Professional Documents
Culture Documents
Fiscal Policy and Educational Attainment in The United States: A Generational Accounting Perspective
Fiscal Policy and Educational Attainment in The United States: A Generational Accounting Perspective
doi:10.1111/j.1468-0335.2006.00546.x
INTRODUCTION
Since the seminal studies by Auerbach, Gokhale and Kotlikoff (1991; 1994)
and Kotlikoff (1992), generational accounting has usually been perceived as a
meaningful way to evaluating the sustainability of fiscal policy. It builds on an
intertemporal treatment of the government budget constraint: at any date, the
present value of government purchases must be covered by the current net
public wealth, the present value of the net taxes that will be paid by living
people over the rest of their lifetimes, and the present value of the net taxes that
will be paid by future generations over their whole lifetimes. The basic
questions are: (i) What burden must the government leave for future
generations if it is to remain solvent? and (ii) Is the resulting fiscal treatment
of future generations identical to that of the current newborns?
Economically, there is no reason for equalizing individual taxes and
transfers. Most fiscal regimes involve high taxes for high-income individuals
and lower taxes for less productive workers. In the same way, if one generation
is economically more productive than another (for example because its
proportion of rich people is higher), it seems justified to make it pay more taxes
or receive less transfers. The generational accounting methodology partially
takes account of differences in generational wealth by comparing generations
in terms of fiscal pressure rather than gross burden. Using balanced growth
assumptions, the classical methodology defines the generationally balanced
policy as a situation in which individual taxes and transfers increase at the
same pace as labour productivity. However, existing generational accounting
exercises rely on very simple assumptions about the changes in labour
productivity and/or lifetime labour income across generations. Usually, these
changes are related to exogenous growth rates that have no explicit link with
the skill composition of the generations. Hence existing studies do not take
account of an important source of heterogeneity within and between
r The London School of Economics and Political Science 2006
330 ECONOMICA [MAY
must account for the real lifetime income from the labour of these cohorts. By
disregarding the skill structure of future generations, the accounting technique
is likely to underestimate the lifetime labour income of such cohorts.2
This paper shows that the rise in educational attainment strongly modifies
GPS’s conclusion. To allow comparisons, our analysis is also based on the
1995 fiscal year, and our adjustment calculations rely on the counterfactual
hypothesis that all changes began in 1995. Our conclusions are more optimistic
and raise numerous issues about: (i) educational policies and social mobility
(how can the incentives for education be increased? will the cost of increasing
average schooling levels exceed the fiscal gains?); (ii) the political sustainability
of current fiscal policy (if the government deficit decreases, will there be fiscal
pressure to reduce taxes or to increase transfers?) and (iii) the structure
of the labour market (can these fiscal gains survive the increasing supply of
skills?). Nevertheless, the sole purpose of our contribution is to compute the
impact of educational changes on the government budget constraint, all other
things being constant, i.e. taking the economic environment as given.
Consequently, our results must be appreciated exclusively in terms of fiscal
sustainability.
The rest of this paper is organized as follows. Section I discusses data issues
and the calibration of net tax profiles by age and education level. Section II
considers methodological issues of generational accounts with skill hetero-
geneity. As in GPS, our basic assumptions about the growth rates of per capita
taxes and benefits build on the official projections of the Congressional Budget
Office (CBO). The term ‘fiscal policy’ then reflects both the current age profile
of net taxes (by age and educational level) and official growth rates for per
capita amounts. Generational accounts by skill level are computed in Section
III. Our results indicate that low-skill newborns are characterized by a negative
generational account ( 15.4% of their lifetime labour income). However,
lifetime net tax rates for medium and high-skill newborns are positive,
amounting to 26.8% and 32.3%, respectively.
Taking account of the rise in educational attainment has a strong impact
on the results. Compared with GPS, we show that, in our baseline scenario, the
total burden left for future generations is reduced by about 30.7%.
Nevertheless, the dramatic rise in educational attainment does not restore
fiscal sustainability. Balancing the long-run deficit requires taxes to be
increased (by about 1.2%) or transfers reduced (by about 2.7%). This
contrasts with GPS’s study, which suggests that restoring the generational
balance in 1995 would require all transfers to be cut by 17.5% or all taxes
increased by 8.2%.
Section IV describes a similar exercise, when public education expenditure
is treated as a transfer rather than as part of government purchases. Similar
results are obtained, but the government deficit becomes slightly higher than in
our baseline scenario. A sensitivity analysis is presented in Section V. Our
results are quite robust to growth and discounting assumptions, to the
treatment of education spending and to the evolution of the skill premium.
They are more sensitive to assumptions about the schooling level of
future generations. Section VI summarizes our conclusions. An Appendix
provides the mathematical tools for generational accounting with hetero-
geneous skills.
r The London School of Economics and Political Science 2006
332 ECONOMICA [MAY
I. DATA ISSUES
Estimating the age profiles of taxes and benefits for a reference year is the
building block for any longitudinal calculation. In this paper we use GPS’s
estimates of the profiles of men and women in 1995. We aggregate men’s and
women’s profiles using gender weights by age. These profiles cover six types of
tax (labour income taxes, FICA taxes, excise taxes, capital income taxes,
seignorage taxes and property taxes), seven types of benefit (OASDI, Medicare,
Medicaid, unemployment insurance, general welfare, AFDC and food stamps)
and three items of government purchases (education expenditures, federal
government purchases and state government purchases). The age profiles are
scaled so as to match the aggregated observations (i.e. the observed proportion
of GDP). The bold lines in Figure 1(a) and (b) represent the total taxes and
total benefits paid or received (excluding education spending) by a
representative individual in each age cohort in 1995. The bold line in Figure
1(c) gives the difference between taxes paid and benefits received. It appears
that net taxes are positive for individuals aged 15–70 (net contributors) and
negative for individuals aged over 70 (net beneficiaries).
These representative profiles depict the weighted average of net taxes paid
by all US residents in 1995. The age-specific amounts are thus dependent on the
average characteristics of each age group, in particular their average skill levels.
In this study, we disaggregate each tax or benefit item of the GPS dataset by
educational level. Three educational levels were distinguished: less than high
school diploma (oHS), high school diploma only (HS), and more than high
school diploma (4HS). Two statistical sources were used:
1. For labour income taxes, capital income taxes and OASDI benefits, we used
data extracted from the PSID (Panel Study on Income Dynamics3) of the
US population aged 20 and over in 1993. After elimination of illogical
information and non-responses, we computed the representative profiles for
these items from 17,943 observations. For those aged under 20, we assumed
that the profiles were identical, irrespective of education. PSID data were
also used to compute the lifetime labour income for each level of schooling.
2. For other items of taxes and benefits (including public education expendi-
tures), we used the profiles estimated by Lee and Miller (1997) by educational
level, cohort and immigration status. These profiles were based on 1994 and
1995 data from the CPS (Current Population Surveys). For children (aged 0–
19), education was forecast on the basis of parental education and ethnicity.
We did not distinguish between immigrants and natives.
40,000
35,000
30,000
25,000
20,000
15,000
10,000
5,000
0
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96
< HS HS > HS GPS average
FIGURE 1(a). Tax profile by age and education (present value in 1995 dollars).
25,000
20,000
15,000
10,000
5,000
0
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96
< HS HS > HS GPS average
FIGURE 1(b). Transfer profile by age and education (present value in 1995 dollars).
40,000
30,000
20,000
10,000
0
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96
−10,000
−20,000
< HS HS > HS GPS average
FIGURE 1(c). Net tax profile by age and education (present value in 1995 dollars).
r The London School of Economics and Political Science 2006
334 ECONOMICA [MAY
70%
60%
50%
40%
30%
20%
10%
0%
1940
1950
1960
1970
1980
1990
2000
2010
2020
2030
2040
2050
2060
2070
2080
2090
2100
<HS (%) HS (%) >HS (%)
1.4%
1.2%
1.0%
0.8%
0.6%
0.4%
0.2%
0.0%
2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
GPS This study
FIGURE 2(b). Common growth rate of labor income per capita (same growth rate per educational
group).
3.5%
3.0%
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
− 0.5%
1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070
− 1.0%
Maddison GPS This study Average Madison 1960-2000
FIGURE 2(c). Growth rate of the US GDP per capita (historical data and projections).
Profiles of taxes and transfers The net tax profiles of each educational group
are assumed to be stable over time. For example, the work effort of educated
r The London School of Economics and Political Science 2006
336 ECONOMICA [MAY
The evolution of taxes and transfers We took our growth rates of per capita
taxes and benefits from the GPS study. These in turn built on the reference
projection of the Congressional Budget Office (1997a, b) through 2070.
A growth path was calculated for each item of tax and benefit. We applied
these growth rates to per capita amounts paid (or received) by the members
of each educational category. For a given path of growth rates, our fiscal
projections are affected by changes in educational structure and demo-
graphic size (although the GPS/CBO aggregates are affected only by the
demography).
Does our procedure involve double-counting the effect of the rise in
educational attainment on productivity growth? This would clearly be the case
if the CBO growth assumptions reflected expected changes in human capital. In
that case an alternative scenario would be to calibrate the growth rates so as to
match the CBO aggregate projections for each item. We did not follow this
procedure, for three reasons. First, the CBO projection does not take account
of the rise in educational level of the population. Second, as shown on Figure
2(b), our baseline scenario using per capita amounts is consistent with the 1.2%
exogenous growth rate fixed for the post-2070 period. (The GPS scenario based
on CBO aggregates induces smaller wage increases in the transition period.)
Finally, as shown in Figure 2(c), our scenario predicts a growth rate of US
GDP per capita (sum of labour income and capital income per capita) which is
very similar to the average growth rate experienced between 1960 and 2000.
Consequently, our aggregated projections do not correspond to those of the
Congressional Budget Office.
Other assumptions
The present value of government purchases (PVG) is the discounted sum of
government purchases (state and federal purchases, including educational
expenditures) in 1995 under the assumed fiscal policy. The baseline scenario
builds on the GPS study. We used the path of government purchases simulated
by the Congressional Budget Office up to 2070. Beyond 2070, government
purchases are assumed to grow by 1.2% per year. Note that government
purchases include expenditure on public education. It could be argued that our
baseline assumption is incompatible with the future rise in school attendance.
For this reason, in Section IV we examine an alternative scenario in which
r The London School of Economics and Political Science 2006
2007] FISCAL POLICY AND EDUCATIONAL ATTAINMENT 337
education expenditures are treated as transfers (i.e. they vary with changes in
school attendance). The growth rate of individual taxes and transfers in the
PVG beyond 2070 is also set at 1.2%. The baseline discount rate is 6%. The
public net wealth is the only observable item of the budget constraint: as in the
GPS study, it amounted to $2.1 trillion in 1995.
TABLE 1
Generational Accounts of Living Generations: Baseline Scenario
(present value in 1995 dollars)
TABLE 2
Generational Imbalance and Lifetime Net Tax Rates: Baseline Scenario
(present value in 1995 dollars)
250,000
200,000
150,000
100,000
50,000
0
−50,000 0 10 20 30 40 50 60 70 80 90 100
−100,000
GPS average GA
FIGURE 3. Average generational account per living cohort (Baseline scenarioFpresent value in
1995 dollars).
estimates, compared with $9398 billion in the GPS methodology. This is due to
the fact that the GPS methodology underestimates the total payments made by
living generations.
TABLE 3
Generational Accounts of Living Generations: Education Benefit
Scenario (present value in 1995 dollars)
TABLE 4
Generational Imbalance and Lifetime Net Tax Rates: Education Benefit
Scenario (present value in 1995 dollars)
present value at birth). The changes are important for individuals aged 25 and
under. They are minor for older generations for whom mandatory education
belongs to the past. As will be shown in Table 4, the lifetime net tax rates of
newborns become 68.2%, 2.5% and 13.3% for the three educational groups,
respectively.
Once again, our average generational accounts per cohort are not identical
to GPS amounts. The differences are small for old-age cohorts but are larger
for young cohorts. Extrapolating the future taxes and transfers of newborns on
the basis of the contemporary profile, the classical method underestimates
newborns’ average account by about 35%.
In this scenario, the total burden left for future generations amounts to
$2063 billion (instead of $4881 billion as in the GPS methodology). The current
fiscal policy is still unsustainable and generates a long-run budgetary deficit.
The deficit is larger than in the baseline case. Education spending increases
more quickly than in the baseline scenario owing to the changes in the
educational structure of the population. Table 4 shows that restoring the
balance through tax adjustments would necessitate a rise in all taxes by about
1.39% (against 5.0% in the GPS framework). The lifetime net tax rates then
become 67.5%, 3.2% and 13.8% for the three educational groups. Restoring
the balance through transfer adjustments means a cut in all transfers by about
2.38% (against 13.0% in the GPS framework), and lifetime net tax rates
become 65.3%, 3.5% and 13.9%, respectively.
V. SENSITIVITY ANALYSIS
Generational accounts depend on uncertain assumptions about demographic
changes, labour productivity growth rates, budgetary policy, interest rates and
r The London School of Economics and Political Science 2006
2007] FISCAL POLICY AND EDUCATIONAL ATTAINMENT 343
TABLE 5
Sensitivity to the Return to Skill ( 10% of the Tax Ratio): Generational
Accounts and Budgetary Adjustments
Compared with the ‘high’ and ‘low’ variants in Cheeseman Day and Bauman
(2000), our baseline scenario can be considered a reasonable middle path.
Two alternative forecasts can be simulated. Cheeseman Day and Bauman’s
(2000) the high projection is based on very optimistic assumptions about future
educational attainment. In the long run, it assumes that the proportions of
high-skilled workers aged 30 will reach 70.2% in 2030 (against 5.2% for low-
skilled workers and 24.6% for medium-skilled). Our ‘low variant’ takes the
most pessimistic annual forecast from Lee and Miller. We keep the 2005
educational structure (16% for the low-skilled, 32% for the medium-skilled
and 52% for the high-skilled) constant after that year.
Table 6 gives the results. Clearly, newborns’ generational accounts and
lifetime net tax rates are quite stable. In terms of generational balance, the low
variant has a substantial impact on the long-run deficit. Taxes would have to
be increased by 5.69% (instead of 1.16% in the baseline scenario) or transfers
reduced by 12.52% (instead of 2.68%). However, the high variant generates a
long-run budgetary surplus. A continued rise in educational attainment would
make the present fiscal policy sustainable: taxes could be reduced by 1.3%
or transfers increased by 3.13%. Of course, reaching the high variant’s
educational attainment would probably require a strongly expansionary
education policy. If the marginal cost of education increases, the discounted
r The London School of Economics and Political Science 2006
2007] FISCAL POLICY AND EDUCATIONAL ATTAINMENT 345
TABLE 6
Sensitivity to the Long-run Educational Structure (High and Low
Variants): Generational Accounts and Budgetary Adjustments
Newborns’ generational account and lifetime tax rate without policy adjustment
Generational accounts Lifetime net tax rate
cost of such a policy could exceed the discounted fiscal gains. In some sense
the baseline scenario seems more realistic, since it is based on current state
commitments. Nevertheless, our sensitivity analysis demonstrates the crucial
role of education policies in the debate on ageing and public finance.
TABLE 7
Sensitivity to Discount and Growth Rate Assumptions: Generational
Accounts and Budgetary Adjustments
Newborns’ generational account and lifetime tax rate without policy adjustment
Generational accounts Lifetime net tax rate
oHS g ¼ 0.007 g ¼ 0.012 g ¼ 0.017 g ¼ 0.007 g ¼ 0.012 g ¼ 0.017
about the interest rate modify the calculations. The lifetime net tax rate of low-
skill individuals is particularly affected by interest rate changes. However,
interest rates and growth rates do not modify the main result: in all cases, the
current fiscal policy is unsustainable. The policy adjustments needed to restore
the generational balance are robust (especially if the interest rate increases).
The tax variation ranges from 1.0% to 4.7% in the two extreme scenarios.
Transfer cuts range from 2.5% to 9.0%.
VI. CONCLUSION
It is usually argued that expected demographic changes threaten the
sustainability of fiscal policies. In most industrialized countries, social policies
involve considerable transfers from young cohorts to old cohorts. If the
number of beneficiaries increases, the financial viability of the system is
obviously called into question. Generational accounting is generally seen as a
meaningful way of evaluating the sustainability of fiscal policy. Recent
r The London School of Economics and Political Science 2006
2007] FISCAL POLICY AND EDUCATIONAL ATTAINMENT 347
where PVLt is the present value of net tax payments by living generations over the rest
of their lives; PVFt is the present value of net tax payments by future generations; PVGt
is the present value of prospective government purchases of goods and services; and
NWt is the net public wealth.
The net wealth at time t is observed. Two of the remaining terms, PVGt and PVLt,
are projected using current observations and official projections. The fourth term, PVFt,
can thus be calculated as the residual burden bequeathed to future generations.
The present value of government purchases is the discounted sum of public
expenditures,
X
1
Gs
ðA2Þ PVGt ¼ ;
s¼t ð1 þ iÞst
where Gs is the value of public purchases projected at time s X t, and i is the interest rate.
In practice, the path of Gs can be projected partly on the basis of budgetary forecasts
(i.e. between t and tn) and partly using balanced growth assumptions (between tn and
1). In the long run, it is assumed that Gs will grow at the same rate as total factor
productivity, g.
The present value of net tax payments by living generations can be obtained by
summing the present value of the net taxes these generations will pay to the government
over the rest of their lives, i.e. summing the generational accounts of living cohorts. We
distinguish three educational levels (L ¼ low, M ¼ medium and H ¼ high) and assume
that each individual lives a maximum of D years. The present value of payments by
living generations, PVLt, can be written as
D
X
ðA3Þ PVLt ¼ nLj;t pLj;t þ nM M H H
j;t pj;t þ nj;t pj;t ;
j¼0
where pX
j;tis the size of type-X population (X ¼ L, M, H) of age j at time t, and nX
j;t
measures the generational account of these agents.
Generational accounts sum the value of net taxes to be paid by each type of
individual over the rest of his or her life:
D yX
1 X
X
k;tþkj pk;tþkj
ðA4Þ nX
j;t ¼ X
; j ¼ 0; . . . ; D;
pj;t k¼j ð1 þ iÞkj
where yXk;tþkj is the net tax payment by an agent of type X and age k at time t þ k j.
In practice, pX k;tþkj can be projected using demographic forecasts (including
mortality and net immigration flows), data on schooling levels by age, and estimates of
the educational attainment of young living generations when their education is
complete. The net taxes yX k;tþkj can be extrapolated partly on the basis of short-run
forecasts (taking account of official budgetary projections and potential fiscal reforms
between t and tn) and partly using balanced growth assumptions (between tn and 1).
Typically, different assumptions can be considered for the components of yX k;tþkj .
It should be noted that the generational accounts of newborns, measuring the
present value of net taxes they can be expected to pay over their whole lifetimes, need
not all be the same sign. They can be negative for low-skill individuals and positive for
the high-skilled ones. These generational accounts can be expressed as percentages of
X
the average discounted lifetime labour income, denoted by W0;t for a newborn agent of
type X. In GPS’s scheme, this defines the lifetime net tax rate of the newborns as
nX
0;t
ðA5Þ LNRX
0;t ¼ X
:
W0;t
future generations, PVFt, does not, in itself, answer this question. To go further, the
aggregate burden must be transformed into an individual amount, the average account
of future cohorts.
One way to proceed is to compute the hypothetical generational accounts of future
cohorts under current fiscal policy. Using same reasoning as in equations (3) and (4)
above, we can write
min½st1;D
X yLj;s pLj;s
X
1 þ yM M H H
j;sþj pj;s þ yj;sþj pj;s
ðA6Þ PVFtn ¼ ;
s¼tþ1 j¼0 ð1 þ iÞst
where PVFtn is the present value of net payments by future generations if current fiscal
policy is unsustainable.
This hypothetical value can then be compared with the residual value PVFt
computed from equation (1):
if PVFtn ¼ PVFt the policy is sustainable and there is no need to make fiscal
adjustment;
if PVFtn > PVFt the government budget is in surplus and benefits could be increased
without increasing taxes;
if PVFtn < PVFt the current policy is not sustainable or is not generationally balanced:
is implies that either future generations must pay more net tax than current
generations or current policy must be adjusted to restore sustainability.
If the present policy is unsustainable, the obvious strategy is to adjust taxes and/or
transfers at some future date. In this paper we use an adjustment method that covers all
the members of all generations. If a gap has to be financed (to cover a deficit) or a
surplus reallocated, we compute the proportional adjustment in all taxes (or all
transfers) required to balance the budget.5
Let us decompose the net taxes on all generations into two basic components, taxes
and benefits: yX X X
j;s ¼ yT;j;s yB;j;s . A time-invariant adjustment factor can be applied to
each of these components (ZT for taxes and ZB for benefits) so as to restore
sustainability. We then apply these proportional changes to both living generations
(over the rest of their lifetimes) and future generations, so as to balance the budget
constraint. Our adjustment rule is summarized by the following set of equations:
h i
X D X D X yX X X
T;k;tþkj ð1 þ ZT Þ yB;k;tþkj ð1 ZB Þ pk;tþkj
adj
PVLt ¼
j¼0 k¼j X¼L;M;H ð1 þ iÞkj
h i
1 min½st1;D
X X X yXT;j;s ð1 þ ZT Þ y X
B;j;s ð1 Z Þ
B pj;s
X
PVFtadj ¼
s¼tþ1 j¼0 X¼L;M;H ð1 þ iÞst
There is a continuum of pairs (ZT, ZB) restoring the balance. Two specific pairs are
usually considered, one with ZT ¼ 0 if the balance is achieved through cuts in transfers
and one with ZB ¼ 0 if the balance is achieved through tax increases. For each scenario,
the lifetime net rate of tax for future generations can be computed and compared with
that of current newborns.
ACKNOWLEDGMENTS
We thank two anonymous referees, Gilles Duranton, Joël Hellier and Philip Oreopoulos
for very helpful comments. We are grateful to Alan Auerbach, Tim Miller and Philip
Oreopoulos for transmitting their program and dataset. The usual disclaimers apply.
r The London School of Economics and Political Science 2006
350 ECONOMICA [MAY
NOTES
1. Indeed, it makes little sense to assume that the average net taxes of the generation aged 20 in
1995 will be the same as the average net taxes paid by current older generations.
2. This bias is less important than the previous one, since future changes in educational
attainment are likely to be much smaller than the differences observed among living
generations.
3. These data are available from the website: http://psidonline.isr.umich.edu
4. Taking into account differences in mortality rates would probably make our results slightly
less optimistic. However, as shown in Figure 1, the main differences in generational accounts
are related to differences in taxes (and not to differences in transfers in old age).
5. Our strategy differs slightly from GPS, who compute the changes in taxes and/or benefits
needed to equalize the lifetime net tax rates of current and future generations. It should be
noted that, in line with GPS’s approach, the balance could also be restored through changes
in government purchases.
REFERENCES
ACEMOGLU, D. (2002). Technical change, inequality and the labour market. Journal of Economic
Literature, 40, 7–72.
AUERBACH, A. J., GOKHALE, J. and KOTLIKOFF, L. J. (1991). Generational accounts: a meaningful
alternative to deficit accounting. In D. Bradford (ed.), Tax Policy and the Economy, Vol. 5, pp.
55–110. Cambridge, Mass.: MIT Press.
FFF, FFF and FFF (1994). Generational accounts: a meaningful way to evaluate fiscal
policy. Journal of Economic Perspectives, 8, 73–94.
CHEESEMAN DAY, J. and BAUMAN, K. J. (2000). Have We Reached the Top? Educational
Attainment Projections of the US Population. Population Division Working Paper, no. 43, US
Census Bureau.
CONGRESSIONAL BUDGET OFFICE (CBO) (1997a). An Economic Model for Long-run Budget
Simulations. Washington: CBO, July.
FFF (1997b). The Economic and Budgetary Outlook: An Update. Washington: US Government
Printing Office, September.
GOKHALE, J., PAGE, B. R. and STURROCK, J. R. (1999). Generational accounts for the United
States: an update. In A. J. Auerbach, L. J. Kotlikoff and W. Leibfritz (eds.), Generational
Accounting Around the World. NBER Books, Chicago: University of Chicago Press.
HAVEMAN, R. (1994). Should generational accounts replace public budgets and deficits? Journal of
Economic Perspectives, 8, 95–111.
KOTLIKOFF, L. G. (1992). Generational Accounting: Knowing Who Pays, and When, for What We
Spend. New York: Free Press.
LEE, R. D. and MILLER, T. (1997). Immigrants and their descendants. Project on the Economic
Demography of Interage Income Reallocation. University of California: Berkeley.