The Decision to take up Adult Education: a Dynamic Programming Analysis

1
The Decision to take up Adult Education a
Dynamic Programming Analysis

• Justin van de Ven and Martin Weale
• National Institute of Economic and Social Research

2

• Why dont more people participate in life-long
  learning?
• Educational upgrading leads to higher pay and
  better employment prospects.
• We consider men upgrading from level 2
  (qualifications above GSCE) to level 4 (higher
  education).

3
The Low Prevalence of Lifelong Learning
BHPS data Initial Qualifications Initial Qualifications
Upgrade Level 2 Level 3 Total
Pooled Panel Data
No 1,577 2,815 4,392
Yes 82 188 270
Total 1,659 3,003 4,662
Percentage upgraded 4.90 6.20 5.80
Individual Time Series
No 232 382 614
Yes 11 31 42
Total 243 413 656
Percentage upgraded 4.50 7.50 6.40
Probability upgrade over life-time 5.40 18.60
4
Earnings are Unpredictable
Qualification Level Median Mean St. Dev Number
2 385 419 179 1577
2 with Upgrade 514 519 160 82
3 445 493 229 2815
3 with Upgrade 479 512 238 188
Full Sample 425 469 216 4662

Are the uncertain benefits from upgrading a
deterrent?
5

• The increase in earnings associated with
  upgrading is not statistically significant at a
  5 level.
• But the uncertainty surrounding the effects of
  upgrading suggest that 90 confidence limits for
  its individual effects in multiplying up earnings
  are 0.52 to 1.92.
• A high chance of upgrading going wrong.

6
A Structural Analysis (i)

• Welfare comes from consumption and leisure
• Assume people maximise their expected life-time
  welfare but are risk-averse
• 1 is better than equal chances of getting 50p or
  1.50.
• Life-long learning has costs in terms of fees and
  leisure time given up.
• It yields uncertainty benefits in terms of better
  pay and employment prospects.
• Is this uncertainty enough of a deterrent.

7
A Structural Analysis (ii)

• Choose parameters to define welfare function so
  as to match observed consumption and labour
  supply decisions over the life-course as closely
  as possible.
• Implies a reasonably strong degree of
  risk-aversion.
• Earnings rise sharply for people in their 20s and
  more slowly to about 45 then fall off
• But there is a strong random element.

8
A Structural Analysis (iii)

• Simulate a panel of 10000 people moving through
  the life course.
• At each age they can decide how much to spend and
  whether to work full-time, part-time or not at
  all.
• They can also decide whether to upgrade their
  qualifications or not. Upgrading is possible only
  once.
• Observe the number of people deciding to upgrade
  by age.

9
Costs of Education

• University fees introduced at 1000 in 1998/9
  currently 9000
• Many higher education sub-degree courses are much
  cheaper and take one or two years.
• Full-time education makes full-time work
  impossible but does allow part-time work.
• Part-time study is compatible with full-time work
  but leaves little spare time.
• Assumed two years of full-time study or three
  years of part-time study.

10
Drop-out Rates

• 15-17 of mature students drop out after the
  first year
• Resumption rates are under 20
• 37 of part-time students under 30 dropped out
  and 34 of those aged 30
• No information on failure we explore the role of
  a subjective risk of failure.

11
Simulated Life-time Probabilities of Upgrading
from Level 2 to Level 4
Risk of Failure Risk of Failure Risk of Failure Risk of Failure Risk of Failure
Cost of full-time course ( p.a.) 0 20 40 60 80
0 81 75 64 46 11
500 80 73 61 43 8
1000 79 71 60 39 5
3000 73 64 51 28 0
9000 59 45 27 1 0
The observed probability is 5
12
Simulated Life-time Probability of Upgrading from
Level 3 to Level 4
Risk of Failure Risk of Failure Risk of Failure Risk of Failure Risk of Failure
Cost of full-time course ( p.a.) 0 20 40 60 80
0 71 62 52 37 7
500 69 60 50 34 5
1000 67 59 48 31 3
3000 62 54 42 23 0
9000 50 39 25 3 0

The observed probability of upgrading is 18
13
The Simulated Participation Profile
60 chance of failure fees of 1000 p.a.
14
Conclusions

• Risk is a factor affecting the take-up of
  life-long learning.
• But both uncertainty about earnings and fees go
  only a part of the way to explaining low
  participation.
• High subjective risks of failure are needed to
  account for observed participation rates.