Cover photo: Job-seekers queue at the Allahabad Employment Exchange...

68

Transcript of Cover photo: Job-seekers queue at the Allahabad Employment Exchange...

Page 1: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue
Page 2: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office. 2.3 million

people applied for 368 government jobs advertised in the state of Uttar Pradesh.

Source: New York Post

Page 3: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

My sincere thanks to:

Lant Pritchett, for being an invaluable thought partner.

Your ideas have had a deep impact on my thinking, beyond the present work.

Michael Walton, for mentorship throughout the research process.

Your feedback pushed me to make my work focused and practical.

Varun Aggarwal, for helping me understand recruiting trends in India.

Your insights grounded my work in some semblance of the real world.

Page 4: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

ii

Page 5: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

iii

1. EXECUTIVE SUMMARY

Despite two decades of economic growth, more than 70 percent of urban Indian workers don’t

have formal sector jobs. This vast majority is excluded from India’s remarkable progress: their

wages are twenty times lower, they don’t enjoy worker protections, and they can’t access formal

credit. Failing to put the next generation of Indians in formal sector jobs will turn India’s

anticipated demographic dividend into a demographic disaster.

There is no magic potion that can cure India’s ailing formal sector. Regulatory reforms and

investment in infrastructure are necessary, but will take time. The most effective medium-term

strategy to increase formal employment is to increase the skill level of India’s labor force.

Employers are scrambling to find skilled candidates to fill formal sector jobs, paying a large and

increasing premium for the right talent. Students, in response, are rushing to get degrees to improve

their likelihood of getting jobs, fueling a dramatic boom in higher education. A staggering 20,000

colleges have opened in the last decade.

The quality of many of these colleges, however, is notoriously low. Low quality colleges survive

by hiding in the crowd of an expanding, fragmented industry, in which employers and students are

still learning about college reputations. The proliferation of low quality colleges is leading

employers to question the value of degrees as signals of skill. Facing employer skepticism, job

seekers are finding it harder to demonstrate their ability and earn higher wages. The tertiary wage

premium we observe in the labor market belies the even higher premium employers are willing to

pay for quality tertiary graduates.

An intervention that generates reliable information about the quality of graduates can help firms

hire better workers, reward skilled workers with higher wages, and reveal the quality of colleges

to both employers and students. We propose implementing sector specific Job Readiness Tests for

job seekers that (1) assess their employability; (2) are voluntary initially, with a longer term view

to universal coverage; (3) are championed by a group of employers that graduates aspire to work

for; and (4) use independent assessment agencies for implementation.

Job Readiness Tests will be implemented by sector level Job Readiness Councils (JRCs), led by

employers motivated by self-interest in recruiting better talent. Importantly, JRCs must remain

firmly independent from government to ensure industry trust of the tests.

Page 6: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

iv

Putting India to Work

Resolving Information Failures to Fill the Skill Gap

March 14, 2016

1. Executive Summary ................................................................................................................... iii

2. Motivation: A Good Job for Every Indian .................................................................................. 1

3. Defining the Problem: Why So Few Good Jobs? ...................................................................... 2

4. Diagnosing the Problem: Why So Few Skills? ........................................................................ 13

5. Designing a Solution: Solving Information Failures in the Labor Market ............................ 25

6. Implementing: An Industry Led, Government Supported Intervention ................................ 35

Appendix 1. Admissions Sorting and Wage Premia ................................................................... 42

Appendix 2. Assumptions for Industry Cost-Benefit Analysis .................................................... 56

Appendix 3. Bibliography ............................................................................................................. 57

Page 7: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

1

2. MOTIVATION: A GOOD JOB FOR EVERY INDIAN

2.1. Growth without “good” jobs

India’s has undergone remarkable economic transformation over the past two decades. Per capita

income grew at a decadal rate of 6.1 percent in the 2000s, almost twice as quickly as any previous

decade,1 and pundits predict India will soon be the fastest growing major economy in the world.2

But, as many Indians know only too well, aggregate economic growth does not necessarily imply

shared prosperity. Excluding farmers, 72 percent of Indians worked in the informal sector in 2011-

12. This means they work in small, unincorporated businesses owned by households. 58 percent

informal workers are self-employed, and 70 percent work in an enterprise with 5 or less workers.

They work without contractual agreements (in many cases for family or social relations) and

outside the purview of government regulations. 75 percent of urban informal workers work in

manufacturing, construction, wholesale and retail trade, and transportation and storage.3

As India’s National Economic Survey put it: these jobs are not, for the most part, “good jobs.”

Wages in the formal sector are, on average, 20 times higher than the informal sector. Formal sector

employees build formal work histories, which gives them access to formal credit. They also enjoy

worker protections enacted by the government regulations.4 Informality also has aggregate

efficiency costs. Estimates suggest formal enterprises (excluding the public sector) are more than

9 times more productive than non-agricultural informal enterprises.5

India’s growth, remarkable as it has been, is simply not producing enough “good jobs” for its

citizens. Of the 10.5 million new manufacturing jobs created between 1989 and 2010, 65 percent

have been in the informal sector. A staggering 98.8 percent of the new manufacturing

establishments created in this period were informal. 6 Of those formal sector jobs that do exist,

1 Kar et. Al., “The Dynamics of Economic Growth”, Effective States and Inclusive Development Research, 2010 2 Growth predictions from Atlas of Economic Complexity, Harvard Center for International Development,

http://atlas.cid.harvard.edu/rankings/growth-predictions/ 3 Statistics from 68th round of the National Sample Survey, 2011-12, as reported in “Informal Sector and Conditions of

Employment in India”, Ministry of Statistics and Program Implementation, 2014. 4 Chapter 10, National Economic Survey 2015-16, Ministry of Finance, 2016 5 2003 estimates; Dougherty et. Al, “What is Holding Back Productivity Growth in India?”, OECD Journal of Economic

Studies, 2009 6 Chapter 10, National Economic Survey 2015-16, Ministry of Finance, 2016

Page 8: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

2

over 60 percent are in the public sector.7 In this report, we explore why the odds are stacked against

companies that create “good jobs” and drive India’s productivity growth. This failure of the formal

sector limits India’s growth potential and excludes millions of Indians from sharing in the

prosperity generated by India’s growth.

2.2. A demographic window

Failure to create good jobs could prove particularly disastrous for India as it undergoes a

demographic boom of unprecedented scale. Over the next ten years, more than 250 million young

Indians will come of working age for the first time8 - a figure comparable in magnitude to the

entire working age population of Europe. By 2025, two-thirds of all Indians will be of working

age, and almost one in five workers around the world will be Indian. 9

As urbanization gathers speed, this demographic boom will be particularly concentrated in India’s

cities. Over the next ten years, more than 100 million people will migrate from rural to urban areas,

and the urban share of India’s population will grow from 32% to 37%. By 2050, almost 400 million

additional Indians will live in cities, accounting for 50% of the total population.10

The window in which this next generation of Indians enter the urban labor force is a critical

opportunity. Failing to put citizens in good jobs could consign them to a life of under-employment

and exclusion, and squander the economic windfall of a one-off demographic dividend. Acting

now could put India on a trajectory to inclusive prosperity.

Caught between a history of few formal sector jobs and a future of many workers, this report

proposes a strategy to create “good jobs” for the next generation of Indians.

3. DEFINING THE PROBLEM: WHY SO FEW GOOD JOBS?

Theories put forward to explain the lack of formal sector jobs fall into three broad categories: lack

of adequate infrastructure to enable private investment, regulatory distortions in factor markets

7 Ministry of Statistics and Program Implementation, accessed at https://data.gov.in/catalog/employment-organised-

sectors-public-and-private 8 United Nations World Population Prospects: The 2015 Revision, United Nations, 2015 9 Ibid. 10 United Nations Urbanization Prospects: The 2014 Revision, United Nations, 2014

Page 9: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

3

(particularly labor), and inadequate supply of human capital to fill the skill-intensive jobs being

created in our economy. 11 We consider each of these theories in turn.

3.1. Infrastructure: not the full story

A popular explanation offered is that low quality infrastructure in urban India has contributed to

the failure of labor intensive industries such as manufacturing. Confronted with poor transport

infrastructure in particular, Indian firms specialize in industries that don’t need to physically

transport goods. IT and IT enabled service companies such as Infosys, Wipro, HCL Technologies

and Tata Consultancy Services have flourished, beaming their inputs and outputs over the internet

and bypassing infrastructural constraints. These service firms hire mid-and-high-skill workers, but

do little to create jobs for India’s abundant low-skill labor. Low skill workers, unable to find jobs

in traditionally labor intensive manufacturing industry, are forced into less productive, small scale

commercial activities in the informal sector to make ends meet.

There is at least some truth in this line of argument. We look at India’s road as an example. Despite

having the second largest network in the world – denser per square kilometer than the US, China

and Mexico – most of India’s roads (59 percent) are small and rural, 12 and less than half are paved.

13 0.6 percent of roads are multilane highways, compared to 3.5 percent in China and 12.9 percent

in Mexico.14 Unsurprisingly, road congestion has been a perennial concern for businesses and

residents in India’s major hubs.

India, however, performs better than countries at a similar level of development on aggregate

infrastructure indices published by the World Economic Forum (WEF) and the World Bank

(Exhibits 1&2). These (and other) measures suggest that, while undoubtedly bad, infrastructure

might not be a binding constraint to private investment in India.

Importantly, infrastructure can only explain some of the facts we observe in the Indian economy.

It can plausibly explain why skill-intensive service industries (particularly IT) have prospered, but

11 Chapter 7, National Economic Survey 2014-15, Ministry of Finance, 2015 12 AT Kearney, “Trends in Indian Infrastructure Development”, 2013, accessed at http://www.i-

cema.in/pdf/trends_india_infra_dev_page.pdf 13 Bureau of Transportation Statistics; RITA, US Department of Transportation, accessed through Statista 14 AT Kearney, Trends in Indian Infrastructure Development, 2013

Page 10: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

4

it cannot explain the peculiar patterns dynamics in firm size and growth. We explore these

unexplained facts further in the next section.

Exhibits 1&2: conditioning for income level, India performs well on World Bank and World Economic Forum

aggregate infrastructure indices

Source: World Economic Forum, World Development Indicators

3.2. Regulatory distortions: second-best solutions deployed

A second line of argument contends that business regulations, many of which are as old as India

itself, create perverse incentives that prevent firms from formalizing or gaining scale. In attempts

to protect small firms from competition, a broad range of policies have been implemented:

subsidized credit for small firms, tax exemptions, preferential government procurement, cheaper

electricity, and small-scale reservations, which prevent larger firms from producing certain

products altogether.15

To illustrate these perverse incentives, we look at the effect of labor regulations in particular. The

Factories Act stipulates rules for work hours and conditions that must be followed in

manufacturing establishments that employ more than 10 people. The Industrial Disputes Act

requires firms with more than 100 employees to obtain government permission before firing a

worker, and firms with more than 50 employees to obtain worker consent before modifying job

descriptions or moving workers between plants.

15 Martin et. Al, “In with the Big, Out with the Small: Removing Small-Scale Reservations in India”, RAND corporation

working paper, February 2014

Page 11: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

5

These laws can only be enforced on formal sector firms above a threshold size of employment,

leading to distortions in the number and size of formal sector firms. In India, the average 40-year-

old firm hires 40 percent more workers than a 5-year-old firm. In the US, it hires 7 times as many

workers (Exhibit 3). Research also suggests a “kink” in the distribution of manufacturing firm size

at 10 workers – the threshold at which the Factories Act kicks in – suggesting firms are indeed

altering behavior to avoid regulations (Exhibit 4).16

Exhibit 3: Indian firms grow slowly

Source: Hsieh and Klenow (2014)

These distortions in firm dynamics misallocate resources in the economy, channeling them away

from larger, formal, more productive firm to small, informal sector businesses. This can have a

real negative effect on the economy. Studies suggest that states that amended labor laws in pro-

worker direction (and so exacerbated perverse firm incentives) experience lower output,

employment, investment and productivity in formal sector manufacturing.17

Although there’s near consensus among academics that India’s labor laws need revision, reform

efforts are politically fraught. The current government proposed modest amendments to the

Factories Act, Contract Labor Act and Industrial Disputes Act, raising thresholds at which these

acts kicked in. Even this limited reform led to a 150 million worker strike in September 2015.18

Firms, realizing that these laws may be here to stay (at least in the mid-term), have evolved to cope

with regulatory distortions. In some instances, they simply don’t comply,19 though this opens up

avenues for “harassment corruption” by government inspectors.20

16 Amriapu, A. & Gechter, M. “Indian Labor Regulations and the Cost of Corruption: Evidence from the Firm Size

Distribution,” November 2014 17 Besley, T and Burgess, R. “Can Labor Regulation Hinder Economic Performance? Evidence from India”, 2004 18 “Indian worker strike over Modi labour reforms”, BBC news, 2 September 2015 19 Bhattacharjea, A. “Labour Market Regulation and Industrial Performance in India: A Critical Review of the Empirical

Evidence”, Centre for Development Economics, DSE, Working Paper No. 141, June 2006 20 Amriapu, A. & Gechter, M. “Indian Labor Regulations and the Cost of Corruption: Evidence from the Firm Size

Distribution,” November 2014

Page 12: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

6

Exhibit 4: kink in the firm size distribution at 10 employees (threshold for the Factories Act)

Source: Amirapu and Gechter (2014)

In many cases, firms use contract workers through third party agencies in lieu of permanent

workers. The proportion of contractors in the formal manufacturing sector increased from 12

percent in 1999 to 25 percent in 2010.21 Contractors don’t show as permanent employees of the

company, allowing firms to stay small enough on paper to circumvent labor laws and wash their

hands of the need to deal with government inspectors.22 The contract share of employment is higher

in states with rigid labor laws, suggesting that firms are actively using this strategy. 23

Even firms that are large enough to be subject to labor laws increasingly use contract workers to

avoid making permanent hires who might become impossible to fire in the future (due to either

demand shocks or performance issues). Indeed, the contractor share of employment is higher

among plants with more than 100 workers, to whom the Industrial Disputes Act that makes firing

difficult already applies.24

The increasing use of contract workers mitigates the worst effects of regulatory distortions.

Without touching labor laws, firms are effectively limiting their scope by reducing the share of the

labor force they apply to. By reducing the cost of hiring in the formal sector, the use of contractors

has led to an increase in the formal employment in manufacturing, although it remains at one third

of informal levels (Exhibit 5). The re-allocation of resources to more productive, formal sector

21 Chapter 10, National Economic Survey 2015-16, Ministry of Finance, 2016 22 Ibid. 23 Ibid. 24 Ibid.

Page 13: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

7

firms also has broad economic benefits. Research suggests that increased use of contract labor

boosted manufacturing GDP by 0.5 percent annually between 1998-99 and 2011-12.25

Exhibit 5: formal sector employment has

picked up since 2000

Source: Ghani et. Al (2015)

Though they have taken the edge of labor regulations, contract workers remain a second-best

solution for firms. Recent empirical work suggests that firms that hire more contract workers are

less productive. Contract workers stay with employers for shorter periods of time, reducing a

firm’s incentive to invest in firm-specific human capital. Though contract workers are sometimes

more motivated (eager to earn a permanent position), the net effect of a higher contract share on

firm productivity is negative. 26 The use of contract workers introduces middlemen into the hiring

process, increasing hiring costs. The Indian Cellular Association estimates hiring a contract worker

is 14 percent more expensive than hiring a regular worker.27 Finally, firing permanent employees

remains very difficult, reducing incentive for all the but the most motivated to work.

In sum, while regulatory distortions remain an issue, industry workarounds are diluting the worst

of their effect.

25 Bertrand et. Al. (2015) as cited in Chapter 10, National Economic Survey 2015-16, Ministry of Finance, 2016 26 Soundararajan, V. “Contract Work and Endogenous Firm Productivity in the Indian Manufacturing Sector”, 2015 27 Chapter 10, National Economic Survey 2015-16, Ministry of Finance, 2016

Page 14: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

8

3.3. Human capital: jobs, but no skills

A third line of argument contends that low levels of human capital28 in the labor force are

preventing the creation of more formal sector jobs. Companies simply can’t find the skills they

need to make investments viable (at acceptable cost). Two facts lie at the heart of this argument.

First, despite an abundance of cheap low skill labor, India’s growth has been driven by skill29

intensive industries. Skill intensive industries, measured as those with higher years of schooling

among employees, account for a large and growing share of output and employment (Exhibit 6).30

This proportion has historically been high when compared to countries at similar levels of

development, such as China and Indonesia (Exhibit 7).

We note this is consistent with our discussion in chapter 2. Regulatory distortions that make hiring

expensive may be causing companies to substitute towards higher skill (higher value add) workers.

Poor infrastructure may similarly be causing a shift towards high skill service sectors.

Exhibit 6: Skill intensive sectors dominate

the economy

Source: Amirapu and Subramanian (2015)

Second, India’s labor force has particularly low levels of skill. 29 percent of the labor force is

illiterate.31 More than 40 percent of the labor force has less than primary education – much higher

28 By “human capital”, we mean the skills, knowledge and attributes that make individuals productive in the workforce.

Given human capital is difficult to measure, we metrics such as years of schooling, level of educational attainment and test

scores as proxies. We also use the term “skills” interchangeably with human capital throughout the paper. 29 For expositional convenience, we use the terms “skill” and “human capital” interchangeably for the remainder of this

paper. 30 Amriapu, A. & Subramanian, A. “Manufacturing or Services? An Indian Illustration of a Development Dilemma,” Center for

Global Development Working paper, June 2015. 31 Mehrotra et. Al, “Estimating the Skill Gap on a Realistic Basis for 2022”, IAMR Occasional Paper No. 1, 2013

Page 15: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

9

than what we would expect for a country at India’s level of development (Exhibit 8). In 2010, only

28 percent of India’s labor force had secondary or tertiary education.32

Exhibit 7: Skill intensity has been high

compared to similar countries (1981-1996)

Source: Kochhar et. Al (2006)

Exhibit 8: India’s labor force is particularly

under-educated by global standard

Source: World Development Indicators

These numbers illustrate the low levels of educational attainment and say nothing of what many

pundits cite as the real problem: quality of education. Even India’s best states perform poorly in

international standardized school level assessments (Exhibit 9). At home, Indian employers

complain loudly and frequently about the quality of college graduates. One famous (but by no

means only) example is a 2005 study by NASSCOM (the Indian IT trade association) and

McKinsey, which found that “only 25 per cent of engineering graduates in India have the skills to

be employed in IT jobs without prior training.”33

32 World Development Indicators 33 NASSCOM & McKinsey, “Extending India’s Leadership of the Global IT and BPO Industries”, 2005

Page 16: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

10

Exhibit 9: Poor performance on standardized

PISA test

Source: Asian Development Bank, Key Indicators for

Asia and the Pacific (2015)

These facts suggest that (a) Indian employers have a high demand for skilled labor, and (b) skill is

in short supply in the Indian labor force. Given an excess demand for skills in the economy, tertiary

graduates have a much higher likelihood of being employed (Exhibit 10). They also enjoy a large

and rising wage premium (Exhibit 11). This premium is increasing more quickly in states where

industry productivity is growing faster (Exhibit 12). The co-movement of the skill premium and

growth suggest that skills may be a binding constraint to private sector growth.

Despite a rising skill premium, employers struggle to find qualified candidates. A McKinsey

survey of 2,832 employers in nine countries found that the average large employer in India had 36

entry level vacancies – more than any other country surveyed. When asked why, 53 percent of

Indian employers cited lack of skills as a common reason for entry-level vacancies – second after

Turkey and well above the group average of 39 percent. 82 percent of Indian employers said they

would be willing to pay more for graduates with the right skills – second only to the United States,

and again well above the group average of 70 percent.34

34 McKinsey & Company, “Education to Employment: Designing a System that Works”, 2012

Page 17: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

11

Exhibit 10: share of employment by Exhibit 11: Rising tertiary skill premium

education level

Source: Education, Skill Development and Labour Source: Azam (2009)

Force, Ministry of Labour & Employment (2014)

Exhibit 12: Skill premium is

correlated with manufacturing

productivity growth across states.

faster growing states exhaust a

limited pool of skilled workers, and

skill premium rise faster

Source: Amirapu and Subramanian (2015)

McKinsey’s survey is not the only one finding loud complaints of low skill job-seekers among

Indian employers. A Manpower Group survey of 41,700 employers in 42 countries found that 58

percent of Indian employers had difficulty filling jobs, seventh highest in the world. By way of

48.7

53.958.6

54.2

43.440.1

44.2

54.6

64

0.34

0.5

0.3

0.37

0.130.15

0.27

0.3

0

0.1

0.2

0.3

0.4

0.5

0.6

1983 1987 1993 1999 2004

Tertiary-Secondary gap Secondary-Middle gap

Middle-Primary gap Primary-Below Primary gap

Page 18: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

12

comparison, the highest complaints were in Japan, a nation in a very different demographic

situation to the booming Indian working age population.35

Employers’ complaints are borne out in the data. A study by the Asian Development Bank found

that, among Asian countries, India had the second-lowest proportion of tertiary workers in high-

skilled occupations (Exhibit 13).36 Employers are struggling to fill these high-skill positions.

Exhibit 13: education share in high-skilled

occupations. Few tertiary workers in high

skill jobs.

Source: Asian Development Bank (2015) from country

labor force surveys

PRC = People’s Republic of China

3.4. Homing in on skills

As our discussion to this point hopefully illustrates, no single factor can be isolated as the cause

low levels of formal sector employment is a difficult. A range of policy issues that stack the odds

against formal sector, job-creating companies need to be addressed. In subsequent chapters, we

focus our discussion on skills for three reasons:

1. Action is politically supportable and feasible in the short term. Labor regulation reform in

India has been mooted for decades without a shift in its political likelihood. Infrastructure

investments are subject to lengthy time frames for approval and construction, and are already

being aggressively pursued by the government.37 Human capital development efforts, on the

other hand, are riding on a wave of political support. The government has made major

investments in driving primary school enrolment and is now increasingly being lobbied to

35 Manpower Group, “Annual Talent Shortage Survey”, 2015 36 Asian Development Bank, “Key Indicators for Asia and the Pacific”, 2015 37 See, for example, reports of the Prime Minister’s Office directly driving infrastructure project: “PMO steps in to push

stalled infrastructure projects”, The Times of India, Dec 30, 2014

Page 19: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

13

focus on higher education and education quality. Both the previous Congress and the current

BJP government announced a major emphasis on a “Skill India” campaign, aimed at increasing

the level of vocational skills in the labor force. There is a political hunger for action in this

space, and interventions may have a shorter pay back than longer term infrastructure

development projects.

2. Industry “work-arounds” for low skill levels may prove difficult. Industry responded to limited

infrastructure by focusing on infrastructure-light service industries. They responded to

regulatory distortions through use of contract workers. How will industry respond to an

exhausted pool of skilled workers? Despite their best efforts, successive governments have

failed to catalyze the take-off of low-skill-labor intensive manufacturing sectors. Given the

fastest-growing sectors in the economy remain skill intensive, companies will struggle to find

a substitute for human capital.

3. Interventions can serve a dual purpose of growth and inclusion. Helping job-seekers acquire

skills helps business find and hire more productive workers. It also empowers job-seekers to

participate in a growing economy and reap its dividends through higher incomes. In this way,

skills are both a pre-requisite for growth and a guarantee that citizens are included in India’s

rising prosperity.

For these reasons, we home in on skills and explore issues blocking (and solutions that might

enable) the development of human capital in India.

4. DIAGNOSING THE PROBLEM: WHY SO FEW SKILLS?

We noted in chapter three that skill levels are low in the Indian workforce, leading to a skill

shortage among Indian employers. In this chapter, we go one level deeper and ask: why? Given

demand for skilled labor among employers, why aren’t job-seekers acquiring skills?

4.1. Training sector response

In some cases, they are. Individuals are responding to the growing tertiary skill premium in

particular and seeking to acquire skills. College enrolment increase 2.4 times in the 11 years from

2000-01 to 2011-12. While Arts courses dominate (36 percent of total enrolments), enrolment

Page 20: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

14

growth has been fastest in engineering (11.8 percent compound annual growth from 1985 to 2010),

followed by education (8.0 percent) and science (6.2 percent) (Exhibit 14).

This has created extraordinary demand for higher education institutions (HEIs). As of 2011-12,

India had 36,113 HEIs – the highest number in the world, and more than 5 times as many as the

US and 7 times as many as China.38 More colleges opened in the six-year period 2005-06 to 2011-

12 than had opened in the thirty years prior (16,444 new colleges vs. 15,238).39

Almost two thirds of these are private, 40 suggesting the private sector has stepped in to plug the

public shortfall in training capacity. Private sector response has been particularly strong in

professional disciplines such as physiotherapy, pharmacy and hotel management (Exhibit 15).

Training supply is responding to skilled labor demand, though there is a long way to go. Gross

tertiary enrolment stood at 24 percent in 2012, up from 10 percent in 2002. It is now higher than

what we would expect from a country at India’s income level.41

Exhibit 14: enrolment growth by course of study

Source: Ministry of Human Resource Development

38 Ernst & Young and FICCI, Private sector participation in Indian higher education, FICCI Higher Education Summit, 2011 39 Ministry of Human Resource Development statistics 40 Ernst & Young and FICCI, Private sector participation in Indian higher education, FICCI Higher Education Summit, 2011 41 World Development Indicators

Page 21: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

15

Exhibit 15: private sector

participation in

professional higher

education courses

Source: Ernst & Young and

FICCI, Private sector

participation in Indian higher

education, FICCI Higher

Education Summit, 2011

While the quantity of higher education supply has grown impressively, the quality of colleges and

college graduates is notoriously bad. According to India’s National Assessment and Accreditation

Council (NAAC), 89 percent of India’s colleges are rated either average or below average.42

According to recent reports of graduate employability, 34 to 53 percent of graduates are considered

employable.43 The top rated Indian institution in university rankings by Times Higher Education,

US News and QS rankings came in at ranks 251-300, 354 and 147 respectively. Training

companies are churning out more and more graduates, but many of them are simply not

employable.

4.2. Explaining low training quality

We argue that the extremely low quality of higher education institutions can be explained by

information failures in the labor and training markets

Rapid expansion – over 20,000 new colleges have opened in over the past decade, an average of

2,000 each year – has allowed low quality colleges to hide in the crowd. While employers are still

in the process of learning the quality of colleges, graduates of high and low quality college receive

similar wage premiums for their educational investment. Similar labor market outcomes upon

42 As reported in “Indian universities second-grade?”, The Times of India, Feb 21, 2014 43 Pushkar P, “Why India’s universities can’t keep up with the masses”, The Conversation, April 20, 2014

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Page 22: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

16

graduation slow down student learning about college quality as well, and prevent admission sorting

by ability into high and low quality institutions. Limited sorting in turn further slows down

employer learning about college quality.

This dual information failure – between employers and colleges and students and colleges – leads

to a vicious cycle that prevents, or at least slows, learning and reputation building about the quality

of training institutions, allowing low quality colleges to survive for longer than they otherwise

would.

There are exceptions to this trend. At the very top end of the training market, an elite group of

institutions have built formidable reputations for quality. In a marketplace where information on

quality higher education is scarce, graduates from these elite institutions enjoy super-normal

returns. According to the Financial Times MBA rankings, a graduate from the prestigious Indian

Institute of Management Ahmedabad (IIM-A) makes $167,676 on average three years after

graduation (PPP adjusted). This is the fifth highest salary in the world (after HBS, Stanford,

Wharton and Columbia), and the only non-OECD school in the top 10.44

Excluding this elite, however, the prospect for Indian graduates seeking jobs are relatively bleak.

In 2005, one in three unemployed in India had tertiary degrees – the seventh highest proportion in

the world. For this large group of graduates, having a degree is a necessary but by no means

sufficient condition to securing a well-paying job.

To understand the role of information in educations and labor markets more thoroughly, we test

the extent to which better quality colleges are able to (a) attract better students in the admissions

process, and (b) secure better labor market outcomes of its graduates.

4.3. Can students recognize good colleges?

We study education and employment outcomes for a sample of 1,959 engineering college

graduates, diving deeper into a sector that has experienced the most drastic boom in training supply

in recent years.45

44 Financial Times MBA rankings 2015, accessed at http://rankings.ft.com/businessschoolrankings/global-mba-ranking-

2015 45 Details of our study are included in Appendix 1. We apply the methodology laid out by Macleod et. Al (2015) on a dataset

provided by an Indian assessment agency – the Aspiring Minds Employment Outcomes 2015 database.

Page 23: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

17

If students have information about college quality, we would expect there to be fiercer competition

to get into better colleges. Students with better high school marks would win sought after seats,

leading to a sorting in the admissions process of better students to better colleges.

It’s important to note that a range of factors may prevent perfect sorting by ability. Students may

choose not to go better quality colleges because cost or convenience (distance) considerations.

Colleges may also choose not admit the best students. A prime example of this are India’s policies

of caste-based affirmative action. 49.5 percent of seats in government educational institutions are

reserved for students who belong to Scheduled Castes, Scheduled Tribes or Other Backward

Castes (these groups account for 20 percent, 9 percent and 40 percent of households in India).46

These reservations do not apply to private institutions, however, which comprise 91 percent of

engineering colleges.47

We test admissions sorting among the 2014 graduating cohort in our sample. Measuring the

candidate is relatively straightforward: we use the year 12 scores of students (scaled for their board

of education), which form a basis of the application students would put forward to colleges.

Measuring college quality is more difficult, given its multi-dimensional nature. We argue that,

from a labor market perspective, the best measure of college quality are the wages of its graduates.

In a world of perfect information, 2014 graduates would have had access to the post-graduation

salaries of 2011 graduates (three years senior). Better students would choose to go to colleges

whose graduates have higher wages upon graduation. The average wage of 2011 graduates,

therefore, serves as a measure of college quality.

In Exhibit 16, we compare the quality of each candidate with the quality of the college he/she

attended. If there is no sorting by ability, we would expect to see a horizontal line implying there

is no relationship between candidate and college quality. If there is some sorting, we would expect

to see a positively sloped line. The higher the level of sorting, the steeper we would expect the line

to be. The green line in Exhibit 16 shows the sorting that actually occurs. Admissions sorting is

limited and inconsistent. Better students do go to better colleges, but the relationship is slight. 2014

graduates with year 12 scores in the top ten percentile of our sample went to colleges with an

average 2011 graduate salary of Rs. 341,833. Students with scores 50 percentile ranks lower in our

46 National Sample Survey 61st Round, “Employment and Unemployment Situation Among Social Groups in India 2004-05”,

2006 47 Ernst & Young and FICCI, “Private sector participation in Indian higher education”, 2011

Page 24: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

18

sample, in the 40 to 50 percentile range, went to colleges with an average salary of Rs. 321,551 –

a difference of just Rs. 20,000, less than 5 percent of the difference in average salary between the

best and worst college.

Exhibit 16: Limited sorting of students

by ability at time of admission

Source: AMEO database (2015)

Exhibit 17: Significant intra-college

ability variation

Source: AMEO database (2015)

In Exhibit 17, we rank colleges by quality on the x-axis. We then look at the quality of the median

candidate enrolled at each college (blue dot), as well as the range of ability from the 25th to the

75th percentile college student enrolled at each college (red line range). As we can see, while

admissions sorting across colleges does occur, intra-college candidate quality variation is in many

cases as large as variation across colleges. There are students in the worst quality college in our

sample with year 12 scores higher than the median student in the best quality college in our sample.

Page 25: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

19

While we can’t rule out other factors (cost, convenience, quotas), our findings support the

existence of information failures between students and colleges. Students may not know which

colleges are better for their labor market outcomes and may be making enrollment decisions based

on misinformation. Literature also supports the view that new institutions market themselves based

on more visible features, such as campus facilities, resources and the use of technology.48 The US

Department of Education notes that the US News university ranking “weights spending and school

resources as nearly thirty percent of the evaluation, scored six times greater than how students fare

after their education experience.”49

4.3. Can employers recognize good colleges?

Next, we consider whether employers can recognize good colleges and are willing to pay a wage

premium for their graduates. Given admissions sorting in India is substantially incomplete, we

might expect the conditional return to college return to be significantly smaller than it otherwise

would. In a world with better sorting, better colleges would have better graduates not only because

of better standards of learning, but also because of their ability to screen candidates at the time of

admission. To the extent this screening is imperfect, the premium a degree from a better quality

college is diluted.

Returns to college reputation, controlling for the student’s own ability (as measured by scores in

high school, college and on a third party test), are small but statistically significant.50 Attending a

college with a student body scored ten percentage points higher in their year 12 exams (controlling

for a student’s own score) is associated with 3 percent higher wages. By comparison, the same

analysis completed in Colombia concluded that attending a better college was associated with an

8 percent higher wages. That is, the wage premium for college reputation is 2.7 times higher in

Colombia than it is in India.51

This provides further evidence consistent with our argument that information failures are at play

in the training and labor markets. Students and employers are not able to identify the best colleges,

48 Ivy, J. “Higher education institution image: a correspondence analysis approach.” The International Journal of

Educational Management. 15/6, 276-282, 2001 49 U.S. Department of Education, “Better Information for Better College Choice & Institutional Performance”, 2015 50 Details are provided in Appendix 1 51 Macleod et. Al. “The Big Sort: College Reputation and Labor Market Outcomes”, May 2015

Page 26: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

20

and low quality institutions are able to hide in the crowd and continue to provide sub-standard

education.

4.4. Industry response to information failures52

Employers have access to three imperfect signals of job-seeker quality on a typical fresh graduate’s

CV: her year 12 score, college GPA, and the name of the college she attended.53

Year 12 scores are typically reported along with the board of education the candidate graduated

from. The Indian education system offers students a choice to complete secondary education under

the auspices of one of 54 state or central boards of education.54 The lack of standardized scores

makes it harder for employer to compare across job-seekers and make judgments of relative ability.

College GPAs are also not trusted. As we have discussed, employers don’t know and don’t trust

the reputation of colleges job-seekers have attended and, by extension, the value of the college

GPA as a signal of skill is also marginal.

In the absence of reliable measures of ability, many employers hire on the basis of connections

and social relations. Sociological research suggests managers tend to hire “recruits who are

socially similar to themselves”55 leading to “homosocial reproduction”56 or “status

groups…monopoli[zing] valued economic opportunities.”57 To quantify this discrimination,

researchers submitted otherwise identical CVs with names associated with different religious or

ethnic groups and measured interview call-up rates. An applicant with a Scheduled Caste Hindu

name had a 67 percent chance of getting an interview as compared to an identical applicant with a

high caste Hindu name. A Muslim name had 33 percent odds.58 Network-based hiring is not

52 I thank Varun Aggarwal, co-Founder of Aspiring Minds, an assessment agency, for sharing his insights on recruitment

trends that informed this section of the report. Errors remain mine. 53 Employers may also value of extracurricular activities 54 Council on Boards of School Education in India board list, accessed at

http://www.unishivaji.ac.in/uploads/admin/circ/board_list.pdf 55 Thorath and Attewell, “The Legacy of Social Exclusion: A Correspondence Study of Job Discrimination in India”, Economic

and Political Weekly, Vol. 42, No. 41 (2007) 56 Kanter, Rosabeth (1977): Men and Women of the Corporation, Basic Books, New York. as cited in Thorath and Attewell

(2007) 57 Weber, Max (1968): Economy and Society (edited by Guenther Roth and Claus Wittich), Bedminister Press, New York, as

cited in Thorath and Attewell (2007) 58 Thorath and Attewell, “The Legacy of Social Exclusion: A Correspondence Study of Job Discrimination in India”, Economic

and Political Weekly, Vol. 42, No. 41 (2007)

Page 27: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

21

meritocratic and a hotbed for discrimination in employment practices. It also does not help

employers find the skill they need to grow.

Unable to find the skills they need, employers have adopted two broad approaches, which we refer

to as “hire and train” and “screen heavily.”

In the “hire and train” approach, employers stop trusting signals altogether and assume they’re

always hiring the median (or worse) worker. They then invest heavily in training new hires. A

mid-size textile company CEO we interviewed summarized: “If I’m hiring a fresh graduate, I

assume he knows nothing. I don’t care what college he comes from. We have our own training

programs that teach everything the guy needs to know to work in our company.” This company is

not alone – 93 percent of employers train new hires for an average of 31 days – the highest

investment among nine surveyed countries (Exhibit 18).59

Exhibit 18: Training of entry level workers:

employers invest significant amounts in new hire

training

1. Does your company provide training for new hires?

2. On average, how many days of training does a new hire

receive in the first year?

Source: McKinsey Education to Employment report

The “hire and train” approach exacerbates the vicious cycle created by information failures.

Employers place little premium on skills acquired in college or training institutions as they are

investing significantly in re-training their hires. Graduates from better colleges receive smaller

premiums, reducing demand for better quality colleges.

59 Ibid.

Page 28: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

22

However, not all industries have responded to the paucity of information in this way. Some

companies take the “screen heavily” approach, investing in more rigorous screening methods to

find and recruit the best candidates. This generally requires generating new data on candidate

quality through employer led interviews or tests. Wipro, a leading IT company, has a three round

selection procedure for fresh graduates. First, an online test of programming, quantitative, logical

and verbal skills. Second, a group discussion on a broad range of social and civic issues. Third,

one-on-one technical and HR interviews (strengths and weaknesses, etc.)60 Based on the accounts

of company employees, Tata Motors follows a similar procedure: shortlisting candidates based on

college scores, administering their own written test, group discussion, technical interview followed

by an HR interview.61

The IT industry is a good example of how the “screen heavily” approach changes recruiting

dynamics within a sector. The largest IT companies invest heavily in screening by conducting on-

campus recruitment at select university campuses. This includes, in many cases, purchasing and

delivering standardized employment tests from assessment agencies to supplement traditional

recruiting methods.

Small or mid-size companies don’t hire directly from campus at all. The investment required is

too great. Instead, they recruit 1-2 year experienced hires from the top ten IT companies, piggy-

backing on their screening processes. The top ten companies “keep the best and train the rest”,

retaining the top graduates and churning other screened and now experienced hires to the rest of

the industry. In some cases, the top ten may even be “training the best and keeping the rest,” as

smaller, high value add companies hire the very best graduates once they’ve proven themselves at

the bigger IT firms.

The skill acquisition outcomes that result from the “screen heavily” approach are very different

from the “hire and train” approach. In the “screen heavily” approach, employers generate new data

that effectively helps solve the information failure between employers and colleges/job-seekers.

Armed with this information, firms are willing to pay a premium for skills, and students have an

incentive to acquire skills that allow them to secure better jobs. Bad colleges whose graduates fail

60 “WIPRO interview process selection procedure”, CampusFold.com, 2014, accessed at

http://www.campusfold.com/2014/08/wipro-interview-process-recruitment.html 61 “What is the placement procedure for Tata Motors”, Quora, 2015, accessed at https://www.quora.com/What-is-the-

placement-procedure-for-Tata-Motors

Page 29: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

23

to pass employer screening efforts will eventually be found out, breaking the vicious cycle created

by a dual information failure.

The IT sector’s approach, while better than nothing, is still sub-optimal. First, it imposes a

screening cost associated with new hires that might reduce the number of people a company is

willing to hire. Second, IT companies that invest in screening and hiring fresh graduates don’t reap

the full benefits of their rewards. As employees are poached by other companies over time, some

of the benefits leak to other employers and, of course, in large part to the employees themselves.

This means firm investment in screening will be lower than the “socially optimal” level.

Nevertheless, many industry commentators consider the IT sector’s approach to skill a success.

Wage premiums have led to graduates rushing to train in technical skills applicable in IT. The

industry’s own analysis of its talent situation concludes that “the talent pool available is expected

to stay in excess of the talent demand for the coming few years”62 – a rare and privileged position

for employers to be in in India’s otherwise skill constrained economy.

We argue that the IT industry’s position is not (just) the result of prescient foresight, better

employer coordination, or luck. Rather, three structural features can help explain which sectors

adopt a “screen heavily” strategy that leads to better talent outcomes.

1. High labor productivity. In human capital intensive industries with high marginal product of

labor, search costs will be relatively small compared to large value gain of attracting better

employees. In these industries, companies would be willing to incur screening costs to find

even slightly better employees, because each employee is worth more to the company. This

helps explain why human capital dependent industries such as IT and professional service firms

invest heavily in candidate sourcing and screening.

2. High variance in job-seeker quality. If there is high variance in graduate quality, employers

would have higher incentive to invest in screening solutions. Let us assume employers are able

implement a screening technology that identifies the top 20 percent of candidates from a

graduating cohort. In industries with high variance in graduate quality, the average screened

candidate will be much better than the average candidate without screening. If the quality of

the graduate cohort has low variance, the average screened candidate will be better, but not

62 NASSCOM, Analysis of Talent Supply and Demand. Employment Requirements and Skill Gaps in the Indian IT-BPM

Industry, 2014

Page 30: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

24

that much better, than the average unscreened candidate (Exhibit 19). Research suggests that

the variance of employee performance increases with skill level of employees. One study found

that a blue collar worker who is one standard deviation above the mean is 20 percent more

productive than the average worker. In white collar jobs (selling insurance), she is 120 percent

more productive. In a management job, she is 600 percent more productive.63

Exhibit 19: higher job-seeker

quality variance increases the

return on screening

3. Known, low cost screening technology. Technologies for screening certain skills and attributes

are more readily available, reliable and cheaper. Testing technical knowledge, for example, is

easier than testing a candidate’s behavioral attributes or ability to learn. IT companies can

reliably and cheaply test a coder’s programming skill. It’s might be more difficult and

expensive for a bank to test a new employee’s sales skills, or for an auto-manufacturer to test

an electrician’s ability to respond to possible maintenance cases. Evidence suggests that “skill

is relatively much more important for professional/technical employees while behavioral traits

are more important for blue-collar workers.”64

We note that each of these three structural factors predict that skill-intensive industries with higher

value-add per worker (like IT) are more likely to invest in screening. Companies in relatively lower

skill but more labor intensive sectors such as automobile manufacturing or construction are

unlikely to follow this socially preferable path.

63 Staffing.org, Staffing Effectiveness & Retention, 2013 64 Osterman, P. “Skill, Training, and Work Organization,” Industrial Relations, Vol. 34, No. 2, 1995

Page 31: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

25

5. DESIGNING A SOLUTION: SOLVING INFORMATION FAILURES IN THE LABOR

MARKET

So far, we have argued that low levels of skill are one key reason for low levels of formal sector

employment in India. Low levels of skill arise due to the low quality of education providers, who

are able to survive (and even thrive) due to information failures in the education and employment

markets. Employers find it hard to know which job-seekers (and which colleges) are good.

Students find it hard to know the quality of colleges at the time of admission. These information

failures form a vicious cycle that reduce incentives for individuals to acquire skills and make it

harder for companies to find the skills they need to fill “good jobs.”

While some industries have responded to this problem by investing in screening technologies that

create more information about job-seeker skill levels, not all industries will solve this problem on

their own. Companies in lower skill but labor intensive industries are less likely to invest in

screening of job-seekers, which would in the long run increase skill levels of graduates and create

more formal sector jobs.

A successful solution must help bridge information failures in the labor and education markets,

and particularly in sectors which need help the most. An intervention that creates a credible signal

of job-seeker quality can increase incentives for skill-acquisition and reduce search costs for filling

“good jobs.”

5.1. Learning from current efforts

The Government has implemented various schemes aimed at improving quality in tertiary

education. The National Assessment and Accreditation Council (NAAC), set up in 1994 as an

autonomous institution under the University Grants Commission (UGC), uses self-evaluations and

peer reviews to assess higher education institutions across seven input and process based criteria

and assigns them a letter grade. Similarly, The All India Council for Technical Education (AICTE),

the accrediting agency for engineering and technical higher education institutions, made

accreditation by the National Board of Accreditation (NBA) compulsory for all institutions under

its purview. The NBA is now an autonomous body, independent of AICTE. The Indian Council

Page 32: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

26

of Agricultural Research also established an Accreditation Board (AB) to perform a similar

function.

The majority of these efforts have focused on assessing college inputs and processes, rather than

student outcomes. Results have been mixed. A self-commissioned impact analysis of NAAC found

that colleges have “started copying the top-bracket institutions….[and some institutions] had spent

much time preparing documents and plans that would impress the peer team.”65 This copying leads

to what Pritchett et. Al. term isomorphic mimicry: “the adoption of the forms of other functional

states and organizations which camouflages a persistent lack of function.”66

Upholding the “objectivity of accreditation” is also emerging as a challenge.67 Assessments with

unfavorable results lead to Universities questioning of the NAAC methodology.68 Most damaging,

however, is the perception that government-led subjective assessments are corrupt. Media reports

of bribery are common. One such report alleges a University spent Rs. 1,289,137 (~US$20,000)

on NAAC members, including gifts of rare shawls, saffron and five-star accommodation, to secure

an “A” rank.69 The AICTE is also perceived as having significant issues with corruption,70 which

in the past have led to suspension of the AICTE Chairman.71

A few outcomes focused quality assurance efforts have also been tried. AICTE requires all colleges

to submit Mandatory Disclosures each year, which includes labor market data (including salaries)

for all degree graduates.72 One study found that 82 colleges out of a sample 500 had not submitted

mandatory disclosures, and many others had obvious mistakes or misreporting. Even when

mandatory disclosures had been submitted, they were difficult to access and compare.73

In vocational education, the government is in the process of implementing an industry-sourced

National Skill Qualification Framework (NSQF) to serve as a curriculum against which vocational

skills can be assessed. Developed by Sector Skill Councils (SSCs) comprised of companies and

other industry bodies, the NSQF outlines National Occupational Standards (NOS) for 1,641 job

65 Stella, A. “Institutional Accreditation in India”, International Higher Education, 2015 66 Pritchett, Woolcock, Andrews, “Capability Traps? The Mechanism of Persistent Implementation Failure”, 2010 67 Ibid. 68 “Shock over grading of 38 deemed universities”, Times of India, November 22 2015 69 “Kashmir university ‘spices’ hands in order to obtain A-ranking”, Daily Mail, May 10 2012 70 “All India Council for Technical Education caught in a cleft”, Business Standard, November 14 2013 71 “AICTE chairman suspended over corruption case”, Indian Express, July 30 2009 72 AICTE, Approval Process Handbook, January 2010 73 Jindal, N. “AICTE, the Regulator which doesn’t share data!”, Careers 360, 6 Jan 2014

Page 33: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

27

roles across 33 sectors.74 Independent assessment agencies impaneled by SSCs carry out

assessment of candidates against the NSQF upon graduation from government sponsored

vocational education programs.

While the jury is still out on the success of NSQF implementation (with no formal evaluations),

anecdotal evidence of industry recognition and ownership of the certification is mixed. A key issue

raised is that government focus on scaling quickly is compromising the value of SSC certification.

Chasing a highly improbably target (training ~500 million additional Indians by 2022), politicians

and bureaucrats have incentives to maximize the quantity of graduates, rather than their quality.

Having outsourced assessment to third party assessment agencies, the government has not built

any in-house capability to check or assure quality of graduates, and is facing the tough reality that

industry may not be willing to recognize SSC certificates. In an effort to bolster its uptake, the

government has announced plans to make SSC certification mandatory for application to relevant

government jobs by 2020.75

5.2. Design choices

A successful intervention needs to credibly assess quality of colleges and/or job seekers to provide

employers with a reliable signal of job-seeker quality. To design a credible assessment, we must

answer four fundamental design questions:

1. What should be assessed? Should we assess inputs and processes of institutions or training

courses? Should we assess graduate learning outcomes or employability? Should we report the

salaries graduates are earning post-graduation?

2. Who should be assessed? Should assessment efforts focus on colleges as a unit or on individual

students? Should assessment be mandatory, or should people choose to opt in?

3. What should be the role of the government vs. the private sector? Who should govern the

implementation of the proposed intervention? Who should fund it?

4. Who should do the assessing? Should government agencies such as NAAC lead assessment?

Should assessment be handed over to private sector assessment agencies?

74 National Skill Development Corporation website, accessed at

http://www.nsdcindia.org/sites/default/files/files/Summary_QP-NOS_list_as_on_30th_Dec_2015.pdf 75 “More than a degree, NSQF to become mandatory for Government PSY jobs by 2020”, The Economic Times, April 9 2015

Page 34: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

28

In answering these design questions, we keep three broad criteria in mind:

Does the answer lead us to an intervention that, if successfully implemented, will address the

information failure in question (technical correctness)?

Does the answer lead us to an intervention that is likely to be implementable in the short to

medium term (administrative feasibility)?

Does the answer lead to an intervention that will garner enough political support (political

supportability)?

We consider each of the four design questions in turn.

5.3. What should be assessed?

Two categories of metrics are available to assess college quality: inputs or outcomes.

A vast literature exists on the potential gap between schooling inputs and learning outcomes.76

Measuring input metrics such as college faculty, resources or processes provide no guarantee of

what we ultimately care about: the quality of college graduates. Measuring inputs may create

perverse incentives for over-investment in areas captured by assessment at the expense of what’s

truly needed to improve the quality of graduates. This is evident in the “isomorphic mimicry”

arising out of the NAAC’s current efforts.

In measuring outcomes, a metric of chief importance to incoming students is their expected salary

after graduation. Departments of Education in several countries, including Chile and the US, report

average salary of graduates by institutions and degree program.77 Post-graduation labor market

information can help students determine the quality of colleges and create a healthy competition

for admission.

While useful for students, this metric would be telling employers what they already know: which

colleges are already recognized as superior. By publishing salaries, we would we amplify this

existing information rather than generate new signals. Salary data is also biased by caste or religion

based discrimination in the hiring process, as discussed in the previous chapter. For these reasons,

reporting wages alone is not the technically optimal solution. Further, while administrative data

76 Pritchett L, “The Rebirth of Education: Schooling Ain’t Learning”, Center for Global Development, October 2013 77 See, for example, collegescorecard.ed.gov

Page 35: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

29

on salaries is readily available in advanced countries, it is less readily available in India making it

administratively difficult to collect and share this data.

A second outcome metric is the employability of college-leaving graduates. In Colombia, for

example, all higher education graduates are required to sit a standardized test that measures both

generic and domain competencies specific to their degree.78 It is important to note that this does

not necessarily mean testing mastery of the curriculum, which can lead to rote learning on the part

of students. Rather, a combination of hard and soft skills could be tested, aimed at predicting the

productivity of a graduate in a median firm. A great deal of private sector effort is underway to

develop such predictive tests (discussed in the next design question).

Assuming reliable testing is possible, these test scores would provide employers with a new signal

of job-seeker ability, while also improving their learning of college reputation. Changing labor

market outcomes of college graduates would feed back to students making decisions about

admissions, helping them identify better colleges.

Introducing a standardized test for college graduates raises concerns of administrative and political

feasibility. It is not unprecedented, however. The National Board of Examinations (NBE),

established in 1975, administers a standardized test for graduates of Medical Colleges across the

country, following standards laid out by the Medical Council of India.79 We consider these

concerns in greater detail in the next section.

5.4. Who should be assessed?

Efforts to bring transparency to higher education are likely face stiff political opposition. Two

groups are likely to be the most vociferous attackers: low quality colleges who don’t wish to be

revealed, and elite colleges who stand to lose too much from a standardized testing regime.

78 Velosa et. Al, “Returns to Higher Education in Chile and Colombia”, IDB Working Paper Series No. IDB-WP-587, March

2015 79 See http://www.natboard.edu.in/aboutus.php

DESIGN PRINCIPLE 1: Measure employability of graduating students. Testing inputs or

labor outcomes post-graduation is technically sub-optimal.

Page 36: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

30

Low quality colleges benefit from the opacity of the current education and labor markets, which

allows them to hide in misinformation and confusion. The strength of opposition is compounded

when we consider that it’s widely reported that many low quality colleges are owned by

politicians.80 As one report puts it: “in the over-regulated education sector, politicians with clout,

connections and insider knowledge, can cut through red tape and quickly expand capacity.”81

Elite colleges already enjoy a privileged position and have little to gain from a new assessment of

quality. Indeed, efforts to impose a new assessment could show them in bad light. Given the

visibility and influence of many of these institutions, they are well placed to shut down perceived

threats. They might, for example, be well placed to question the validity of any test that were

imposed upon them.

Given these vested interests, we argue that mandatory assessment of colleges is politically

untenable. We can, however, adopt two tactics that help us mitigate and largely avoid this political

opposition.

First, we propose that assessment be focused on students, not institutions. Adopting this approach

has both political and technical benefits. Politically, assessing students side steps a frontal attack

on colleges. Technically, it gives employers precisely what they want: a signal of the ability of the

job-seeker sitting in front of them, rather than of the quality of the college that she attended.

Second, we propose that the assessment be voluntary rather than mandatory (at least initially).

Voluntary schemes will be perceived as less of a threat by low quality education and providers,

buying an intervention time to scale. Voluntary testing, however, creates a technical problem in

the validity of results. It’s unlikely that the poorest performing colleges or students will opt to sit

for an optional standardized test if they believe that the results will carry negative signals to future

employers. As a consequence, below average students, knowing that the poorest students are not

sitting the exam, will also be reluctant to taking knowing that they will be compared to students

who are better than them. Finally, students from colleges that already have stellar reputations are

unlikely to sit an exam given they have everything to lose and little to gain.

80 “For Political Class, Money Flows via TN Engg Colleges”, The New Indian Express, March 13 2016 81 Pandit, V. “Politicians in education”, Education World Online, March 5 2009

Page 37: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

31

Only in the scenario when testing reaches a critical mass such that employers demand a test score

from job-seekers, or not providing a test score carries negative information value, will an opt-in

model truly help reveal the worst performing colleges. Initially, it’s more likely that voluntary

testing provides an avenue for some good graduates to signal their ability.

We argue that this technical compromise is necessary to maintain political viability. Launching as

a voluntary scheme with long term ambitions for universal coverage is also a proven tactic among

government programs in India. The NAAC began as a voluntary scheme with a focus on self-

improvement rather than accountability or the meeting of minimum standards. 82 NAAC framed

their accreditation process as “meant for quality institutions…[and] not suitable for the ones that

might be still struggling with some basic problems.”83 As of October 2006, a total of 2,956 colleges

or 13 percent of India’s higher education institutions had been accredited.84 While commendable

from the point of view of self-improvement, this did not constitute a broad signal of college quality.

Accreditation was made mandatory by UGC (in 2013) and various state governments in order to

qualify for state and central funding. In the next year alone, 2,808 colleges and 102 Universities

“volunteered” for accreditation – equal to roughly half the total number of colleges assessed by

NAAC in 18 years of existence.85

Another Indian policy initiative illustrates that voluntary schemes, if executed well, can reach

extraordinary scale. Aadhaar, India’s unique identification scheme, started as and remains a

voluntary scheme. Its voluntary nature helped Unique Identification Authority of India (UIDAI)

avoid political opposition from other Ministries and scale at its own (albeit rapid) pace. At the time

of writing, over 950 million Aadhaar numbers have been issued.86 While much has been written

about the success and challenges of Aadhaar, we simply highlight that its design as a voluntary

scheme did not preclude its rapid scale up to near universal coverage.

Further, voluntary schemes have the benefit of being administrative feasible. The rapid ramp up

of implementation required in mandatory schemes can lead to what Pritchett et. Al. term premature

load bearing “in which wishful thinking about the pace of progress and unrealistic expectations

82 Stella, A. “External Quality Assurance in Indian Higher Education: developments of a decade”, Quality in Higher

Education, Vol. 10, No. 2, July 2004 83 Ibid. 84 Agarwal, P. “Indian Higher Education: Envisioning the Future,” SAGE publishing 2006 85 NAAC Annual Report 2013-14 86 Aadhar online portal, accessed at https://portal.uidai.gov.in/uidwebportal/dashboard.do

Page 38: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

32

about the level and rate of improvement of capability lead to stresses and demands on systems that

cause capability to weaken (if not collapse).”87 We find evidence of premature load bearing in

Indian higher education quality assurance efforts. In the case of the NAAC, the Chairman of the

UGC acknowledged that, once accreditation was made mandatory, “the present capacity of NAAC

is so very inadequate...”88 to assess all colleges. As discussed in section 5.1, unrealistic targets for

vocational education have also put NSQF under strain. Under pressure to deliver on targets, the

Skill India machinery is scrambling to maintain quality control in its programs and link trainees to

jobs.

We make one final note on eligibility for the eligibility of test-takers. Implementing a standardized

assessment opens up the playing field for students from all universities (not just the elite) to

compete for jobs. It could also create opportunities for non-University students who do not have a

formal education but have informally acquired skills to access formal sector jobs. Equal

opportunity to compete for jobs is a large benefit of a standardized assessment.

5.5. What should be the role of government vs. private sector?

A government led standardized assessment is likely to fail both politically and administratively.

Politically, government interference in University affairs is an age old topic of controversy and

debate in the Indian educational sphere, most recently cited by the Chief Minister of Delhi as

“responsible for the erosion of the education system.”89 If government led tests are seen as a

political imposition, accusations of interference will haunt their legitimacy.

Administratively, implementing a test that predicts graduates’ job performance requires keeping

assessment frameworks up to date with continuously evolving industry knowledge and

requirements. This is a difficult undertaking for bureaucracy. Even the NSQF, which relies heavily

87 Pritchett, Woolcock, Andrews, “Capability Traps? The Mechanism of Persistent Implementation Failure”, 2010 88 “Rethink required to improve higher education: UGC chief”, Business Standard, August 14 2014 89 “Political interference has eroded education system, says Arvind Kejriwal”, The Indian Express, March 27 2015

DESIGN PRINCIPLE 2: Implement voluntary assessment of students. While universal

testing is a desirable end point, mandating testing may lead to premature loadbearing in

implementation and create political opposition.

Page 39: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

33

on industry involvement through Sector Skill Councils (SSCs), has faced challenges in sourcing

standards from industry. While the aggregate coverage figure of 1,641 job roles across 32 sectors

seems impressive, just five sectors have contributed over 40% of this total. 90 Progress has been

highly variable, depending in part on the abilities and motivation of SSC members.

An employer led intervention has better chances of designing and maintaining an industry relevant

curriculum. It also has a better chance of linking education to employment for graduates.

Examining the beginnings of global standardized tests is instructive. The GMAT, now the standard

for evaluating business school applicants, began in 1953 when representatives from nine leading

business schools (Columbia, Harvard, Northwestern, Rutgers, Seton Hall, University of Chicago,

University of Michigan, University of Pennsylvania and Washington University in St. Louis) met

to establish a standardized assessment.91 The SAT was first administered by a teacher at Princeton

and subsequently adopted by Harvard to administer a scholarship program for gifted students.92 In

both cases, a group of aspirational Universities led the launch of the standardized tests.

A standardized test for college graduates would do well to similarly start with a community of

aspirational end-users of the information it generates. In our case, implementation could be

anchored with a group of companies that graduates aspire to work for, and who are motivated to

participate for self-interest in better talent sourcing.

As we noted in chapter 4, however, not every industry has the features necessary to create

incentives for firm-level investments in assessment and screening of graduates. Because the

benefits of screening are not captured by any one employer, the economic case will not stack up

for firms in industries with low levels of human capital, low variance in quality of graduates, or

high cost of testing.

The government can catalyze screening by improving the economics of assessment in these

industries. This could be in the form of a per-student-assessed subsidy to assessment agencies,

similar to payment currently being made for vocational education assessments under the Pradhan

90 National Skill Development Corporation website, accessed at

http://www.nsdcindia.org/sites/default/files/files/Summary_QP-NOS_list_as_on_30th_Dec_2015.pdf 91 “The History of the GMAT”, Economist GMAT blog, November 5 2014 92 Leeman, N. The Big Test: The Secret History of the American Meritocracy, 2000

Page 40: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

34

Mantri Kaushal Vikas Yojana (PMKVY) scheme. The subsidy would cover only a part of the

assessment costs, with the remainder being covered by employers and the student.

It’s important to note that this manner of subsidy might create incentives for assessment agencies

to complete as many assessments as possible rather than focus on the quality of each individual

assessment. We discuss mechanisms for quality assurance in the next section.

5.6. Who should do the assessing?

Reliability and industry relevance are key to credible assessment. A large government role in

assessment design and administration is unlikely to succeed. Recognizing this, and responding to

limited in-house capabilities to assess, government has started engaging independent assessors of

employment skills. NSDC, for example, has impaneled 21 assessment agencies to evaluate the

learning of graduates from its vocational training programs.93 Many of these agencies also assess

non-vocational skills and skills specific to white-collar industries, such as engineering. This

expertise could be leveraged in broader assessment across many sectors.

As mentioned in the previous section, subsidy payments to independent assessment agencies

should be made with strong safeguards in place to ensure the quality of assessment remains high.

We argue that the most powerful safeguard is a reputational incentive of certifying agencies. If the

GMAT, for example, started producing unreliable scores, business schools would quickly stop it

as a benchmark test for admissions.

This powerful incentive should not be lost when assessment agencies partner with government.

Certificates issued under the auspices of NSDC, for example, do not feature the name of the

assessment agency who evaluated the candidate. This removes the incentive for quality assurance

93 NSDC website, accessed at http://nscsindia.org/NSCSAssessmentAgency.aspx

DESIGN PRINCIPLE 3: Assessment should be led by a group of companies that graduates

aspire to work for. Government could subsidize a part of assessment costs in order to

catalyze company action in low-skill industries.

Page 41: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

35

by impaneled assessment agencies, who instead are driven to chase a per student assessment fee

and push for quantity rather than quality of evaluation.

Learning from this, industry should retain control of assessment agencies to reduce likelihood of

(perceived) capture of assessment quality in the public sector. Companies in a sector may then

decide to award multi-year contracts to agencies that are able to prove reliability of their tests in

trials. Companies in a sector may choose to award contracts to multiple assessment agencies to

maintain competition that leads to improvement in assessment quality. If this approach were

followed, certificates issued by each assessment agency should carry their name, so employers can

continue to learn about their performance. It is important to note that the use of multiple assessment

agencies within a sector may have technical trade-offs, as it would be harder to compare scores

across assessors. In the long run, however, competition may lead to a single assessment agency

emerging as a market leader with a superior product to service the industry. This decision could

be made by companies at the sector level, depending on the level of maturity of the screening and

assessment technologies already available in the market.

6. IMPLEMENTING: AN INDUSTRY LED, GOVERNMENT SUPPORTED

INTERVENTION

As our discussion so far has made clear, private sector leadership is critical to the success of our

proposed intervention. In this section, we outline a framework for cooperation between

government and employers that can enable implementation.

6.1. Employer led Job Readiness Councils

We recommend that our program be anchored at the sector level. This allows employers the

flexibility to choose a screening technology most suited to their needs. As we noted in the previous

chapter, employers of low-skill workers focus on behavioral attributes that are often measured

DESIGN PRINCIPLE 4: Third party assessment agencies can be engaged to design and

deliver tests, provided a reputational incentive is in place to ensure they maintain the

quality of assessments.

Page 42: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

36

through personality or culture-fit tests. Employers looking for high-skill candidates are likely

emphasize technical skill or soft skills much more.94

Priority sectors could be chosen based on (a) the size of labor force affected, (b) the levels of

human capital of sector employees, and (c) levels of employer coordination. The rationale for

criteria (b) lies in our discussion in section 4.4: the economic case for an employer-led screening

solution is more likely to succeed in industries with higher levels of human capital. Sectors in the

“sweet spot” for our intervention fall just short of the levels of human capital required to

incentivize employers to implement firm-level screening solutions (e.g., IT companies,

professional service companies). Levels of human capital within each sector could be measured

by the share of sector employees who have tertiary education.

Within each sector, “Job Readiness Councils” (JRCs) should be established. 1-2 representatives

from industry associations, such as the Confederation of Indian Industry (CII), will play an

important convening role in JRCs. The authority of JRCs will come from the 3-4 major employers

in the industry, who should all have 1 representative on the council. A representative from a mid-

size company in the industry would also lend the JRC legitimacy among a broader employer base.

Given they are partial funders of the program, a representative from Government would also be

present on the JRC. The representative could be a Director or Joint Secretary level bureaucrat from

the Ministry of Human Resource Development (in industries where job seekers are tertiary

graduates) or the Ministry of Skill Development and Entrepreneurship (where job seekers are

vocational program graduates).

It’s important to note that, while industry associations may play a convening role, employers need

to be the champions and motivators of this proposal. Employers should have full confidence that

the proposed intervention is in their best interests, and intervention design should be re-worked

until it has the full buy-in of the JRC. A motivated group of employers and JRCs should be a pre-

condition for implementing the intervention in any given sector.

94 Osterman, P. “Skill, Training, and Work Organization,” Industrial Relations, Vol. 34, No. 2, 1995

Page 43: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

37

6.2. Implementation plan

The JRC has three broad functions:

Prepare

1. Raise funds from industry employers to cover implementation and industry share of assessment

costs. This will also serve as a litmus test of industry support.

2. Set up a Standards Unit (SU) tasked with creating industry standards for key job roles. The SU

can comprise of HR representatives from industry employers and education and assessment

experts from industry. The SU should also draw on existing job standards created by member

employers and by government bodies such as Sector Skill Councils (SSCs). Job Readiness

Standards should be sent to industry employers for feedback and ultimately approved by the

JRC.

3. Contract assessment agencies to test against Job Readiness Standards.

Promote and Execute

4. Secure commitment from at least 3-4 employers to at least interview all candidates who place

in the top quartile (threshold can be revised) of Job Readiness Tests, regardless of their

background or other characteristics.

5. Market this commitment aggressively in their recruiting efforts and in colleges. We note that

it will also be in the commercial interest of assessment agencies to source test-takers, as they

are paid on a per-test basis (with reputational incentive to ensure quality).

6. Oversee assessment agency implementation of tests.

Publish

7. Publish anonymized data on test performance, broken down by college/institution of test-sitter,

on a public website.

8. Collect data on performance of Job Ready test-takers in interviews and, over time, on-the-job,

and share with JRC to inform program improvements.

It is important to note that government should play a minimal role in implementation. Government

should not co-brand the initiative, leaving it to the JRC to establish an autonomous brand.

Government should not link its funding to scale or outcome targets that may compromise the

Page 44: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

38

quality and reliability of tests. Perhaps most importantly, government should also resist the

temptation to merge JRCs with existing Sector Skill Councils (SSCs) or other government skill

agencies. Prima facie, JRCs and SSCs have very similar functions: set industry occupational

standards and test against them. SSCs, however, are focused on graduates of government-

supported short term vocational training programs (though this is changing with a focus on

“Recognition of Prior Learning” in the Ministry). JRCs will focus on all job-seekers, with an initial

focus (as discussed in section 6.1) on sectors with higher levels of human capital. This means that

test-takers initially are likely to be tertiary graduates, who are not the audience of SSCs.

Where conflicts do arise, however, we argue that industry doubts about SSCs arise precisely from

the amount of influence government is able to exert over them. This has led to a premature focus

on scale and universal coverage with insufficient efforts to ensuring quality and connection to

employer needs. Establishing industry led JRCs will duplicate the work of SSCs in some sectors,

but with the important design change that they JRCs are owned and championed by industry. SSC

job standards are approved, for example, by the National Skill Development Agency (NSDA) – a

government body – and not industry employers. SSCs respond to Ministry and NSDA targets to

create occupational standards and certify trainees, not to industry employers.

In cases of conflict, the Ministry of Skill Development and Entrepreneurship (MSDE) may

understandably oppose funding of JRCs, given they have invested considerable effort setting up

their own SSCs. In response, JRCs should:

Remain firm on independence. Government plays a minority role in funding and project

governance precisely to prevent it from exerting undue influence. If insisting on autonomous

operation costs JRCs their government funding, this is an acceptable compromise. Government

funding could be replaced by employer contributions or by the increasing amount of

philanthropic funds being directed to education-to-employment issues by organizations such

as the Rockefeller Foundation.95 These funds could be accessed through CII or a non-profit set

up to oversee the JRC’s activities.

Avoid. A tactical focus on tertiary graduates early on will help JRCs gain scale and establish

credibility before dealing with possible opposition from MSDE.

95 See, for example, Rockefeller Foudnmation’s employment initiative in the US:

https://www.rockefellerfoundation.org/our-work/initiatives/youth-employment/

Page 45: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

39

Collaborate. Cooperation is a preferred outcome to confrontation. The JRC should remain

open to adopting SSC job standards where they exist and are of high quality. In the longer

term, JRCs should look for politically feasible ways to subsume SSC activities and funding

into JRCs (while still maintaining autonomy).

6.3. Monitor and course-correct

Data should be collected at all stages of program to troubleshoot implementation bottlenecks or

program design failures. A preliminary list of metrics to measure the progress, outputs and

outcomes of the intervention are presented in the table below.

Table 6.3.1: Proposed metrics and course-corrections

Metrics Potential causes Course correction

Progress Job readiness

standards not

being created at

expected pace

1. Low industry buy-in. Not

confident in the solution

2. Expertise to write standards is

lacking

1. Consult major employers on

solution. What don’t they like?

2. Engage education experts or

assessment agencies in standard-

writing procedure

Assessment

agencies have not

been contracted

1. Low industry buy-in

2. Economics for assessment are

off

1. Consult major employers

2. Revisit assumptions on

government subsidy, employer

contribution and student

willingness to pay

Unable to raise

money from

employers to fund

initiative

1. Low industry buy-in

1. Consult major employers

Outputs Few candidates

assessed (low test

volume)

1. Candidates not aware of test

2. Candidates don’t think the test

helps their chances of getting a

job

3. Assessment agencies have

limited capacity to implement

1. Increase program marketing

efforts

2. Investigate job-search benefits to

test-takers. If they exist, publicize. If

they don’t, consult employers

3. Re-contract assessment to

more/a different assessment

agency

Average test score 1. Low score + low variance =

test is too hard, inadequate

supply of skills

2. High score + low variance =

test is too easy

3. High variance = test is

performing its task screening

applicants

1. Revisit test methodology/re-

contract assessment agency

1.b. Consider efforts to deliver

training directly

2. Revisit test methodology/ re-

contract assessment agency

Outcomes Collect additional

data from

employers about

1. Candidates with higher test

scores not performing better on

the job (wage, promotion,

retention)

1. Revisit testing methodology

Page 46: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

40

on-the-job

performance

Impact on higher

education sector

1. Data not being used by

students for admissions

decisions

1. Redouble marketing and

awareness efforts

Continued

implementation

1. Employers stop wanting to give

funds to the program

2. Test volumes decline

Consult employers and/or students

about possible dissatisfaction with

the program

6.4. Costs and Benefits

Selecting three pilot sectors, we estimate the cost to industry per hired test-taker to range from 3

to 10 percent of the average annual salary of workers. Productivity gains are between 1.2 and 2.9

times as large as the cost (with very conservative assumptions). Even if the program fails, the total

expenditure from all employers in the industry is small: ranging from Rs. 128 million to Rs. 319

million, depending on the sector. This represents a small fraction of the total recruiting spending

by companies. Details of our assumptions are included in Appendix 2.

The outlay for government is even smaller, barely meriting a line item in any Ministry’s budget.

Social returns on this small investment could be extremely large, however. Industry will benefit

from higher productivity workers and employees will benefit from higher wages. By fixing

information failures, the test will also create incentive for individuals to acquire skill and improve

the quality of higher education institutions. While difficult to quantify, these benefits dwarf the

associated costs.

Perhaps more important than the overall economic case for implementation will be the ability for

industry and government to communicate “program wins” to the public and other stakeholders.

Communicating metrics such as the number of tests administered, proportion of test takers hired,

performance of test takers on the job, and performance of results by college will all help

demonstrate progress and value to program stakeholders.

6.5. Next steps

An industry association (such as the Confederation of Industry Association) is well placed to

convene groups of employers in priority sectors to discuss (a) the need for, and (b) the design of

the proposed solution would be a good first step. In industries with employer buy in, these groups

of employers could form JRCs and continue implementation, as described above.

Page 47: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

41

Page 48: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

42

APPENDIX 1. ADMISSIONS SORTING AND WAGE PREMIA96

A1.1. Introduction

To understand the extent of information failure in India’s labor and education markets, we tested

college’s reputation for quality affects (a) the students it attracts in the admissions process, and (b)

the labor market outcomes of its graduates. To do this, we follow the methodology outlined by

Macleod et. Al (2015) to measure college quality, admissions sorting and reputational return. We

apply this methodology to a dataset obtained from a leading Indian employment assessment

agency. This dataset is not representative of the labor market in general, so findings are limited to

our sample.

Section A1.2 describes our approach to measuring candidate and college quality. Section A1.3

describes our dataset. Section A1.4 reports our empirical results on admissions sorting and section

A1.5 reports results on income return on college quality. Section A1.6 highlights areas for future

research and concludes.

A1.2. Candidate quality and college quality

Employers hiring fresh graduates typically observe (albeit imperfectly) four measures of candidate

quality on CVs: their year 12 score, college GPA, AMCAT scores (explained below) and the

reputation of the college from which the candidate graduated.

The first three measures are signals of candidate quality that are directly measured and reported in

our database. Year 12 scores are typically reported along with the board of education the candidate

graduated from. The Indian education system offers students a choice to complete secondary

education under the auspices of one of 54 state or central boards of education. 97 Anecdotally,

central boards – the most popular of which is the Central Board of Secondary Education (CBSE)

– have a reputation for being tougher, and therefore scores are expected to be lower. College GPAs

are also measured and reported directly in our data.

96 This appendix reproduces, with some modifications, work completed by the author in ECO2810a Labor Market Analysis,

with Professor Katz. While additional analysis has been conducted, large sections of the text are reproduced verbatim. 97 Council on Boards of School Education in India, http://www.unishivaji.ac.in/uploads/admin/circ/board_list.pdf

Page 49: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

43

In addition to school and college scores, we are also able to measure ability using scores from the

standardized Aspiring Minds Computer Adaptive Test (AMCAT). AMCAT is a third party

employability assessment administered by Aspiring Minds (an assessment agency) to evaluate job

seeker skills. Approximately one million test takers pay Rs 900 (~$15) to sit the test each year.

The AMCAT comprises of three general tests assessing candidates’ logical, quantitative and

English language skills. Each of these is scored on a scale of 0-800, giving a combined total of

2400 points. Candidates may also opt to take domain specific tests in Computer Programming and

in various branches of Engineering (Electrical, Mechanical, etc.). Since different candidates give

different domain-specific tests, scores are reported as the percentile rank of the candidates in their

respective test. We note that candidates with bad scores may choose not to share them with

employers. It is costlier to hide college GPAs or year 12 scores, as their omission may carry

negative informational value about the candidate.

The fourth measure – college reputation – is more difficult to assess, but nevertheless considered

important by employers when hiring fresh graduates. Especially given competitive entrance in

many Indian colleges, reputation is used as a signal of candidate quality. Apart from signaling,

going to a better college may also improve the labor market prospects of a student by improving

their skills or giving them access to alumni networks. College reputation is multi-dimensional,

incorporating quality of student, faculty, financial and other inputs. To measure it, we adopt the

approach outlined by Macleod et. Al (2015), who reason that we would expect to see colleges with

better reputations attract better quality students to enroll and would churn out higher quality

graduates. Following their logic, we offer four methods of measuring college reputation.

Our first two measures of college quality are the mean AMCAT scores of graduates from each

college. We separately report a college’s mean student AMCAT scores in general tests (English,

logical, quantitative) and AMCAT percentile ranks in domain specific tests (computer

programming, engineering, etc.). One might argue, however, that the AMCAT is not widespread

enough for the labor market to use it as the basis for reputation-formation. Given the absence of a

standardized graduation test, the reputation of colleges among employers may depend simply on

the difficulty of college admission.

Our third measure of college reputation follows from this and reports the quality of enrolling

students, measured by their mean year 12 percentile rank (scaled for central or state board

Page 50: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

44

attendance). Using year 12 percentile ranks measures quality of enrollees rather than the quality

of graduates may be cause for concern if colleges with higher reputation also add more value to

their students. In this case, measuring the quality of enrollees would understate the difference in

quality of graduates from good and bad colleges. To allow for this, following Macleod et. Al

(2015), we make the additional assumption that college value add is positively correlated by with

labor market reputation, and so is preferred by higher ability enrolling students. Therefore, using

year 12 scores gives us a noisy measure of what we – and employers – actually care about: the

quality of college graduates.

Our fourth measure of college reputation adopts the point of view of enrolling students, rather than

employers. We argue that the most important information at the time of enrolment are the salaries

obtained by college graduates. From the point of view of students, we measure college quality at

time t as the average salary of a graduating cohort at time t-3. Students learn from the salaries of

the graduating class at the time of their enrollment.

Using these measures, we test the extent of ability sorting by plotting college quality against

candidate quality. As alluded to in the main report, we note there may be non-information related

reasons for incomplete sorting by ability. We also note that, while admissions sorting undoubtedly

does take place elite Indian institutions, our sample focuses on the more amorphous “middle” of

Indian colleges.

To test wage premiums for college quality, we assume that firms set wages based on expected

ability of workers, given the information available to them: year 12 scores (adjusted for school

board), college GPA, AMCAT scores and college reputation. If this is indeed the case, we would

expect to see positive and significant unconditional returns to year 12 scores, college GPA,

AMCAT scores and college reputation. Conditional returns would be smaller in magnitude, but

still significant.

A1.3. Data

This paper draws on Aspiring Minds’ Employability Outcomes (AMEO) 2015, a dataset released

by Aspiring Minds, an assessment company headquartered in India. AMEO 2015 provides

AMCAT scores for 3,998 engineering undergraduates, together with demographic information,

educational background (high school and college scores and other information) and employment

Page 51: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

45

outcomes (first job salaries). Data was collected via an email survey sent to previous test-takers

who are now in the labor market.

The dataset is not representative of the Indian labor market, and so our results cannot be

generalized beyond our sample. Descriptive statistics are provided in Table 3.1, split by year of

graduation from university. The average age of the candidate in the dataset is 25 years and the

average annual salary is Rs 307,700 (~US$4,600), roughly four times India’s GDP per capita.

Graduates from 1350 colleges are covered. The average test taker scored 1517 out of a total

possible 2400 points in the compulsory module of the AMCAT and had a college GPA of 71.5%.

The average year 12 graduating percentage is 74.5%. It is widely acknowledged that central board

exams are more difficult than state board exams. Anecdotally, employers use rules of thumb such

as subtracting five percentage points from a state board exam score to compare to a central board.

This is corroborated in our data, which shows that despite having lower average year 12 scores,

central board students have higher average AMCAT test scores and higher average salaries.

To more formally compare state and central board scores, we adopt a two-step approach. First, we

convert scores to percentile ranks of students within their board. Second, we scale these percentile

ranks using the AMCAT score as an objective third party assessment. We shift the state board

distribution downwards, multiplying each state percentile rank by the ratio of the average state

board AMCAT to the average central board AMCAT score. Using the ratio of median AMCAT

yields similar results: scaling all state board year 12 scores by a factor of 0.91.

This methodology is crude. Scaling using the AMCAT test, which is taken nearer to the time of

graduation than enrolment, conflates talent of incoming students with value-add of colleges. If

higher value-add central colleges prefer central board students, we would expect their AMCAT

scores to be higher regardless of initial ability. In this case, our scaling factor of 0.91 would be too

small. We consider the scaled percentile to be a lower bookend for the true estimate, and compare

with the unscaled percentile as the upper bookend.

Our analysis excludes respondents who failed to provide a graduation year or who are still in

university (graduation years 2016/17), leaving us with 3,982 candidates. We also exclude graduate

candidates, focusing only on test-takers with a Bachelor of Technology or Bachelor of

Engineering. Given our analysis focuses on college reputation, we only consider the 221 colleges

with five or more graduates represented in the dataset, reducing our sample to 1,959 candidates.

Page 52: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

46

Table A.1.3.1: descriptive statistics of the AMEO 2015 database, by graduation cohort

Graduation year

All

years

Variable N/A 2007 2009 2010 2011 2012 2013 2014 2015 2016 2017

Salary (Rs p.a.)

325,00

0

120,00

0

365,62

5

441,31

8

369,68

4

315,47

8

277,81

0

267,92

5

299,57

4

295,71

4

195,62

5

307,70

0

Age 24 29 29 27 26 25 24 23 23 23 24 25

Year 12 score

(%)

85

87

67

74

73

75

75

75

74

78

69

74

College GPA

(%)

69

65

68

70

69

72

72

72

71

64

68

71

AMCAT

English

695

425

456

475

494

488

509

516

511

574

504

502

AMCAT Logic

655

575

471

472

493

498

512

507

493

503

454

502

AMCAT Quant

680

645

485

508

512

512

524

507

482

587

481

513

AMCAT

domain

percentile rank 0.76 . 0.61 0.53 0.53 0.56 0.47 0.50 0.48 0.66 0.42 0.51

Candidates

1

1

24

292

507

847

1,181

1,036

94

7

8

3,998

Proportion

male 100% 100% 75% 82% 74% 73% 76% 77% 84% 86% 88% 76%

Page 53: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

47

A1.4. Admissions sorting

To understand admissions sorting in Indian colleges, we plot candidate quality (measured by

scaled year 12 percentile rank) against each of our measures of college quality (AMCAT general,

AMCAT domain, year 12 scores, graduate salaries). When measuring college quality by graduate

salary, we introduce a time lag of three years. I.e., the information 2014 graduates are able to see

are the labor market outcomes of graduates three years their senior. In all other cases, we measure

college quality using the 2014 cohort only. Our results are presented in figures 4.1 to 4.4.

In each of figures 4.1 to 4.4, the horizontal line represents the outcome that would be expected if

there was no sorting by ability at admissions and students were allocated randomly to colleges. A

positive slope suggests some sorting. The steeper the slope, the greater the extent of admissions

sorting.

Figure 4.1 shows college reputation measured by AMCAT general scores against year 12

percentile rank. We see admission sorting does indeed take place, but is limited. The local linear

regression line of best fit shows some – but not much – sorting by ability. Figure 4.2 shows

measuring college reputation by AMCAT domain percentile ranks does not change the picture by

much. While these figures show scaled year 12 percentile rank (adjusted for central vs state board

of education), using unscaled year 12 percentile rank does not change the picture.

Figure 4.3 measures college reputation by year 12 percentile rank. As we would expect, measuring

college reputation by year 12 scores mechanically shows stronger sorting. But sorting is still far

from complete, suggesting that factors such as cost-constraints, information failures and

affirmative action are indeed at play.

Finally, figure 4.4 measures college reputation by 2011 graduate salaries. Sorting is incomplete

and inconsistent, suggesting the flow of information of labor market outcomes back to enrolling

students is incomplete.

Page 54: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

48

Page 55: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

49

We also look at the variation of ability levels within any given college. Figures 4.5 to 4.8 rank

colleges by quality, according to each of our four measures, on the x-axis. On the y-axis, we show

the year 12 percentile score of the median college student (dot) as well as the range from the 25th

to the 75th percentile college student (red line range). As we can see, while admissions sorting

across colleges does occur, intra-college ability variation is just as significant (if not more so) than

variation across colleges.

For suggestive comparison, we include a similar analysis of sorting in Chile (Macleod et. Al

(2005)), who use scores on a standardized entrance test – Icfes – to measure ability, and measure

college reputation as the mean Icfes percentile of college graduates. The authors conclude that

admissions sorting is incomplete in Chile, and a visual comparison of the two charts suggests

similar levels of intra-college ability variation in the two countries.

Page 56: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

50

Page 57: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

51

A1.5. Earnings

To test our prediction on earnings, we estimate the following regression

𝑤𝑖 = 𝛽0 + 𝛽1𝑦𝑒𝑎𝑟 12 𝑠𝑐𝑜𝑟𝑒𝑖 + 𝛽2𝑐𝑜𝑙𝑙𝑒𝑔𝑒𝐺𝑃𝐴𝑖 + 𝛽3𝐴𝑀𝐶𝐴𝑇𝑔𝑒𝑛𝑒𝑟𝑎𝑙𝑖 + 𝛽4𝐴𝑀𝐶𝐴𝑇𝑑𝑜𝑚𝑎𝑖𝑛𝑖

+ 𝛽5𝑐𝑜𝑙𝑙𝑒𝑔𝑒𝑟𝑒𝑝𝑢𝑡𝑎𝑡𝑖𝑜𝑛𝑖 + 𝛽5𝑔𝑟𝑎𝑑𝑢𝑎𝑡𝑖𝑜𝑛𝑦𝑒𝑎𝑟𝑖 + 𝛽6𝑏𝑜𝑎𝑟𝑑𝑜𝑓𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛𝑖

+ 𝛽7𝑔𝑒𝑛𝑑𝑒𝑟𝑖 + 𝜀𝑖

The dependent variable, 𝑤𝑖, measures log annual salary of workers. We regress log wages against

an individual’s year 12 score (unscaled), college GPA, AMCAT score for general skills (logic,

quantitative and English) and domain specific skills (computer programing, engineering, etc.),

college reputation (measure either by AMCAT scores or year 12 scores) and dummies for

graduation cohort, school board of education and gender.

Estimating 𝛽5 in the equation above gives us the “college reputation premium” in our sample,

describing the average return to going to a better college. Note, we cannot identify this as a causal

effect, but as a descriptive return of going to a better college for students in our sample.

Unconditional return to college quality

Table 5.1 estimates the regression above with each explanatory in turn (together with dummies),

illustrating the unconditional return to reputation (and ability). We see that a one percent point

increase in college reputation measured by year 12 scores is associated with a 0.8% increase in

wages. A 10-point increase in college reputation measured by AMCAT general scores (which

range from 0-2400) is associated with 1.2% higher wages. This is as theory would predict: a

positive unconditional return to reputation. Measuring college reputation by AMCAT domain

scores similarly shows that a one percentile improvement in your domain score rank is associated

with 1.0% increase in wages.

Unconditional return to ability

Unsurprisingly, we note similar positive unconditional returns to ability as measured by year 12

score, college GPA or AMCAT score. One percentage point higher year 12 scores or college GPA

are associated with 1.6% and 1.8% higher wages respectively. Scoring one percentile higher in the

Page 58: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

52

AMCAT domain test is associated with an 0.4% increase in wages, while scoring 10 points higher

in the AMCAT general tests is associated with a 0.8% increase.

Conditional returns to college reputation and ability

Estimating the full equation above gives us the conditional returns to each explanatory variable.

We expect conditional returns to be smaller than unconditional returns, but still positive. Given

admissions sorting in India is substantially incomplete, we might expect the conditional return to

college return to be significantly smaller. With limited admissions sorting, the signal value of

college reputation is diluted. If, however, the college reputation premium reflects college value-

add apart from its signaling value, we may still expect positive conditional returns. This would be

the case when college value-add (skill improvement, networks) is heterogeneous and positively

correlated with reputation.

Table 5.2 estimates conditional returns. We see that returns to signal of general ability roughly

halve when conditioned on college reputation, reducing from 1.6% to 0.6%-0.8% for year 12 score

(depending on how college reputation is measured) and 1.8% to 0.7-0.8% for college GPA.

Returns to the AMCAT general test reduce from 0.8% to 0.5% per 10 points. Returns to AMCAT

domain scores drop more drastically, from 0.8% to 0.1% once conditioned on college reputation

measured by year 12 scores or AMCAT general scores. Conditioning on college reputation based

on AMCAT domain scores reduces returns even further to 0.08%.

Conditional returns to college reputation are also lower, but still significant, once controlling for

other signals of ability. Regressions 1 and 2 show that the conditional return to reputation as

measured by year 12 scores and AMCAT general scores falls by almost two thirds, from 0.8% to

0.3% and 1.2% to 0.4% respectively. Conditional return to reputation measured by AMCAT

domain, in regression 3, also falls, but only by half – from 0.8% to 0.4%.

This suggests the reputation premium is larger at colleges renowned for their technical skill, rather

than their ability to produce good generalists. This would make sense in the context of hiring

engineers or computer programmers. Regression 4 “throws the kitchen sink” at the data and shows

that, once employers have information on an individual’s test performance, they only pay more for

graduates from colleges with good reputations as measured by technical skill.

Page 59: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

53

Table 5.1. Regression results, unconditional returns

(1) (2) (3) (4) (5) (6) (7)

VARIABLES lnsalary lnsalary lnsalary lnsalary lnsalary lnsalary lnsalary

Year 12 Percentage

(total 100)

0.0164***

(0.00105)

College GPA

(total 100)

0.0180***

(0.00147)

AM general (total 2400) 0.000830***

(4.55e-05)

AM domain (total 100) 0.00366***

(0.000380)

College quality

(year 12 score)

0.00874***

(0.000716)

College quality

(AMCAT general)

0.00116***

(8.93e-05)

College quality

(AMCAT domain)

0.0102***

(0.000990)

Constant 11.41*** 11.42*** 11.52*** 12.44*** 12.20*** 10.90*** 12.04***

(0.178) (0.191) (0.168) (0.167) (0.167) (0.211) (0.175)

Observations 1,959 1,959 1,959 1,959 1,959 1,959 1,959

R-squared 0.175 0.138 0.207 0.113 0.137 0.146 0.119

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Note: estimates on dummies for graduation cohort, board of education, gender and regression constant are not reported.

Page 60: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

54

Table 5.2. Regression results, conditional returns

(1) (2) (3) (4)

VARIABLES lnsalary lnsalary lnsalary lnsalary

Year 12 score 0.00599*** 0.00843*** 0.00827*** 0.00712***

(0.00137) (0.00114) (0.00114) (0.00142)

College GPA 0.00768*** 0.00698*** 0.00757*** 0.00742***

(0.00151) (0.00150) (0.00150) (0.00151)

AMCAT general 0.000547*** 0.000460*** 0.000519*** 0.000482***

(5.05e-05) (5.50e-05) (5.11e-05) (5.58e-05)

AMCAT domain 0.00123*** 0.00116*** 0.000785** 0.000901**

(0.000371) (0.000371) (0.000387) (0.000389)

Reputation (measured 0.00288*** 0.00133

by year 12 scores)

(0.000834) (0.000963)

Reputation (measured 0.000417*** 0.000203

by AMCAT general)

(9.56e-05) (0.000125)

Reputation (measured 0.00425*** 0.00263**

by AMCAT domain)

(0.000994) (0.00118)

Observations 1,959 1,959 1,959 1,959

R-squared 0.263 0.265 0.265 0.268

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Note: estimates on dummies for graduation cohort, board of education, gender and regression constant are not reported.

Page 61: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

55

A1.6. Conclusion and future research

Colleges with better reputations are able to attract better students, but admissions sorting is far

from complete. Measuring reputation using a third party assessment (the AMCAT test) shows very

limited sorting. Factors such as cost, information failures and affirmative action quotas could be

preventing admissions sorting by pure ability.

Nevertheless, graduates from better colleges command a reputation premium in their first jobs.

This is particularly true of colleges with strong technical reputations. It is unclear, however,

whether this premium reflects the value of reputation as a signal of ability, or the value-add of

attending a better college. Data on the earnings growth of graduates would help us distinguish

between signaling and human capital models of college reputation.

While graduates do command a reputation premium, it goes without saying that this premium

might be larger if admissions sorting improved and became a better signal of candidate ability.

Assuming the binding constraint to more sorting are information failures (rather than cost or

quotas), interventions to bridge these failures could trigger a virtuous cycle between greater

admissions sorting improved college reputation premium in the labor market.

We are limited in data to the starting salary of graduates, and do not have access to their subsequent

earnings growth. Efforts to obtain time series data are underway for future research, which could

test a richer set of predictions along the lines of Farber and Gibbons (1996), Altonji and Pierret

(2001) and Macleod et. Al (2015).

Specifically, in future research we hope to test whether college reputation pure signaling, in which

case we expect the unconditional return to college reputation to stay flat with experience, and the

conditional return to fall. We would also expect to see the both the unconditional and conditional

return to unobserved ability to rise with experience.

Page 62: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

56

APPENDIX 2. ASSUMPTIONS FOR INDUSTRY COST-BENEFIT ANALYSIS

Assumption rationale

Tourism, Hospitality

and Travel

Auto and auto

components

Electronic and IT

hardware

Volume

Estimated annual manpower requirement National Skill Development Corporation estimates, annualised 648,000 390,000 461,000

Share covered by JRTs Voluntary, so intial coverage is low 30% 30% 30%

Total number of assessments 194,400 117,000 138,300

Share of test-takers hired Conservative, given above average candidates volunteer for test 10% 10% 10%

Total number of test-takers hired 19,440 11,700 13,830

Cost

Cost per assessment Multiple of Rs 900 cost in IT 1,000 2,000 3,000

Total assessment cost 194,400,000 234,000,000 414,900,000

Other implementation costs 10% of assessment cost 19,440,000 23,400,000 41,490,000

Total cost 213,840,000 257,400,000 456,390,000

Industry share Higher in industries with higher skill level 60% 60% 60%

Total cost to industry 128,304,000 154,440,000 273,834,000

Cost per test-taker hired 6,600 13,200 19,800

Benefit

Average annual salary

Auto and electronics based on US Bureau of Labor Statistics

hourly wage estimates. Tourism assumed by author 192,000 275,000 240,000

Productivity gain of a better hire

Staffing.org estimate of productivity advantage of a blue collar

worker one standard deviation above the mean 20% 20% 20%

Total productivy gain All test-takers hired are one standard deviation above the mean 746,496,000 643,500,000 663,840,000

Productivity gain captured by industry Assuming 50 percent lost to higher wages 373,248,000 321,750,000 331,920,000

Productivity gain to industry per test-taker hired 19,200 27,500 24,000

Productivity gain : cost ratio 2.9 2.1 1.2

Page 63: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

57

APPENDIX 3. BIBLIOGRAPHY

Aadhar online portal, accessed at https://portal.uidai.gov.in/uidwebportal/dashboard.do

Agarwal, P. “Indian Higher Education: Envisioning the Future,” SAGE, 2006

Ahsan, A and Pages, C. “Are All Labor Regulations Equal? Evidence from Indian Manufacturing”,

Institute for the Study of Labor (IZA), Discussion Paper No. 3394, March 2008

Amriapu, A. & Subramanian, A. “Manufacturing or Services? An Indian Illustration of a

DevelopmentDilemma,” Center for Global Development, Working paper, June 2015.

Amriapu, A. & Gechter, M. “Indian Labor Regulations and the Cost of Corruption: Evidence from

the Firm Size Distribution,” November 2014.

Altonji, J. & Pierret, C. “Employer Learning and Statistical Discrimination,” Quarterly Journal of

Economics, 116, February 2001

Asian Development Bank, “Key Indicators for Asia and the Pacific”, 2015

Aspiring Minds, “National Employability Report”, 2013

Aspiring Minds Employment Outcomes Database, 2015

AT Kearney, “Trends in Indian Infrastructure Development”, 2013, accessed at http://www.i-

cema.in/pdf/trends_india_infra_dev_page.pdf

Atlas of Economic Complexity, Harvard Center for International Development, accessed at

http://atlas.cid.harvard.edu/rankings/growth-predictions/

Azam, M. “India’s Increasing Skill Premium: Role of Demand and Supply,” Institute for the Study

of Labor (IZA), Discussion Paper No. 3968, Jan 2009.

Banerjee et. Al, “Remedying Education: Evidence from Two Randomized Experiments in India”,

2007

Bertrand, M et. Al., “Contract Labor and Firm Growth in India”, Working Paper, 2015

Besley, T and Burgess, R. Can Labor Regulation Hinder Economic Performance? Evidence from

India, 2004

Bhattacharjea, A. “Labour Market Regulation and Industrial Performance in India: A Critical

Review of the Empirical Evidence”, Centre for Development Economics, Delhi School of

Economics, Working Paper No. 141, June 2006

Blatman, C and Ralston, L., “Generating Employment in Poor and Fragile States: Evidence from

Labor Market and Entrepreneurship Programs”, 2015

Page 64: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

58

Blom and Saeki, “Employability and Skill Set of Newly Graduated Engineers in India”, World

Bank Policy Research Working Paper 5640, April 2011

Cho et. Al, “Impact of vocational training on the employability, earning potential and sexual

behavior of youth in Malawi”, J-PAL, 2011

Council on Boards of School Education in India board list, accessed at

http://www.unishivaji.ac.in/uploads/admin/circ/board_list.pdf

Dougherty et. Al, “What is Holding Back Productivity Growth in India?”, OECD Journal of

Economic Studies, 2009

Gibbons and Katz, “Layoffs and Lemons”, Journal of Labor Economics 9(4), 199

Hasan, R and Jandoc K, “Labor Regulations and the Firm Size Distribution in Indian

Manufacturing”, Columbian Program on Indian Economic Policies Working Paper No. 2012-3,

January 2012

Ernst & Young and FICCI, Private sector participation in Indian higher education, FICCI Higher

Education Summit, 2011

Farber, H. and Gibbons, R. “Learning and Wage Dunamics”, Quarterly Journal of Economics,

111.4, November 1996

Financial Times MBA rankings 2015, accessed at

http://rankings.ft.com/businessschoolrankings/global-mba-ranking-2015

Government of India, Ministry of Finance, National Economic Survey, 2014-15

Government of India, Ministry of Finance, National Economic Survey, 2015-16

Government of India, Ministry of Labour, Report on Employment-Unemployment Survey, 2013-

14

Government of India, Ministry of Statistics and Program Implementation website, accessed at

https://data.gov.in/catalog/employment-organised-sectors-public-and-private

Hasan, R and Jandoc K, “The Distribution of Firm Size in India: What Can Survey Data Tell us?”,

ADB Economics Working Paper Series No. 213, August 2010

Hausmann, R et. Al, “Doing Growth Diagnostics in Practice: A ‘Mindbook’, CID Working Paper

No. 177, 2008

Hirshleifer et. Al, The impact of vocational training for the unemployed : experimental evidence

from Turkey”, World Bank, 2014

Page 65: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

59

Hsieh, Chang-Tai and Klenow, Peter. “Misallocation and Manufacturing TFP in China and India,”

The Quarterly Journal of Economics, Volume 124 Issue 4, November 2009.

Hsieh, Chang-Tai and Klenow, Peter. “The Life Cycle of Plants in India and Mexico,” The

Quarterly Journal of Economics, Volume 129 Issue 3, August 2014.

Ivy, J. “Higher education institution image: a correspondence analysis approach.” The

International Journal of Educational Management. 15/6, 276-282, 2001

Jindal, N. “AICTE, the Regulator which doesn’t share data!”, Careers 360, 6 Jan 2014

Kanter, Rosabeth (1977): Men and Women of the Corporation, Basic Books, New York.

Kar, Pritchett et. Al, “The Dynamics of Economic Growth: A Visual Handbook of Growth Rates,

Regimes, Transitions and Volatility”, 2013

Kochhar, Kalpana, Utsav Kumar, Raghuram Rajan, Arvind Subramanian, and Ioannis Tokatlidis.

“India’s Pattern of Development: What Happened, What Follows?” Journal of Monetary

Economics 53, no. 5 (2006): 981–1019.

Leeman, N. The Big Test: The Secret History of the American Meritocracy, 2000

Macleod, Reihl, Saavedra and Urquiola. “The Big Sort: College Reputation and Labor Market

Outcomes”, May 2015

Manpower Group, “Annual Talent Shortage Survey”, 2015

Martin et. Al, “In with the Big, Out with the Small: Removing Small-Scale Reservations in India”,

RAND corporation working paper, February 2014

McKinsey & Company, “Education to Employment: Designing a System that Works”, 2012

Mehrotra, S et. Al, “Estimating the Skill Gap on a Realistic Basis for 2022”, Institute of Applied

Manpower Research, Planning Commission, Government of India, 2013

Muralidharan, K & Sundararaman, V, “Teacher Performance Pay: Experimental Evidence from

India”, Journal of Political Economy 119(1), 2011

National Assessment and Accreditation Council (NAAC) Annual Report 2013-14

NASSCOM & McKinsey, “Extending India’s Leadership of the Global IT and BPO Industries”,

2005

NASSCOM, Analysis of Talent Supply and Demand. Employment Requirements and Skill Gaps in

the Indian IT-BPM Industry, 2014

National Assessment and Accreditation Council Annual Report 2013-14

Page 66: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

60

National Sample Survey 61st Round, “Employment and Unemployment Situation Among Social

Groups in India 2004-05”, 2006

National Skill Development Corporation, Skill Gap Reports, 2013

National Skill Development Corporation website, accessed at

http://www.nsdcindia.org/sites/default/files/files/Summary_QP-

NOS_list_as_on_30th_Dec_2015.pdf

Nguyen, T, “Information, Role Models and Perceived Returns to Education: Experimental

Evidence from Madagascar”, 2008

Osterman, P. “Skill, Training, and Work Organization,” Industrial Relations, Vol. 34, No. 2, 1995

Pallais A, Sands, “Why the Referential Treatment? Evidence from Field Experiments on

Referrals.” Journal of Political Economy. Forthcoming.

Pallais A. Inefficient Hiring in Entry-Level Labor Markets. American Economic Review 104(11).

2014

Pandit, V. “Politicians in education”, Education World Online, March 5 2009

Pritchett L, Woolcock M and Andrews M. “Capability Traps? The Mechanisms of Persistent

Implementation Failure”, May 2010

Pritchett L, “The Rebirth of Education: Schooling Ain’t Learning”, Center for Global

Development, October 2013

Pushkar P, “Why India’s universities can’t keep up with the masses”, The Conversation, April 20,

2014

Spence, M, “Signaling in Retrospect and the Informational Structure of Markets”, American

Economic Review 92(3), 2002

Soundararajan, V. “Contract Work and Endogenous Firm Productivity in the Indian Manufacturing

Sector”, 2015

Staffing.org, Staffing Effectiveness & Retention, 2013

Stella, A. “External Quality Assurance in Indian Higher Education: developments of a decade”,

Quality in Higher Education, Vol. 10, No. 2, July 2004

Stella, A. “Institutional Accreditation in India”, International Higher Education, 2015

Thorath and Attewell, “The Legacy of Social Exclusion: A Correspondence Study of Job

Discrimination in India”, Economic and Political Weekly, Vol. 42, No. 41 (2007)

UNESCO, “BRICS: Building Education for the Future”, 2014

Page 67: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue

Sahil Shekhar Putting India to Work

61

United Nations World Population Prospects: The 2015 Revision

United Nations Urbanization Prospects: The 2014 Revision

U.S. Department of Education, “Better Information for Better College Choice & Institutional

Performance”, 2015

Velosa et. Al, “Returns to Higher Education in Chile and Colombia”, IDB Working Paper Series

No. IDB-WP-587, March 2015

Weber, Max (1968): Economy and Society (edited by Guenther Roth and Claus Wittich),

Bedminister Press, New York

World Bank, World Development Report: Jobs, 2013

“AICTE chairman suspended over corruption case”, Indian Express, July 30 2009

“All India Council for Technical Education caught in a cleft”, Business Standard, November 14

2013

“For Political Class, Money Flows via TN Engg Colleges”, The New Indian Express, March 13

2016

“Indian worker strike over Modi labour reforms”, BBC news, 2 September 2015

“Indian universities second-grade?”, The Times of India, Feb 21, 2014

“Kashmir university ‘spices’ hands in order to obtain A-ranking”, Daily Mail, May 10 2012

“More than a degree, NSQF to become mandatory for Government PSY jobs by 2020”, The

Economic Times, April 9 2015

“Political interference has eroded education system, says Arvind Kejriwal”, The Indian Express,

March 27 2015

“PMO steps in to push stalled infrastructure projects”, The Times of India, Dec 30, 2014

“Rethink required to improve higher education: UGC chief”, Business Standard, August 14 2014

“Shock over grading of 38 deemed universities”, Times of India, November 22 2015

“WIPRO interview process selection procedure”, CampusFold.com, 2014, accessed at

http://www.campusfold.com/2014/08/wipro-interview-process-recruitment.html

“The History of the GMAT”, Economist GMAT blog, November 5 2014

“What is the placement procedure for Tata Motors”, Quora, 2015, accessed at

https://www.quora.com/What-is-the-placement-procedure-for-Tata-Motors

Page 68: Cover photo: Job-seekers queue at the Allahabad Employment Exchange Office…research.aspiringminds.com/wp-content/uploads/2016/06/... · 2016-06-21 · Cover photo: Job-seekers queue