Professional Documents
Culture Documents
www.elsevier.com/locate/econedurev
Abstract
Individual absolute risk aversion is measured in a sample of 1583 male house-hold heads, using the data drawn from
the 1995 wave of the Survey on the Income and Wealth of Italian households. This measure, conditional on household
financial wealth, is used as an instrument for attained education in a standard log earnings equation. In line with most
of the literature, I find that the gap between the IV and the OLS estimates of the returns to education is large.
2002 Elsevier Science Ltd. All rights reserved.
1. Introduction (2000) for an attempt). I use the 1995 wave of the Survey
on the Income and Wealth of Italian households and pre-
It is well known that the estimation of the returns to vious work on these data by Guiso and Paiella (2000) to
education is difficult because of the presence of measure- measure individual absolute risk aversion in a sample of
ment errors and because unobserved ability can affect 1583 married Italian male household heads. This variable
both educational choice and the returns to education. One is then used as an instrument, for education in a standard
of the strategies used to deal with this problems consists Mincerian earnings function. In line with most of the
of selecting instrumental variables, that are correlated current literature, I find that the gap between IV and OLS
with schooling but not with earnings (conditional on estimates is substantial.
schooling). The typical instruments used in the literature
are school reforms, family background variables and
smoking. An alternative is to use data on twins. See Card 2. Schooling choice
(1999) for a review of the existing evidence. In this note
I add to the current list an additional candidate, the absol- Following Card (1999) I assume that an individual
ute degree of risk aversion. I start by showing in a simple chooses S, the years of schooling, by maximizing the
static model that risk aversion affects in a natural way following objective function1
educational choice by influencing the marginal utility of
U(y)⫺f(S) (1)
schooling. Perhaps one reason why this variable has not
been used so far is that it is difficult to measure risk where y is (hourly) earnings, U is a concave function of
aversion in survey data (see Barsky, Juster, Kimball, & y and f is a convex function of S. Hourly earnings are
Shapiro (1997) and Hartog, Ferrer i Carbonell, & Jonker related to S by the following function
0272-7757/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 7 2 - 7 7 5 7 ( 0 1 ) 0 0 0 6 2 - 0
636 G. Brunello / Economics of Education Review 21 (2002) 635–640
y ⫽ g(S) ⫽ elS (2) If college wages varied more than, say, high school
wages, there is an additional reason why risk aversion
In this setup, additional schooling increases earnings
can influence educational attainment, because more risk
and utility U at the price of higher investment costs f.
averse individuals will require higher expected earnings
The first order condition associated to the maximization
to invest in college education.
of Eq. (1) is
U⬘g⬘(S) ⫽ f⬘(S) (3)
where the prime if for the first order derivative.2 3. The empirical model
Using a first order Taylor approximation of U⬘ around
y=0, Eq. (3) can be re-written as follows Consider the standard regression model
U⬘(0)[1⫺ARAg(S)]g⬘(S) ⫽ f⬘(S) (4) ln y ⫽ X⬘d ⫹ aS ⫹ e (7)
U⬙(y) S ⫽ X⬘b ⫹ Z⬘g ⫹ h (8)
where ARA ⫽ ⫺ is the Arrow Pratt coefficient of
U⬘(y)
absolute risk aversion (see Laffont (1990)). where X is a vector of controls that affect the marginal
I can now establish the following benefits of education, e and h are error terms, Eq. (7) is
the Mincerian earnings function and Eq. (8) is the attain-
Claim 1 The selected years of schooling S decrease ment function, that depends on X and on variables that
when absolute risk aversion ARA increases. affect marginal costs and capture individual prefer-
Proof. Differentiation of Eq. (4) with respect to ARA ences (Z).
and S yields Ordinary least squares (OLS) estimates of Eq. (7)
再 冎
yield a consistent estimate of a only when e and h are
f⬙(S)
g⬙(S)[1⫺ARAg(S)]⫺ARAg⬘(S)2⫺ ∂S uncorrelated. Unobserved ability and measurement errors
U⬘(0) are two well known factors that affect both schooling S
⫽ g(S)g⬘∂ARA and earnings y conditional on schooling, thereby
inducing correlation between the error terms. This prob-
The expression within curly brackets on the left hand lem can be addressed if one can identify variables that
side is negative because of the second order conditions affect schooling but not (conditional) earnings. These
for a maximum. The right hand side is positive. variables can be used as instruments (IV) to generate
Let the marginal cost of an additional year of school- consistent estimates of the returns to education.
ing, f⬘(S), be equal to r1 ⫹ r2S. Next, assume that the Card (1999) presents a detailed review of previous
utility function U(y) belong to the CARA (constant studies based on instrumental variables and discusses the
absolute risk aversion) class validity of the instruments used in each study. Briefly,
1 these instruments include school reforms and features of
U(y) ⫽ ⫺ exp(⫺sy) (5) the school system, family background and the use of
s
samples of twins. Another instrument recently used but
In this case the coefficient ARA is equal to s, not discussed by Card is smoking. The argument here is
U⬘(0) ⫽ 1 and optimal schooling S* is given by3 that smoking habits are likely to be highly correlated
with the discount rate, but do not influence earnings
S∗ ⫽ S(l,r1,r2,s) (6) directly.4 Therefore, they can be used as a valid instru-
Individual differences in educational attainment can be ment for schooling S.
explained in this simple model both by differences in The simple model presented in the previous section
marginal returns l and marginal costs r and by differ- suggests that a measure of individual absolute risk aver-
ences in the absolute degree of risk aversion. sion ARA is another potential candidate. In the model,
The expected variation of earnings should also matter. the variable ARA affects the schooling decision because
In a slight complication of the model, I can add to Eq. it affects the marginal utility of income (and
(2) the stochastic term n and assume that log earnings consumption), but does not affect the marginal returns
are normally distributed with mean lS and variance sν2. to schooling l.
2
The sufficient condition for an interior maximum is
U⬙g⬘(S)2 ⫹ U⬘g⬙(S)⫺f⬙(S) ⬍ 0 where the two primes are for
the second derivative.
3 4
Card uses the CRRA (constant relative risk aversion) util- See Fersterer and Winter-Ebmer (2000) for a recent dis-
ity function U(y) ⫽ ln y. cussion.
G. Brunello / Economics of Education Review 21 (2002) 635–640 637
a 8
Note: Robust standard errors. All the coefficients are multi- These categories are: blue collars, white collars and mana-
plied by 1000. The regression includes a constant term. gerial employees in the three sectors and school teachers in the
public sector.
9
I use age rather than potential experience because the for-
7
I exclude household net income, because an important part mer variable can be treated as exogenous in the earnings
of it is the household head’s personal net income. The inclusion regression. See Harmon and Walker (1995). There are 18
of this variable would remove part of the correlation between regional dummies and 3 urban residence dummies, depending
risk aversion and log earnings. At the extreme, the instrument on the size of the town of residence.
10
would lose its power completely. The p-values are 0.14 and 0.28 respectively.
G. Brunello / Economics of Education Review 21 (2002) 635–640 639
Table 5
OLS estimate of Eq. (7). Dependent variable: ln ya
OLS IV
Coefficient P-value Coefficient P-value
a
Note: robust standard errors within parentheses. The regression includes also a constant, marital status, regional and urban
residence dummies.
640 G. Brunello / Economics of Education Review 21 (2002) 635–640
effective government control of criminal activities, is and Guglielmo Weber for advice. The usual disclaimer
likely to affect more Southern households, who live in applies.
crime intensive regions, are less educated and have
higher marginal returns to education, and to produce as
a consequence a higher IV estimate of the average References
returns to schooling.
Barsky, R., Juster, T., Kimball, M., & Shapiro, M. (1997). Pref-
erence parameters and behavioural heterogeneity: an experi-
6. Summary mental approach in the health and retirement study. The
Quarterly Journal of Economics, 112(2), 537–580.
In this note I have added to the current list of instru- Card, D. (1999). The causal relationship of education on earn-
ments for endogenous schooling in an earnings ings. In O. Ashenfelter, & D. Card (Eds.), Handbook of
regression a new entry, the absolute degree of risk aver- labor economics, vol. 3 (pp. 1801–1863). Amsterdam:
sion. I have used the 1995 wave of the Survey on the North-Holland.
Income and Wealth of Italian households to measure this Fersterer, J., & Winter Ebmer, R. (2000). Smoking, discount
variable in a sample of 1583 married Italian male house- rates and returns to education. IZA Discussion Paper n.
hold heads. In line with an important part of the relevant 126.
Guiso, L., & Paiella, M. (2000). Risk aversion and financial
literature, I have found that the IV estimate of the mar-
market imperfections, mimeo, Ente Einaudi, Rome.
ginal return to schooling is much higher than the OLS Harmon, C., & Walker, I. (1995). Estimates of the economic
estimate. returns to schooling for the United Kingdom. American
Economic Review, 85, 1278–1286.
Hartog, J., Ferrer i Carbonell, A., & Jonker, N. (2000). On a
Acknowledgements simple survey measure of individual risk aversion. Univer-
sity of Amsterdam.
I am grateful to two anonymous referees, to Daniele Laffont, J. J. (1990). The economics of uncertainty and infor-
Checchi, Luigi Guiso, Tullio Jappelli, Claudio Lucifora mation. Cambridge, MA: MIT Press.