Education by niusheng11


									BNU Research Meeting 2
       John Knight
Meeting 2. Introductory lecture on the
 economics of education
    1. motivation
    2. issues
    3. theory
    4. research topics
    5. China
                    1. Motivation
Two questions for you.
1. Guess the author:

‘When an expensive machine is erected, the extraordinary work
   to be performed by it before it is worn out, it must be
   expected, will replace the capital laid out upon it, with at
   least the ordinary profits. A man educated at the expense of
   much labour and time to any of those employments which
   require extraordinary dexterity and skill, may be compared to
   one of those expensive machines.
The work which he learns to perform, it must be expected, over
  and above the usual wages of common labour, will replace to
  him the whole expense of his education, with at least the
  ordinary profits of an equally valuable capital. It must do so
  too, in a reasonable time, regard being had to the very
  uncertain duration of human life, in the same manner as to the
  more certain duration of the machine.
The difference between the wages of skilled labour and those of
  common labour is founded upon this principle’.
[Adam Smith, The Wealth of Nations, 1776,
There we have an excellent statement of the
  theory of human capital!
The pioneering modern book is Human Capital,
  first published by Gary Becker in 1964
2. Guess the country - from a government
‘Workmen are paid through the company shop, rarely in cash.
  They often have no balance in hand to apply to education
Children generally leave school at 8-10 years of age, to go into
  continuous employment. Thereafter, some of them attend
  school on their day off each week.
The workmen support the schools exclusively at their own
  expense, and often they cannot afford to provide textbooks.
  Not having a schoolroom of their own, they have to pay a
  heavy rent for the one they use’.
[The Report of the Commission of Enquiry into
  the State of Children in Employment, 1842 –
  evidence from the sub-commissioners for
  South Wales]
Even in the first industrialising country,
  government-supported mass education does
  not have a long history!
Like the first lecture on migration, this lecture is
  intended to be introductory, and to give you
  helpful background and an overview
The research frontiers of the economics of
  education involve fairly sophisticated micro-
  econometric analyses using sample surveys
The literature is expanding very fast, as more
  data sets become available
A growth industry!
Young people throughout the world, like
  yourselves, are likely still to be in the labour
  force in 2050 and beyond
The 21st century economy promises to be a
  knowledge-based, high-tech economy
What sort of labour force is required for it?
What does that imply for education policies
  today in current ldcs like China?
Growth economists are now giving education
  and knowledge a more central role in the
  growth process
This increases the attention being paid to the
  economics of education
Plausible that education has important positive
                2. The issues
The standard basic questions that are posed in
  the literature:
1. Is the socially optimum amount of resources
    being devoted to education?
Which requires:
2. Can economists provide theoretical criteria
    for expenditure on education?
3. Can the theoretical criteria be translated into
    useful practical criteria?
4. Is there likely to be market failure in the
  provision of education?
5. If so: why, and what can be done about it?
6. What relative emphasis should be placed on
  the different levels and forms of education?
7. How to weigh up the benefit of raising the
  quality of education against the benefit of
  raising the quantity of education?
8. How does education affect the wage structure
  and the distribution of income?
9. What are the trade-offs between reaching
  efficiency objectives and equity objectives in
  educational policies?
There is a growing need for evidence-based
  educational policies
But not an easy task: the pervasive problem of
  establishing causality
Easy to show associations between educational
  variables and economic variables
But what is the causal effect of e.g. education
  on the income or productivity of workers, or
  of household income on child education?
But contrast the economist’s approach with the
  view that (primary and secondary) education
  is a ‘human right’ (e.g., the Millenium
  Development Goals)
The claims made for the human capital
  approach to education:
1. It provides a framework for analysing and
   predicting the private demand for education
2. It provides a framework for assessing the
   social value of expenditure on education
3. In other words, the claim is that it is possible
   to estimate both the private and the social
   returns to education
                 3. Theory
1. The private rate of return
Examine post-compulsory education – where
   there is a choice
Private costs: financial cost + opportunity cost
   (earnings forgone)
Private benefits: consumption good + additional
   lifetime income prospects
Consider e.g. higher education:
Derive the private benefit and the opportunity
  cost from the two income streams from
  entering the labour market after high school
  and entering after a first degree
Discount the private benefit and the private cost
  to the decision point of time, using the
  relevant rate of interest
Prediction that education is demanded if the
  present value (the human capital)> the cost
2. The social rate of return
Conventional to start with the private rate of
  return and to make adjustments
On the cost side, add government subsidies
On the benefit side, assume that the difference
  in average wage between two education levels
  measures the difference in marginal
  productivity, i.e. the marginal product of the
  additional education
This implies profit maximisation and perfect
  competition in markets
The method also implies no statistical screening
  by means of education
Screening and signalling involve a private benefit
  but no (or little) social benefit from the
  educational expenditure
Add any positive (or subtract any negative)
  externalities – if measurable
3. The case for state intervention
The private cost may exceed the social cost of
  educational provision
e.g. credit constraints: human capital cannot
  serve as security
Then a welfare economics case for correcting
  the capital market imperfection at source:
  loans to students
The private benefit may be less than the social
e.g. If there are positive externalities, such as
  knowledge spillovers
Then a welfare economics case for subsidies of
  expenditure on education
           4. Research topics
1. Measuring the causal effect of education on
It is very likely that unobserved ‘ability’ is
     correlated with educational attainment
How to eliminate the effect of this unobserved
• Panel data?
• Instrumenting education
• Matching techniques
• Twin studies
• Measuring human capital and ‘ability’ directly
(i) Panel data?
In a panel survey of adult workers , very few
     people change their education level (and
     they may be unusual)
(ii) Instrumenting
Which variables are well correlated with years of
   education (S) but do not enter the income
Possibly: spouse’s education, distance of home
   from school in childhood, schooling reforms,
Quality of the instruments is crucial
What is the effect of instrumenting?
We expect ‘ability’ to be correlated with S, i.e.
   education appears to do the work that is
   really attributable to unobserved ability
Therefore, we expect the coefficient on S to fall
But it is sometimes found that the coefficient
   rises as a result of instrumenting!
Possible reasons:
1. Measurement error in S causes downward
    attenuation bias
• Instrumenting S corrects for this
2. The returns to schooling may be
• If a school policy intervention (e.g. raising
  minimum school leaving age) is used as the
  instrument, people with highest MR relative
  to MC are most likely to respond
• This can produce an over-estimate of the
  average marginal return to education
(iii) Matching techniques
Matching techniques create a quasi-experiment
Suppose the objective is to estimate the returns
     to higher education
In the survey there is a ‘treatment group ‘(with
     higher education)
Use observed pre-treatment characteristics to
     create a control group
Then match each treated individual with his
  closest match
The matching can be exact (identical observable
Or it can be based on propensity score
  matching, i.e. the estimated probability of
  receiving treatment
Matching is a method of overcoming the
  endogeneity of the education variable
Criticism: whatever the matching technique,
  matching can only be on the observables
(iv) Twin studies
Twin studies are intended to overcome the
   problem of unobserved ability bias by
   standardising for dimensions of ‘ability’
Twins share many things in common, including
   family background
And identical twins have identical genes , e.g. IQ
Sometimes siblings within a family are used
Twins have been much used in Psychology
  research, more recently in Economics research
Use a fixed effects estimator:
Estimate a function to predict the difference in
  earnings between twin pairs
Explanatory variables include their difference in
Their common unobserved variables drop out
Initial studies (using US Twin Festivals) reduced
  the returns to schooling a great deal
More recent studies (e.g. Rouse (EER, 1999)
  show more plausible coefficient s on S:
• Pooled OLS           0.11
• Differenced OLS 0.08
• Differenced IV       0.10
IV (twin’s report of own education) is used to
   overcome downward measurement error
   attenuation bias
Criticism: there may still be omitted ability bias
Is the difference in S truly random?
e.g. Twin differences in birth weight can be
   correlated with differences in IQ and in S,
   differential parental treatment?
(v) Measuring human capital and ‘ability’
Pioneered in East Africa (Knight and Sabot,
Concerned to measure the value of secondary
  education in Kenya and Tanzania
Conducted identical labour force surveys
Used the sample of secondary leavers (S) and
  primary leavers (P)
Included tests of cognitive skill (numeracy and
  literacy) and pure reasoning ability for
  primary- leavers and secondary-leavers
Cognitive skill score = human capital (H)
Reasoning score = natural ability (A)
• Earnings functions (predicting ln Y) , including
  S and P, with and without H and A
• Educational attainment functions (dependent
  variable S or P) , including A, etc.
• Educational production functions (predicting
  H) using A and S, etc.
Main conclusions:
H is more important than S in the earnings
  function; the (small) influence of S might
  represent screening
A is not directly important in the earnings
  function but has an indirect influence via the
  educational attainment and educational
  production functions
(vi) Two methods of estimating returns
1. The conventional method, described above :
    cost-benefit analysis, sometimes applied to
    an educational level, important for
    estimating the private and the social rate for
    policy purposes
2. The short method, useful for showing how
    the returns to S vary over time, or across
    ownership sectors, or across regions, etc:
Based on:
       ln Ys = ln Yo + rS + …
Ln income with S years of schooling equals ln
   income with 0 years of schooling plus S
   multiplied by the rate of return on a year of
i.e. the coefficient on S in the income function
   shows the rate of return (e.g. 0.15 implies 15%
However, there can be non-linearities in the
  effect of S on income (Y), suggesting
     ln Y = …. + αS + βS2 + …
• The returns are sometimes found to be
  concave (α > 0, β < 0)
• But sometimes convex (β > 0), i.e. higher
  returns at higher levels of education
• Possibly because of educational rationing at
  the higher levels
2. Measuring Educational Externalities
A benefit or cost accruing to others in society
  which is not taken into account by the
  decision-maker, i.e. not internalised
The conventional adjustment: exclude direct
  taxes when estimating the private rate of
  return but include them when estimating the
  social rate of return
Positive externalities in the workplace:
The more educated may raise the productivity of
   the less educated with whom they come into
i.e. they create externalities which are not (fully)
   internalised by employers
Positive externalities in the village:
Less educated farmers learn from more
   educated farmers
e.g. they adopt new methods or use existing
  technologies more efficiently
Normally the social marginal product of
  education is measured by calculating the
  difference in income by education level
That difference understates the true benefit
The externality should be added, not
Knowledge as a public good:
Some endogenous growth models assume that
  knowledge generates externalities
Other workers, other firms, future generations
  can gain from it
And some empirical evidence from cross-
  country growth regressions that both growth
  dH/H) and stock (H) of human capital raise the
  growth rate of output
Different economies have different production
In the poorest economies, it is the growth and
  stock of primary education that is important
In intermediate economies, …secondary
In advanced economies, …higher education
Mc Mahon (1999) argues that interaction
   between (from and to) education and several
   other aspects of development generates
   positive feedbacks and dynamic processes of
i.e. the causes and consequences of education
   are more complex than conventional analysis
For instance:
1. Education increases the incentive to invest in
   physical capital
2. Education can lower fertility decisions
3. It can improve child health
4. Educational expansion can reduce income
• Possible effects on civil society, government
  accountability, etc.
These can be linked in interacting ways with
  education, and sometimes with long time lags
The implication is that the social benefits of
  education are likely to be under-estimated
Case study of externalities in farming:
Ethiopia has a very poor, rural economy with
  very little education
We added a special education survey to a sub-
  sample of a national household survey
Does education generate production
Weir and Knight (JAE 2007) estimated ‘average’
  and ‘stochastic’ production functions
Included not only average adult education of the
   household (E h) but also average adult
   education of the village (E v)
Is the coefficient on Ev > 0?
The methodological problem: village-level
   education may be positively correlated with
   unobserved village characteristics, e.g. soil
   fertility, local climate
Solution: include village dummies (to pick up
  village fixed effects), and also average
  education of the most educated household in
  the neighbourhood in these (spread-out)
Coefficient on own education: 0.044
Coefficient on neighbourhood education: 0.034
Evidence of a production externality
Dynamic use of the same data - Weir and Knight
  (EDCC, 2004):
Education was found to have no effect on the
  probability of adopting an innovation. Why?
Timing and social networks are crucial
We found that educated farmers tend to be
  early innovators in a particular area
Once an innovation (e.g. use of fertiliser) has
 been tried and the results are obvious to
 others, neighbours imitate: social learning
An argument for new policies to raise the low
 enrolment rate and extend years of education
 in rural Ethiopia
3. Impact evaluation analysis
A recent development in education research is
  randomised trials to evaluate the impact of
  particular policies
Divide a sample into two groups, the treatment
  group and the control group
Individuals must be randomly assigned
Then introduce a policy intervention, e.g. school
  meals, lower pupil-teacher ratio, etc.
Apply the difference in difference method of
1. Experiments can be expensive
2. Results may not apply elsewhere
3. Assignments and attrition may non-random
4. Educational access
Why study who gets education, how much
   education, and why?
In a perfectly competitive economy, the rate of
   return is everywhere equal to the rate of
Lifetime income is independent of how much
   education one receives
But not likely to be the case, especially in ldcs
More education is likely to raise lifetime income
This raises issues of equity and of efficiency
• Are the criteria by which people are allocated
  higher lifetime income fair?
• Are the people who can use education most
  productively the ones who receive it?
Individual heterogeneity in marginal benefit and
  marginal cost of education can cause people
  to choose different amounts of education:
Different benefits:
• Ability - affecting enjoyment of education and
  success in a competition for places
• Family background – affecting attitudes and
  out-of-school learning
• Location – affecting the nature of the
  production function in a rural context (maybe
  not if there is the prospect of migration)
• School quality – often limited choice, so low
  quality can deter enrolment
Different costs:
• Income - in principle, income should be
  irrelevant: the rate of return determines
  demand; but credit constraints
• Discount rate – depends on availability of
  liquid funds (lender’s rate), or possible sources
  of credit (borrower’s rate) ,or the rate at
  which the household discounts future
• Locality – transport costs, extent of
  community subsidy
• User charges – affecting both the private
  return and the funding problem
Estimation of educational attainment or
  enrolment functions, e.g. a binary logit
  enrolment model:
Explanatory variables:
1. Individual - child’s age, sex, health, position
   in the order of siblings
2. Community – local availability of school,
   distance to school, school fee level , proxies
   for quality of school
3. Household – household income, and liquid
    assets/debts, number of siblings
4. Parents – education of father, education of
    mother, sex of household head
Intra-household resource allocation:
Examine the proportion of household
  expenditure on education according to the
  number and gender of children (ideally, but
  not necessarily, using information on
  expenditure on each child)
Possible hypotheses:
• Households favour boys?
• Female household heads spend more on
• Female household heads are fairer to girls?
Possible reasons for gender discrimination:
Benefit side:
• Wage or job discrimination in the labour
  market reduces the return to girls (but is
  discrimination greater for the educated or the
         5. Education in China
1. Some economic history
Very little education in China 60 years ago
With educational expansion, China created an
   unusual educational pyramid, in relation to
   countries at the same level of development:
Higher enrolment rate ate primary level
Average at secondary level
Well below average at higher level
In a market economy, this would have produced
  high returns to post-primary education
But an egalitarian wage structure was imposed:
‘Brain workers’ received no more than ‘hand
  workers’ under central planning
Nevertheless, there is evidence that post -
  compulsory educational enrolments have
  been rationed
As with income, there is a huge rural-urban
  divide in education (Knight and Li, OBES, 1996)
Reflecting separate administrative and funding
Urban education is centrally, and well, funded
Rural education depends heavily on local
This affects both the quantity and the quality of
  rural education
2. Some promising research topics
(i) The returns to education
(ii) Higher education
(iii) Access to education: intergenerational
(iv) Poverty traps
(i) The returns to education
The Chinese labour market has been changing in
   two ways:
Movement towards market equilibria
Movement of market equilibria
They need to be distinguished
Rising returns in China could be due to either or
Over time the egalitarian wage structure has
  been eroded by market forces
But only slowly, because erosion requires labour
  mobility between jobs
The returns to education have risen for that
This should be more marked in the more
  competitive parts of the labour market
• More marketised regions
• More marketised ownership categories
The pattern of returns (short method) is worth
  exploring, over time and across regions,
  ownership categories, labour market
  segments, etc.
Need to estimate the causal effect of education
  on income
• Panel data, instrumenting, matching
  techniques, twin studies
(ii) Higher education
A ‘natural experiment’: the sharp change in
   higher education policy
Expansion of higher education enrolment
   starting in 1998
Higher education enrolment rose by almost
   400% between 1997 and 2005 (3.2m to
The effect on the labour market would have
  been felt from about 2002 onwards
How did the labour market adjust?
e.g. unemployment, filtering down into lesser
  jobs, fall in market wages and in rate of return
  to higher education
(iii) Access to education: intergenerational
In every country, there is inequality of access to
   post-compulsory education
e.g. in Britain, huge differences in access to
   higher education based on social class of
   parents – even when higher education was
   fully funded by government
Also true of e.g. East Africa: parental education
   is very important for access to secondary and
   higher education (Knight and Sabot, 1990)
Hence a lack of intergenerational mobility
i.e. the children of parents with relatively low
   education also have relatively low education
What about China?
Need for data on the education of two, ideally
  three, generations
Need to standardise for other determinants of
  education, e.g. household income (at time of
  education), location, gender, etc
The possible consolidating role of ‘assortative
  mating’ , i.e. the educated marry the educated
(iv) Education and poverty traps
The possibility that education and income are
   inter-related so as to produce a ‘poverty trap’:
Education can raise income, but low income
   prevents access to education
Two papers on education and the poverty trap
   in rural China by Knight, Li and Deng (ODS,
   2009, 2010)
They set up a model with 17 different
  relationships to form a system
Involving quantity and quality of education,
  income, community income, community
  enrolment, parental income, parental
  education, health, subjective well-being, etc.
Each of these relationships deserves to be
  studied in more detail

To top