to support employers’ recruiting activities could also be effective. Fostering a manufacturing industry that is symbolized by Make in India is one of the best policies for restructuring and rejuvenating the economy, including the labor market, and for improving its global competitiveness. Because globalization is an irreversible trend including trade liberalization, jumping onto the global value chain through Make in India may be a solution to offset the negative effects of trade liberalization on the job-matching process through job creation in India.

This study represents the first step in exploring empirical matching functions for India, various issues need to be addressed in future work. For example, to develop a better picture of the job-matching process and its applications in India, it is important to fully examine the representativeness of the data generated by the employment exchanges. Access to and utilization of more disaggregated data, particularly at the state-level, must also be realized in order to conduct more general and accurate analyses of the Indian labor market.

References

Alessandrini
,
Michele
,
Bassam
Fattouh
,
Benno
Ferrarini
and
Pasquale
Scaramozzino
.
2011
.
Tariff Liberalization and Trade Specialization: Lessons from India
.
Journal of Comparative Economics
39
(
4
):
499
513
.
Anant
,
T. C. A.
,
Rana
Hasan
,
Prabhu
Mohapatra
,
Rayaprolu
Nagaraj
and
S. K.
Sasikumar
.
2006
.
Labor Markets in India: Issues and Perspectives
. In:
Labor Markets in Asia
, edited by
Jesus
Felipe
and
Rana
Hasan
, pp.
205
300
.
London
:
Palgrave Macmillan
.
Banga
,
Rashmi
.
2005
.
Impact of Liberalisation on Wages and Employment in Indian Manufacturing Industries
.
Working Paper
.
New Delhi
:
Indian Council for Research on International Economic Relations
.
Berman
,
Eli
,
Rohini
Somanathan
, and
Hong W.
Tan
.
2005
.
Is Skill-Biased Technological Change Here Yet? Evidence from Indian Manufacturing in the 1990's
.
Working Paper
.
Washington, DC
:
World Bank
.
Cahuc
,
Pierre
, and
André
Zylberberg
.
2004
.
Labor Economics
.
Cambridge, MA
:
MIT Press
.
CEIC Database
. http://www.ceicdata.com/.
Chamarbagwala
,
Rubiana
.
2006
.
Economic Liberalization and Wage Inequality in India
.
World Development
34
(
12
):
1997
2015
.
Directorate General of Employment & Training (DGE&T)
,
Ministry of Labour and Employment, Government of India
.
Employment Exchange Statistics.
Various years
.
Dutt
,
Pushan
,
Devashish
Mitra
, and
Priya
Ranjan
.
2009
.
International Trade and Unemployment: Theory and Cross-national Evidence
.
Journal of International Economics
78
(
1
):
32
44
.
Fahr
,
René
, and
Uwe
Sunde
.
2004
.
Occupational Job Creation: Patterns and Implications
.
Oxford Economic Papers
56
(
3
):
407
435
.
Felbermayr
,
Gabriel
,
Julien
Prat
, and
Hans-Jörg
Schmerer
.
2011
.
Globalization and Labor Market Outcomes: Wage Bargaining, Search Frictions, and Firm Heterogeneity
.
Journal of Economic Theory
146
(
1
):
39
73
.
Goldar
,
Bishwanath
, and
Suresh Chand
Aggarwal
.
2010
.
Informalization of Industrial Labour in India: Are Labour Market Rigidities and Growing Import Competition to Blame
?
Working Paper
.
New Delhi
:
Institute for Economic Growth
.
Greene
,
William
.
2003
.
Econometric Analysis
, 5th ed.
Upper Saddle River, NJ
:
Prentice Hall
.
Hasan
,
Rana
,
Devashish
Mitra
, and
Krishnarajapet V.
Ramaswamy
.
2007
.
Trade Reforms, Labor Regulations, and Labor-Demand Elasticities: Empirical Evidence from India
.
Review of Economics and Statistics
89
(
3
):
466
481
.
Hasan
,
Rana
,
Devashish
Mitra
,
Priya
Ranjan
, and
Reshad N.
Ahsan
.
2012
.
Trade Liberalization and Unemployment: Theory and Evidence from India
.
Journal of Development Economics
97
(
2
):
269
280
.
Hobijn
,
Bart
, and
Ayşegül
Şahin
.
2013
.
Beveridge Curve Shifts across Countries since the Great Recession
.
IMF Economic Review
61
(
4
):
566
600
.
Kano
,
Shigeki
, and
Makoto
Ohta
.
2004
.
Long-Run Matching Relationship in the Japanese Labor Market
.
Empirical Economics
29
(
4
):
921
937
.
Kano
,
Shigeki
, and
Makoto
Ohta
.
2005
.
Estimating a Matching Function and Regional Matching Efficiencies: Japanese Panel Data for 1973–99
.
Japan and the World Economy
17
(
1
):
25
41
.
Kodama
,
Naomi
, and
Tomohiko
Inui
.
2015
.
The Impact of Globalization on Establishment-Level Employment Dynamics in Japan
.
Asian Economic Papers
14
(
2
):
41
65
.
Kumar
,
Utsav
, and
Prachi
Mishra
.
2008
.
Trade Liberalization and Wage Inequality: Evidence from India
.
Review of Development Economics
12
(
2
):
291
311
.
Liu
,
Yang
.
2011
.
Labor Market Matching with Heterogeneous Job Seekers in China
.
Economics Bulletin
31
(
3
):
1980
1992
.
Mehta
,
Aashish
, and
Rana
Hasan
.
2012
.
The Effects of Trade and Services Liberalization on Wage Inequality in India
.
International Review of Economics & Finance
18
;
23
:
75
90
.
Mitra
,
Arup
.
2009
.
Impact of Trade on Service Sector Employment in India
. Working Paper.
New Delhi
:
Institute for Economic Growth
.
Petrongolo
,
Barbara
, and
Christopher A.
Pissarides
.
2001
.
Looking into the Black Box: A Survey of the Matching Function
.
Journal of Economic Literature
39
(
2
):
390
431
.
Pissarides
,
Christopher A.
2000
.
Equilibrium Unemployment Theory
, 2nd edition.
Cambridge, MA
:
MIT Press
.
Sahoo
,
Pravakar
.
2010
.
Employment in India's Organized Manufacturing Sector during Pre- and Post-Reform Periods
.
Asia Pacific Journal of Economics and Business
14
(
2
):
36
48
.
Topalova
,
Petia
.
2007
.
Trade Liberalization, Poverty and Inequality: Evidence from Indian Districts
. In:
Globalization and Poverty
, edited by
Ann
Harrison
, pp.
291
336
.
Chicago
:
University of Chicago Press
.

Notes

*

I wish to thank the participants at the Asian Economic Panel Meeting that took place at Keio University, Tokyo, Japan, 15–16 September 2014, for their helpful comments. In particular, I deeply appreciate the comments of Wing Thye Woo. His meaningful advice and suggestions substantially improved this paper.

1 

Kodama and Inui (2015) examine the impact of globalization on manufacturing employment in Japan. Here, they measure the job-creation rate, which relies on a search-matching model that is related to the matching function.

2 

Pissarides (2000, 97–203), Cahuc and Zylberberg (2004, 533–537), and Hobijn and Şahin (2013).

3 

The main parameters of the matching function, α and β, are related to the curvature of the Beveridge curve.

4 

The effect of trade liberalization, in this paper, can be considered as an indirect effect that is not captured by theoretical search-matching models. This effect could be called “indirect” because the direct effect is explained by changes in the value of marginal product of labor in the search-matching model.

5 

Other possible proxy variables for trade liberalization, such as tariff rates and principle component indexes, are not used because the main variables in this paper are monthly, whereas other proxy variables are yearly.

6 

The data in this study are drawn from the formal (organized) sector only. Because employment in the formal sector constitutes 10 percent of total employment, the data may not be representative (Anant et al. 2006, 229–230). Nonetheless, employment in the formal sector contributes approximately 75 percent of total GDP (Sahoo 2010, 37).

7 

The Newey-West estimator provides consistent standard errors in the presence of serial correlation. The AR estimator assumes an autoregressive disturbance and corrects it for unbiased estimates. The FGLS corrects the first-order serial correlation of disturbance. This paper adopts a Prais-Winsten regression with a Cochrane-Orcutt procedure. A Prais-Winsten regression uses the FGLS method to estimate the coefficients in the linear model, where the errors are autocorrelated (Greene 2003, 273–276). This paper uses a Prais-Winsten transformation so as not to lose the observation; it also uses a Cochrane-Orcutt iterative procedure for estimation efficiency in estimating the serial correlation coefficient of disturbances.

8 

Controlling for the first lag is equivalent to controlling for all the lags of the independent variables and, hence, in the case where the coefficients of the matching function can be interpreted as the current period or the short-run impact of job seekers and vacancies on new job hires.

9 

Various lagged values of ln Openness are used for instruments and the first and second lags are selected.

10 

The results from using the import penetration rate and the ratio of export to GDP are not different from the main implication in this paper. The results for robustness checks can be provided upon request.