Using data on the iim -Ahmedabad's 2006 batch of mba graduates, we find that graduates belonging to scheduled castes or scheduled tribes get significantly lower wages (19 per cent lower in domestic jobs and 35 per cent lower when foreign jobs are included) than those in the general category. This difference disappears once their lower Grade Point Averages are taken into account, suggesting that the large wage difference is due to the weaker (on average) academic performance of sc/st candidates. The study suggests that in the absence of any serious attempt to equalise school-level opportunities, the current policy of reservations at elite educational institutions will be insufficient to equalise career outcomes even for the minority of sc/st candidates who can benefit from them.
SPECIAL ARTICLEEconomic & Political Weekly EPW november 1, 200845Discrimination in an Elite Labour Market? Job Placements at IIM-AhmedabadSujoy Chakravarty, E SomanathanUsing data on theIIM-Ahmedabad’s 2006 batch ofMBA graduates, we find that graduates belonging to scheduled castes or scheduled tribes get significantly lower wages (19 per cent lower in domestic jobs and 35 per cent lower when foreign jobs are included) than those in the general category. This difference disappears once their lower Grade Point Averages are taken into account, suggesting that the large wage difference is due to the weaker (on average) academic performance of SC/ST candidates. The study suggests that in the absence of any serious attempt to equalise school-level opportunities, the current policy of reservations at elite educational institutions will be insufficient to equalise career outcomes even for the minority of SC/ST candidates who can benefit from them.This work was made possible with data supplied by the Placement Office and the PGP Office, Indian Institute of Management, Ahmedabad, in 2006. We thank Ajay Pandey, Diptesh Ghosh, V K S Nair and Devashish Chakraborty of IIM-Ahmedabad for their help. We also wish to thank Rohini Somanathan, Ashwini Deshpande, Rajiv Sethi and participants at the ISI Conference on Comparative Development, December 2007, for helpful comments. Avinash Bhardwaj provided valuable research assistance. The usual disclaimer applies.Sujoy Chakravarty (sujoy@gmail.com) is at the Department of Humanities and Social Sciences, Indian Institute of Technology, New Delhi and E Somanathan (e.somanathan@gmail.com) is at the Planning Unit of the Indian Statistical Institute, New Delhi.The existence of economic and educational disparities between different castes and genders in India has been extensively documented by social scientists. 1 IntroductionThe proportion of people below the poverty line among sched-uled castes and tribes (SC/STs) is about 50 per cent higher than those among the general population. Access to a reasonable quality of education is far from universal and differs by caste. The fraction of the population that belongs to a SC/ST shrinks as one moves up educational attainment classes. For example, in urban India in 1999-2000, persons belonging to the SC/STs constituted 18.3 per cent of those in the 17-25 age group, but only 11.3 per cent of them had passed high school. Their proportion among college graduateswasonly7.4 per cent [Sundaram 2006]. Similarly, the urbanfemale-to-male wage ratio was found to be 82 per cent for literates and 59 per cent for illiterates [Deshpande and Deshpande 1992] and 78 per cent of this was attributed to differ-ential schooling [Kingdon 1999].It is only recently, however, that social scientists have system-atically studied discrimination against lower castes in the labour market. Madheswaran and Attewell (2007), using National Sample Survey (NSS) data, found that employees from SC/STs in urban salaried jobs in 1999-2000 received wages that were about 30 per cent lower on average than those of other castes. Further, 15 per cent of this differential could not be explained by the measures of education and work experience available in the NSS data. Thorat and Attewell (2007) conducted a field experiment and found that companies discriminate by caste and religion in the frequency with which they contact (fictitious) job applicants with identical resumes. Banerjee et al (2007) conducted similar experiments and found less discrimination in the call-centre industry and none in the software industry.1 It is useful to consider these results in the light of the two ma-jor economic theories of discrimination. In Arrow’s (1972) theory of statistical discrimination, employers have imperfect informa-tion about the productivity of employees, and use group identity to proxy for productivity. This leads them to offer different wages to apparently identical employees from different groups. In con-trast, in Becker’s (1967) description of “taste based discrimina-tion”, employers discriminate even if it means realising a lower level of profit. Prejudice is built into their preferences and the agents performing the discrimination obtain some utility from adversely affecting the economic condition of certain groups even if it means lowering their own earnings. SPECIAL ARTICLEnovember 1, 2008 EPW Economic & Political Weekly46While employers represented in the NSS data presumably have fairly good information about the productivity of their employees, those in the callback studies have far less information regarding the productivity of prospective employees. So it seems that statis-tical discrimination is a possible explanation for why employers discriminate in callback experiments and in entry-level labour markets. But it is much less likely to explain discrimination against employees who have been on the job for a while, which will mostly be the case in theNSS. It is nevertheless possible, that the wage differential observed in the NSS data is due to unobserved productivity differentials because there data include only crude measures of educational at-tainment. On the other hand, in the callback experiments, employers do not see the candidates, only their resumés. One advantage of the study reported here is that data on job candidates’ grades is availa-ble, offering a fine degree of control. At the same time, the labour market is real, not experimental, and candidates typically go through several rounds of interviews with prospective employers,so that employers get a fair amount of information about them. Using data from the Indian Institute of Management, Ahmeda-bad’s (IIM-A) 2006 batch of Post Graduate Diploma in Manage-ment (PGDM, equivalent to anMBA) graduates, we find no evi-dence of discrimination against minorities by employers in place-ments. Controlling for work experience and GPA, there is no wage penalty to being female, or of belonging to a SC/ST. This study is also of interest because IIM-A graduates often come to occupy po-sitions of prestige and power. When historically disadvantaged groups gain access to such positions, this may serve to create role models and break down stereotypes.In addition to examining the possibility of discrimination by gender and caste, we study another form of discrimination that to our knowledge has not been studied in the Indian context. This may take the form of a beauty premium arising from better-looking individuals obtaining a higher reward from economic activity. This has been documented for the US labour market by Hamermesh and Biddle (1994, 1998) and Hamermesh and Parker (2006), and in the UK by Harper (2000). We find weak evidence of this form of discrimination by employers.The next section describes the placement process at IIM-A, while Section 3 describes the data. Section 4 describes the estimation procedure and results, and Section 5 points to the study’s conclusions.2 The Placements Process at the IIMThe business school placements process is the way most business graduates and MBA students obtain employment. Students wish-ing to participate in placement submit their resumés to the place-ment office. After perusing the resumés, companies decide which candidates to interview on campus.2 Interviews take place over several days. Higher paying compa-nies like investment banks and international consulting firms get slots on earlier days, giving them the first chance to make lucra-tive offers to the best students. The interviewers initially screen candidates and progressively shortlist to interview candidates until they make a final offer, so that a company may interview a particular candidate several times. About 5 per cent of the students chose not to go through place-ment because they dropped out of the programme, received other offers, or started their own companies. None of these wereSC/ST students.3 Data and VariablesOur data set comprises 242 final-year students of IIM-A from 2006, who were part of the 250 students who enrolled in the programme in 2004. Of these, we have salary data on 226.3 The salaries of 221 were reported to theIIM by the companies that employed them and five self-reported salaries of those with pre-placement offers or with independently negotiated offers that we obtained through a survey. We have data on gender,SC/ST status,GPA on a 4-point scale, Class 10 and Class 12 public examination marks, college gradua-tion percentage, and years of work experience. The IIM did not, however, provide data on student scores in the Common Admis-sion Test (CAT) and subsequent interviews (in 2004), which together form the basis for selecting students for admission to the IIM’sPGDM programme. A student may get several job offers but only the final offer accepted by a graduating student is reported to the placement office. This is what we use in our analyses. One complication is that about 30 per cent of students receive offers from abroad (Table 1). This necessitates a choice of exchange rate to make rupee and foreign currency salaries comparable.Foreign acceptances (all reported inUS dollars) were converted into rupees as described below. The market exchange rate was Rs44 per dollar at the time of the survey and the World Bank’s purchasing power parity (PPP) rate was approximately eight ac-cording to the Penn World Tables [Heston et al 2006]. The PPP rate probably adjusts too much for cost-of-living differences be-cause executives may consume a larger share of tradable goods than the share of tradables included inGDP. We used the data from an e-mail survey of students who received multiple offers to arrive at the appropriate rate. The idea was to use data from pro-spective employees who received offers in both currencies, and assume that the individual accepted the highest offer using an exchange appropriate for him/her.4 Eleven students who ac-cepted an Indian offer also received foreign offers, and Rs 13.79 per dollar was the mean of the weakest value of the rupee con-sistent with accepting an Indian offer, with these values ranging from Rs 6.35 to Rs 19.33 per dollar. However, looking at the five students who received at least one Indian offer but accepted a foreign offer, we find that the mean of the strongest value of the rupee consistent with this behaviour was Rs 20.53 per dollar, with values ranging from Rs 9.71 to Rs 32.86 per dollar. Owing to this inconsistency, we present the results of our anal-ysis of the determinants of pay using both the mean exchange rates reported above. We also report results using only the sub-sample that accepted Indian offers. Variable definitions and descriptive statistics on the main vari-ables used are given in Tables 2a and 2b (p 47).Table 1: Student Placement by Location in 2006Region PercentageIndia 70Rest of Asia 11US 7Europe 12Source: Placement Office, IIM-A. SCST General SCST General SCST General SPECIAL ARTICLEnovember 1, 2008 EPW Economic & Political Weekly48The main determinants of the starting salary of anIIM gradu-ate are the first-year GPA, GPA in communications courses, and work experience. An increase of one grade point (on a 0-4 scale) in the first-year GPA is estimated to raise the wage by more than 40 per cent (Column (3) of Table 4A and Column (2) of Table 4B). 4.1 Wage Penalty for SC/STs The estimated wage penalty associated with beingSC/ST is between 24 (at Rs 14/$ exchange rate) and 35 per cent (at Rs 20/$) compared to their counter-parts in the general category (Column (1) in Tables 4A and 4B). This estimated penalty is slightly smaller at 19 per cent if only domestic jobs are considered (Table 4A, column (2)). However, once we control for work experience and GPA, the wage penalty to being SC/ST becomes much smaller and not signifi-cant (Columns (3) and (2) in Tables 4A and 4B respectively). SinceSC/ST job candidates have, on average, much lower first-year GPAs, and slightly lower work experience than general category candidates, the wage penalty seems to be operating through these factors. Thus, once we control for the influence of grades, there is no evidence here of discrimination againstSC/ST candidates by employers.The last two columns of Tables 4A and 4B introduce additional controls. GPA in communications courses also has a positive effect on the wage, raising it by 30 or 18 per cent (at Rs 20/$ and Rs 14/$ exchange rates, respectively). An interaction between the SC/ST dummy and first-year GPA is also negative and statistically significant. An F-test reveals that the predicted value of the wage for SC/ST candidates is not significantly different (p = 0.42) from that of general candidates at a first-year GPA of 2, the mean level of GPA for SC/ST candidates. This is seen clearly in Figure 4 (p 49).However, in the same F-test done at a first-year GPA of 2.4, the upper end of the range of SC/STGPAs reveals that the predicted wage forSC/STs is 33 per cent lower than that of the general candidates (p = 0.000). This suggests thatSC/ST candidates are not able to get the same reward as general category stu-dents for higherGPAs. We checked whether this is only becauseSC/ST students have low GPAs (with a maxi-mum of 2.4), by regressing pay against GPA and other variables for general category students within the same range ofGPAs. We find a positive and significant slope coefficient of 0.68. Thus, it is not just the low GPAs of SC/STstudents that accounts for their wages not responding to their GPA.It is worth remarking that the standard deviation in wages between companies is about 2.5 times the standard deviation within companies. Not surprisingly, then, the strong effect of first-year GPA on the wage is mostly due to students with higherGPAs finding employ-ment at higher-paying companies, and not due to their being paid more in the same company. This can be seen from regressions of log pay on the same set of variables, done with company means on the one hand, and with company fixed effects on the other. In the former case, the effect ofGPA is strong (and, as expected, be-ingSC/ST has a strong negative effect ifGPA is omitted from the set of controls). But the latter, within-company, regressions give Table 3: Determinants of Grade Point AveragesCoefficient First-YearGPACommunicationQuantitative GPAGPA (1) (2) (3)SC/ST -0.284***-0.135***-0.404*** (0.038) (0.031) (0.060)Work experience (years) 0.00326 0.00888 -0.00240 (0.0072) (0.0057)(0.011)Attractiveness rating 0.0316** 0.0437*** 0.0448* (0.015) (0.012) (0.023)Female -0.0746*-0.00132-0.166*** (0.039) (0.031) (0.062)Elite college graduate 0.117*** 0.0562** 0.215*** (0.031) (0.024) (0.048)College score (scale 0-4) 0.319*** 0.119*** 0.463*** (0.040) (0.032) (0.063)Constant 1.372***1.434***1.155*** (0.13)(0.10) (0.20)Observations 242242242R-squared 0.530.300.51 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1Table 4A: Determinants of Pay (Exchange Rate: Rs 20/$)Coefficient Log (pay) Log (pay) Log (pay) Log (pay) Log (pay) DomesticJobs (1) (2) (3) (4) (5)SC/ST -0.351***-0.194***-0.08551.411***1.406*** (0.072) (0.043) (0.057) (0.30) (0.31)Work experience (years) 0.0392*** 0.0509*** 0.0409*** 0.0422*** 0.0391*** (0.011) (0.0073)(0.0082) (0.0074)(0.0077)Attractiveness rating 0.0857** 0.0205 0.0774* 0.0718* 0.0623* (0.040) (0.017) (0.039) (0.037) (0.035)First-year GPA 0.606*** 0.738*** 0.596*** (0.092) (0.10) (0.12)SC/ST*GPA -0.730***-0.724*** (0.15)(0.15)Female 0.05070.0341 (0.061) (0.062)Communication GPA 0.303*** (0.11)Second-year GPA 0.0575 (0.052)Constant 13.98***13.77***12.51***12.18***11.82*** (0.081) (0.043) (0.19)(0.22) (0.24)Observations 226160226226226R-squared 0.240.370.380.410.42 Ordinary Least Squares, Robust standard errors in parentheses, clustered bycompany *** p<0.01, ** p<0.05, * p<0.1Table 4B: Determinants of Pay (Exchange rate: Rs 14/$)Coefficient Log (pay) Log (pay) Log (pay) Log (pay) (1) (2) (3) (4)SC/ST -0.243***-0.05420.948***0.920*** (0.051) (0.043) (0.24)(0.24)Work experience (years) 0.0483*** 0.0495*** 0.0498*** 0.0475*** (0.0083)(0.0064)(0.0060) (0.0064)Attractiveness rating 0.0513* 0.0454* 0.0440* 0.0380 (0.027) (0.027) (0.025) (0.024)First-year GPA 0.432*** 0.515*** 0.415*** (0.060) (0.066) (0.071)SC/ST*GPA -0.490***-0.475*** (0.12) (0.12)Female 0.0101-0.00414 (0.041) (0.041)Communication GPA 0.183* (0.093)Second-year GPA 0.0631 (0.047)Constant 13.83***12.79***12.58***12.32*** (0.056) (0.13)(0.15)(0.19)Observations 226226226226R-squared 0.260.380.400.41 Ordinary Least Squares, robust standard errors, clustered by company, inparentheses *** p<0.01, ** p<0.05, * p<0.1
SPECIAL ARTICLEEconomic & Political Weekly EPW november 1, 200845Discrimination in an Elite Labour Market? Job Placements at IIM-AhmedabadSujoy Chakravarty, E SomanathanUsing data on theIIM-Ahmedabad’s 2006 batch ofMBA graduates, we find that graduates belonging to scheduled castes or scheduled tribes get significantly lower wages (19 per cent lower in domestic jobs and 35 per cent lower when foreign jobs are included) than those in the general category. This difference disappears once their lower Grade Point Averages are taken into account, suggesting that the large wage difference is due to the weaker (on average) academic performance of SC/ST candidates. The study suggests that in the absence of any serious attempt to equalise school-level opportunities, the current policy of reservations at elite educational institutions will be insufficient to equalise career outcomes even for the minority of SC/ST candidates who can benefit from them.This work was made possible with data supplied by the Placement Office and the PGP Office, Indian Institute of Management, Ahmedabad, in 2006. We thank Ajay Pandey, Diptesh Ghosh, V K S Nair and Devashish Chakraborty of IIM-Ahmedabad for their help. We also wish to thank Rohini Somanathan, Ashwini Deshpande, Rajiv Sethi and participants at the ISI Conference on Comparative Development, December 2007, for helpful comments. Avinash Bhardwaj provided valuable research assistance. The usual disclaimer applies.Sujoy Chakravarty (sujoy@gmail.com) is at the Department of Humanities and Social Sciences, Indian Institute of Technology, New Delhi and E Somanathan (e.somanathan@gmail.com) is at the Planning Unit of the Indian Statistical Institute, New Delhi.The existence of economic and educational disparities between different castes and genders in India has been extensively documented by social scientists. 1 IntroductionThe proportion of people below the poverty line among sched-uled castes and tribes (SC/STs) is about 50 per cent higher than those among the general population. Access to a reasonable quality of education is far from universal and differs by caste. The fraction of the population that belongs to a SC/ST shrinks as one moves up educational attainment classes. For example, in urban India in 1999-2000, persons belonging to the SC/STs constituted 18.3 per cent of those in the 17-25 age group, but only 11.3 per cent of them had passed high school. Their proportion among college graduateswasonly7.4 per cent [Sundaram 2006]. Similarly, the urbanfemale-to-male wage ratio was found to be 82 per cent for literates and 59 per cent for illiterates [Deshpande and Deshpande 1992] and 78 per cent of this was attributed to differ-ential schooling [Kingdon 1999].It is only recently, however, that social scientists have system-atically studied discrimination against lower castes in the labour market. Madheswaran and Attewell (2007), using National Sample Survey (NSS) data, found that employees from SC/STs in urban salaried jobs in 1999-2000 received wages that were about 30 per cent lower on average than those of other castes. Further, 15 per cent of this differential could not be explained by the measures of education and work experience available in the NSS data. Thorat and Attewell (2007) conducted a field experiment and found that companies discriminate by caste and religion in the frequency with which they contact (fictitious) job applicants with identical resumes. Banerjee et al (2007) conducted similar experiments and found less discrimination in the call-centre industry and none in the software industry.1 It is useful to consider these results in the light of the two ma-jor economic theories of discrimination. In Arrow’s (1972) theory of statistical discrimination, employers have imperfect informa-tion about the productivity of employees, and use group identity to proxy for productivity. This leads them to offer different wages to apparently identical employees from different groups. In con-trast, in Becker’s (1967) description of “taste based discrimina-tion”, employers discriminate even if it means realising a lower level of profit. Prejudice is built into their preferences and the agents performing the discrimination obtain some utility from adversely affecting the economic condition of certain groups even if it means lowering their own earnings.
SPECIAL ARTICLEnovember 1, 2008 EPW Economic & Political Weekly46While employers represented in the NSS data presumably have fairly good information about the productivity of their employees, those in the callback studies have far less information regarding the productivity of prospective employees. So it seems that statis-tical discrimination is a possible explanation for why employers discriminate in callback experiments and in entry-level labour markets. But it is much less likely to explain discrimination against employees who have been on the job for a while, which will mostly be the case in theNSS. It is nevertheless possible, that the wage differential observed in the NSS data is due to unobserved productivity differentials because there data include only crude measures of educational at-tainment. On the other hand, in the callback experiments, employers do not see the candidates, only their resumés. One advantage of the study reported here is that data on job candidates’ grades is availa-ble, offering a fine degree of control. At the same time, the labour market is real, not experimental, and candidates typically go through several rounds of interviews with prospective employers,so that employers get a fair amount of information about them. Using data from the Indian Institute of Management, Ahmeda-bad’s (IIM-A) 2006 batch of Post Graduate Diploma in Manage-ment (PGDM, equivalent to anMBA) graduates, we find no evi-dence of discrimination against minorities by employers in place-ments. Controlling for work experience and GPA, there is no wage penalty to being female, or of belonging to a SC/ST. This study is also of interest because IIM-A graduates often come to occupy po-sitions of prestige and power. When historically disadvantaged groups gain access to such positions, this may serve to create role models and break down stereotypes.In addition to examining the possibility of discrimination by gender and caste, we study another form of discrimination that to our knowledge has not been studied in the Indian context. This may take the form of a beauty premium arising from better-looking individuals obtaining a higher reward from economic activity. This has been documented for the US labour market by Hamermesh and Biddle (1994, 1998) and Hamermesh and Parker (2006), and in the UK by Harper (2000). We find weak evidence of this form of discrimination by employers.The next section describes the placement process at IIM-A, while Section 3 describes the data. Section 4 describes the estimation procedure and results, and Section 5 points to the study’s conclusions.2 The Placements Process at the IIMThe business school placements process is the way most business graduates and MBA students obtain employment. Students wish-ing to participate in placement submit their resumés to the place-ment office. After perusing the resumés, companies decide which candidates to interview on campus.2 Interviews take place over several days. Higher paying compa-nies like investment banks and international consulting firms get slots on earlier days, giving them the first chance to make lucra-tive offers to the best students. The interviewers initially screen candidates and progressively shortlist to interview candidates until they make a final offer, so that a company may interview a particular candidate several times. About 5 per cent of the students chose not to go through place-ment because they dropped out of the programme, received other offers, or started their own companies. None of these wereSC/ST students.3 Data and VariablesOur data set comprises 242 final-year students of IIM-A from 2006, who were part of the 250 students who enrolled in the programme in 2004. Of these, we have salary data on 226.3 The salaries of 221 were reported to theIIM by the companies that employed them and five self-reported salaries of those with pre-placement offers or with independently negotiated offers that we obtained through a survey. We have data on gender,SC/ST status,GPA on a 4-point scale, Class 10 and Class 12 public examination marks, college gradua-tion percentage, and years of work experience. The IIM did not, however, provide data on student scores in the Common Admis-sion Test (CAT) and subsequent interviews (in 2004), which together form the basis for selecting students for admission to the IIM’sPGDM programme. A student may get several job offers but only the final offer accepted by a graduating student is reported to the placement office. This is what we use in our analyses. One complication is that about 30 per cent of students receive offers from abroad (Table 1). This necessitates a choice of exchange rate to make rupee and foreign currency salaries comparable.Foreign acceptances (all reported inUS dollars) were converted into rupees as described below. The market exchange rate was Rs44 per dollar at the time of the survey and the World Bank’s purchasing power parity (PPP) rate was approximately eight ac-cording to the Penn World Tables [Heston et al 2006]. The PPP rate probably adjusts too much for cost-of-living differences be-cause executives may consume a larger share of tradable goods than the share of tradables included inGDP. We used the data from an e-mail survey of students who received multiple offers to arrive at the appropriate rate. The idea was to use data from pro-spective employees who received offers in both currencies, and assume that the individual accepted the highest offer using an exchange appropriate for him/her.4 Eleven students who ac-cepted an Indian offer also received foreign offers, and Rs 13.79 per dollar was the mean of the weakest value of the rupee con-sistent with accepting an Indian offer, with these values ranging from Rs 6.35 to Rs 19.33 per dollar. However, looking at the five students who received at least one Indian offer but accepted a foreign offer, we find that the mean of the strongest value of the rupee consistent with this behaviour was Rs 20.53 per dollar, with values ranging from Rs 9.71 to Rs 32.86 per dollar. Owing to this inconsistency, we present the results of our anal-ysis of the determinants of pay using both the mean exchange rates reported above. We also report results using only the sub-sample that accepted Indian offers. Variable definitions and descriptive statistics on the main vari-ables used are given in Tables 2a and 2b (p 47).Table 1: Student Placement by Location in 2006Region PercentageIndia 70Rest of Asia 11US 7Europe 12Source: Placement Office, IIM-A.
SCST General
SCST General
SCST General
SPECIAL ARTICLEnovember 1, 2008 EPW Economic & Political Weekly48The main determinants of the starting salary of anIIM gradu-ate are the first-year GPA, GPA in communications courses, and work experience. An increase of one grade point (on a 0-4 scale) in the first-year GPA is estimated to raise the wage by more than 40 per cent (Column (3) of Table 4A and Column (2) of Table 4B). 4.1 Wage Penalty for SC/STs The estimated wage penalty associated with beingSC/ST is between 24 (at Rs 14/$ exchange rate) and 35 per cent (at Rs 20/$) compared to their counter-parts in the general category (Column (1) in Tables 4A and 4B). This estimated penalty is slightly smaller at 19 per cent if only domestic jobs are considered (Table 4A, column (2)). However, once we control for work experience and GPA, the wage penalty to being SC/ST becomes much smaller and not signifi-cant (Columns (3) and (2) in Tables 4A and 4B respectively). SinceSC/ST job candidates have, on average, much lower first-year GPAs, and slightly lower work experience than general category candidates, the wage penalty seems to be operating through these factors. Thus, once we control for the influence of grades, there is no evidence here of discrimination againstSC/ST candidates by employers.The last two columns of Tables 4A and 4B introduce additional controls. GPA in communications courses also has a positive effect on the wage, raising it by 30 or 18 per cent (at Rs 20/$ and Rs 14/$ exchange rates, respectively). An interaction between the SC/ST dummy and first-year GPA is also negative and statistically significant. An F-test reveals that the predicted value of the wage for SC/ST candidates is not significantly different (p = 0.42) from that of general candidates at a first-year GPA of 2, the mean level of GPA for SC/ST candidates. This is seen clearly in Figure 4 (p 49).However, in the same F-test done at a first-year GPA of 2.4, the upper end of the range of SC/STGPAs reveals that the predicted wage forSC/STs is 33 per cent lower than that of the general candidates (p = 0.000). This suggests thatSC/ST candidates are not able to get the same reward as general category stu-dents for higherGPAs. We checked whether this is only becauseSC/ST students have low GPAs (with a maxi-mum of 2.4), by regressing pay against GPA and other variables for general category students within the same range ofGPAs. We find a positive and significant slope coefficient of 0.68. Thus, it is not just the low GPAs of SC/STstudents that accounts for their wages not responding to their GPA.It is worth remarking that the standard deviation in wages between companies is about 2.5 times the standard deviation within companies. Not surprisingly, then, the strong effect of first-year GPA on the wage is mostly due to students with higherGPAs finding employ-ment at higher-paying companies, and not due to their being paid more in the same company. This can be seen from regressions of log pay on the same set of variables, done with company means on the one hand, and with company fixed effects on the other. In the former case, the effect ofGPA is strong (and, as expected, be-ingSC/ST has a strong negative effect ifGPA is omitted from the set of controls). But the latter, within-company, regressions give Table 3: Determinants of Grade Point AveragesCoefficient First-YearGPACommunicationQuantitative GPAGPA (1) (2) (3)SC/ST -0.284***-0.135***-0.404*** (0.038) (0.031) (0.060)Work experience (years) 0.00326 0.00888 -0.00240 (0.0072) (0.0057)(0.011)Attractiveness rating 0.0316** 0.0437*** 0.0448* (0.015) (0.012) (0.023)Female -0.0746*-0.00132-0.166*** (0.039) (0.031) (0.062)Elite college graduate 0.117*** 0.0562** 0.215*** (0.031) (0.024) (0.048)College score (scale 0-4) 0.319*** 0.119*** 0.463*** (0.040) (0.032) (0.063)Constant 1.372***1.434***1.155*** (0.13)(0.10) (0.20)Observations 242242242R-squared 0.530.300.51 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1Table 4A: Determinants of Pay (Exchange Rate: Rs 20/$)Coefficient Log (pay) Log (pay) Log (pay) Log (pay) Log (pay) DomesticJobs (1) (2) (3) (4) (5)SC/ST -0.351***-0.194***-0.08551.411***1.406*** (0.072) (0.043) (0.057) (0.30) (0.31)Work experience (years) 0.0392*** 0.0509*** 0.0409*** 0.0422*** 0.0391*** (0.011) (0.0073)(0.0082) (0.0074)(0.0077)Attractiveness rating 0.0857** 0.0205 0.0774* 0.0718* 0.0623* (0.040) (0.017) (0.039) (0.037) (0.035)First-year GPA 0.606*** 0.738*** 0.596*** (0.092) (0.10) (0.12)SC/ST*GPA -0.730***-0.724*** (0.15)(0.15)Female 0.05070.0341 (0.061) (0.062)Communication GPA 0.303*** (0.11)Second-year GPA 0.0575 (0.052)Constant 13.98***13.77***12.51***12.18***11.82*** (0.081) (0.043) (0.19)(0.22) (0.24)Observations 226160226226226R-squared 0.240.370.380.410.42 Ordinary Least Squares, Robust standard errors in parentheses, clustered bycompany *** p<0.01, ** p<0.05, * p<0.1Table 4B: Determinants of Pay (Exchange rate: Rs 14/$)Coefficient Log (pay) Log (pay) Log (pay) Log (pay) (1) (2) (3) (4)SC/ST -0.243***-0.05420.948***0.920*** (0.051) (0.043) (0.24)(0.24)Work experience (years) 0.0483*** 0.0495*** 0.0498*** 0.0475*** (0.0083)(0.0064)(0.0060) (0.0064)Attractiveness rating 0.0513* 0.0454* 0.0440* 0.0380 (0.027) (0.027) (0.025) (0.024)First-year GPA 0.432*** 0.515*** 0.415*** (0.060) (0.066) (0.071)SC/ST*GPA -0.490***-0.475*** (0.12) (0.12)Female 0.0101-0.00414 (0.041) (0.041)Communication GPA 0.183* (0.093)Second-year GPA 0.0631 (0.047)Constant 13.83***12.79***12.58***12.32*** (0.056) (0.13)(0.15)(0.19)Observations 226226226226R-squared 0.260.380.400.41 Ordinary Least Squares, robust standard errors, clustered by company, inparentheses *** p<0.01, ** p<0.05, * p<0.1
