Docstoc

Introduction to the economics of education

Document Sample
Introduction to the economics of education Powered By Docstoc
					Higher education and the
 distribution of income
          Craig Holmes
   Higher Education II seminar
       19th October 2012



         www.skope.ox.ac.uk
             Seminar outline
• Aims:
  – Look at how economists have tried to measure to
    private returns to education, including HE
  – Discuss some of the methodological issues
  – Consider the factors which drive changes in the
    distribution of earnings of graduates over time




                  www.skope.ox.ac.uk
   Higher education and earnings
• In week 1, we discussed links between
  education and training and higher wages:
  – Human capital theory: education  higher
    productivity  higher wages through labour
    market competition
  – Signalling: education  signal of productivity 
    high wages through separation from less able
    workers


                   www.skope.ox.ac.uk
   Higher education and earnings
• Average wages of graduates vs. non-graduates:
                   1200
     Earnings, £

                   1000



                    800



                    600



                    400


                                                                          Graduate
                    200
                                                                          Non-graduate

                      0
                          20-24   25-29   30-34   35-39   40-44   45-49    50-54     55-59



      Source: LFS 2008
                                    www.skope.ox.ac.uk
   Higher education and earnings
• Male wages, graduates vs. non-graduates:
                   1200
     Earnings, £

                   1000



                    800



                    600



                    400


                                                                                     Graduate
                    200
                                                                                     Non-graduate

                      0
                          20-24   25-29     30-34   35-39   40-44   45-49   50-54   55-59   60-64


      Source: LFS 2008
                                          www.skope.ox.ac.uk
   Higher education and earnings
• Female wages, graduates vs. non-graduates:
                    1200
      Earnings, £

                    1000



                     800



                     600



                     400


                                                                            Graduate
                     200
                                                                            Non-graduate

                       0
                           20-24   25-29   30-34   35-39   40-44   45-49   50-54       55-59



      Source: LFS 2008
                                      www.skope.ox.ac.uk
Higher education and earnings




 http://www.oecd.org/edu/highereducationandadultlearning/48630790.pdf


                           www.skope.ox.ac.uk
   Higher education and earnings
• Income benefits to education not solely about
  higher wages
• More educated people:
  – Are less likely to be unemployed
  – Are more likely to find a job whilst unemployed
  – Are more likely to receive further training and
    education
  – Are more likely to progress to better jobs

                   www.skope.ox.ac.uk
             Graduate earnings




http://www.oecd-ilibrary.org/education/what-are-the-returns-on-higher-education-for-
   individuals-and-countries_5k961l69d8tg-en

                           www.skope.ox.ac.uk
   Higher education and earnings
• Average earnings do not take into account all
  the differences between two groups
• Key question: what is the marginal effect of
  investing in education on earnings?
• Economists use wage regressions to calculate
  this (Mincer, 1974)



                  www.skope.ox.ac.uk
     Introduction to regression
Y




– Data sample:                           X

    • Y – dependent variable
    • X – explanatory or independent variable
                    www.skope.ox.ac.uk
     Introduction to regression
Y


                                                b

                                      1




    a


– Estimate using OLS:                       X

    • Yi= a + bXi + errori

                       www.skope.ox.ac.uk
     Introduction to regression
Y
 a+bXj
                                                  errorj




    a
                                             Xj
                                         X

– OLS finds a and b to minimise sum of errors2

                    www.skope.ox.ac.uk
             Introduction to regression
       Y



Ex(Y|Xi)=a+bXi




            a

                                          Xi     X

       – Assumption is that for a given X, Y is distributed normally
         with mean value a + bX
                            www.skope.ox.ac.uk
       Introduction to regression
• When there are many potential explanatory variables
  we use multivariate regression (still using OLS)
   – Y = a + bX + cZ + dW + ... + errors
   – Each coefficient captures the partial correlation between
     the explanatory variables and Y, holding everything else
     constant
• Linear form is convenient, but real life may be more
  complex
   – Missing variables
   – Interactions

                         www.skope.ox.ac.uk
    Returns to higher education
• Wage premia and rates of return conflated
• Internal rate of return:
  – NPV of E = (w0 – c0)+ δ(w1 – c1)+ δ2(w2 – c2)+...
  – w and c are wage increases and costs of acquiring
    education level E
  – δ is discount factor – NPV = 0  δ = 1/(1+IRR)
• Wage premia:
  – Log wage = a + b1.E + b2.exp+ b3.E.exp
  – Under certain assumptions, b1= IRR
                   www.skope.ox.ac.uk
     Returns to higher education
• BIS (2011):
  – Labour Force Survey 1996-2009
  – Sample:
     • Undergraduate or equivalent vs. 2 ‘A’-levels
     • Masters or doctorate vs. undergraduate
  – Controlled for age, family, marital status, gender,
    ethnic background, region and year
  – Breakdown by subject and degree class
  – Max vocational qualification at level 3
                      www.skope.ox.ac.uk
     Returns to higher education
• BIS (2011):




                www.skope.ox.ac.uk
     Returns to higher education
• BIS (2011):
  – Average return: 27.4% (29.7% and 23.5% for
    women and men respectively)
  – Highest returns: medicine (82.8%), maths and
    computer science (41.1%) and law (41.2%)
  – Lowest returns: creative arts (6.3%), mass
    communications (8.4%) and history and
    philosophy (10.4%) – only significant for women.
  – First class degree (32.7%) vs. Lower second
    (21.3%)
                   www.skope.ox.ac.uk
   Higher education and earnings
• Questions:
  – Does completing a degree make a worker more
    productive or are those who get a degree, on
    average, more productive?
  – Does a degree act as a labour market signal to
    employers?




                   www.skope.ox.ac.uk
    Measuring returns to education
     Wage




Δw = b ΔE




                           E1         E2
     – Estimate:                               Education

        • Log wagei = a + b.Ei + d.Xi + errors

                          www.skope.ox.ac.uk
Measuring returns to education
Wage                                                  High ability
                                                  Estimated

   b                                                  Low ability


   b




                      E1         E2
– Estimate:                               Education

   • Log wagei = a + b.Ei + d.Xi + errorsi

                     www.skope.ox.ac.uk
  Measuring returns to education
• Selectivity bias (or ‘ability’ bias):
   – Observed data reflects differences in
     unobservable characteristics
   – Unobservable characteristics are correlated with
     variable of interest
      • Higher ability workers choose more education
      • Some of estimated wage return to education is due to
        differences in ability
• This may, or may not, reflecting signalling
                      www.skope.ox.ac.uk
Measuring returns to education
                                          Estimated
Wage
                                             High ability

  b
                                              Low ability
  b'
  b




                   E1         E2       Education

  • Suppose education and ability are correlated
  • Estimated b > true b
                  www.skope.ox.ac.uk
  Measuring returns to education
• Thompson (2012) – BIS sample comparison
    12%
                  HE
                  Not HE
     8%


     4%


     0%
          120   160   200    240    280    320   360
                         Tariff Points
                      www.skope.ox.ac.uk
  Measuring returns to education
• Signalling vs. ability bias
   – Signalling could cause ability bias to be present in
     OLS estimations of earnings data
   – However, ability bias might exist even if signalling
     is not taking place – employers may be better at
     observing productivity differences than
     econometricians!



                     www.skope.ox.ac.uk
  Measuring returns to education
• Huge literature on dealing with ability bias
  – Twin studies
  – Instrumental variables
  – Natural experiments
• See Card (1999) for attempts to deal with
  ability bias (mostly in schooling, not HE)
• Two HE approaches:
  – Use test scores as proxies for pre-college ability
  – Look at non-completing students
                    www.skope.ox.ac.uk
     Returns to higher education
• Blundell, Dearden and Sianesi (2000):
  – National Child Development Study (1958 cohort)
  – Degree holders compared to ‘A’-levels
  – Wages at age 33
  – Uses school tests (‘A’-levels scores, maths and
    reading aged 7) to proxy for ability




                   www.skope.ox.ac.uk
     Returns to higher education
• Blundell, Dearden and Sianesi (2000):
  – ‘Raw’ returns to undergraduate degree 21% (men)
    and 39% (women)
  – Reduced slightly by controlling for ability
  – Reduced further by controlling for job
    characteristics
  – Full model returns: 12% (men) and 34% (women)



                  www.skope.ox.ac.uk
  Measuring returns to education
• Other issues:
  – Highest qualification or all qualifications?
  – Include occupations?




                    www.skope.ox.ac.uk
  Measuring returns to education
• Exercise:
                 Apprenticeship    No apprenticeship
     Degree      +12%              +10%
     No degree   +5%               Reference group


  – What would happen if you estimated:
    wages = a + b.DEGREE + c.APPRENTICE
     • Would b = 10% and c = 5%?
     • How could you solve this?
     • What practical problems would you encounter?
                        www.skope.ox.ac.uk
  Measuring returns to education
• Occupation and/or industry variables are
  sometimes included in wage equations
• May potentially proxy for unobserved ability
  differences
• However, may also create additional
  selectivity biases (due to unobserved specific
  ability)
  – People self-select into jobs they earn highest at

                    www.skope.ox.ac.uk
    Labour market for graduates
• Most earnings studies offer a snapshot of the
  historical link between (higher) education and
  wages
• Policymakers also need to be forward looking:
  – What are the trends?
  – What are the effects of policy options?
• This will largely depend on the demand for
  skills

                   www.skope.ox.ac.uk
      Labour market for graduates
• Wage premia reflect supply and demand in the
  labour market
  – Assume a competitive labour market
  – Firms demand workers of particular skill until MR = MC
  – MC = wage rate
  – In our week 1 example, MR of each skilled worker was 200.
    Competition led to wage = 200.
  – In real life, MR decreases as more workers are employed
  – Wage set to clear market
   Total demand (sum of all firm demands) = total supply

                     www.skope.ox.ac.uk
       Labour market for graduates

         Firm                        Labour market



Wage                 Wage                            Supply




   w


                                                     Demand
                D=MR

                 L                             L

                www.skope.ox.ac.uk
       Labour market for graduates

Wage
                             Supply
                                          New supply



W**
W*

                                          New demand

                                     Demand

                                      L

                www.skope.ox.ac.uk
    Labour market for graduates
• We saw in week 1 that the supply of graduates
  in UK has increased dramatically since 1989
• What has happened to demand?




                 www.skope.ox.ac.uk
     Labour market for graduates
• Technology is often acknowledged as being a key
  driver of the demand for skill:
     Technical advances since the 1980s have been the main driver in
     helping workers become more productive. This has been strongly
     biased towards those with the skills to adapt and use new technology.
     As a result, more highly skilled workers are in increasing demand by
     employers (‘Skills For Sustainable Growth’ pg. 9, BIS 2010)
• Skill-biased technical change
• Task-biased technical change (Autor, Levy and
  Murnane, 2003; Goos and Manning, 2007)
   – Growth in non-routine jobs – some of which are low skill

                          www.skope.ox.ac.uk
     Changing returns to higher
            education
• Walker and Zhu (2008):
  – LFS 1994-2006:
  – No drop in average male wage premium
  – Small rise in average female wage premium
  – Some evidence of drop-off at bottom of
    distribution




                  www.skope.ox.ac.uk
      Changing returns to higher
             education
• Bratti, Naylor and Smith (2008):
  – Similar methodology to Blundell et al (2000)
  – Uses British Cohort Study (1970 cohort)
  – Male premia has remained same as in NCDS study
    (15%)
  – Female premia has fallen sharply (18%)




                  www.skope.ox.ac.uk
      Changing returns to higher
             education
• Ireland et al(2009):
  – Changing graduate premia reflect
     • Change in return to education
     • Change in return to unobserved ability
     • Changing ability composition of graduates
  – Hence, expansion of HE can have effect on
    signalling role of degrees



                     www.skope.ox.ac.uk
     Labour market for graduates
• There is concern about skill utilisation
   – Will return to this theme is weeks 4-7.
• This may have distributional, not average, effects
• Green and Zhou (2010):
                                   Skill utilisation
                                   Utilised                     Under-utilised
       Qualification   Qualified   Graduate job                  Degree a sorting
       requirements                                              mechanism
                                                                 (retrospective
                                                                 graduatisation)

                       Over-       Employees use graduate        Non-graduate job
                       qualified   skills in non graduate job




                                      www.skope.ox.ac.uk
     Labour market for graduates
• Green and Zhou (2010):




                   www.skope.ox.ac.uk
     Changing returns to higher
            education
• “The Global Auction” – Brown, Lauder and
  Ashton (2011):
  – Four key themes
     •   Mass expansion of higher education
     •   Quality-cost revolution
     •   Digital Taylorism
     •   The War for Talent




                       www.skope.ox.ac.uk
                                 Changing returns to higher
                                        education
• Suggests graduate jobs are segmenting
• No 4.0% thing as “the graduate premium”?
     such
  Change in graduate premium,




                                3.5%
                                3.0%
                                2.5%
           1987-2001




                                2.0%
                                1.5%
                                1.0%
                                0.5%
                                0.0%
                           -0.5% 0.00          0.20        0.40         0.60            0.80             1.00

                           -1.0%
                                                                   Percentile of earnings distribution
                                Source: Holmes and Mayhew (2011)
                                                      www.skope.ox.ac.uk
                            Exercises
• The last section of this lecture focused heavily on the UK
• How do you think the issues discussed apply to other
  countries? To consider:
   –   The size of the HE sector
   –   The skills focus of the HE sector
   –   Transitions into work
   –   Labour market trends that may drive earnings




                            www.skope.ox.ac.uk

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:6/17/2013
language:
pages:46