287
LIFE COURSE PERSPECTIVE ON ECONOMIC SHOCKS AND INCOME 
INEQUALITY THROUGH AGE-PERIOD-COHORT ANALYSIS: 
EVIDENCE FROM FINLAND
by Esa KaronEn* and MiKKo niEMElä
University of Turku
Utilizing age-period-cohort analysis, this paper examines the development of income distribution 
across periodic economic fluctuations in relation to cohorts and age groups. The empirical analysis is 
based on the Finnish Income Distribution Statistics and Household Expenditure Surveys covering the 
period of 1966–2015. The findings suggest that the period and cohort effects can be identified as the 
main effects on relative income, while the age effects have no meaningful impact when the control vari-
ables are taken into account. This result reveals a connection between the effects of economic shocks 
and cohort placement on labor market entry. Additionally, absolute income analysis suggests that eco-
nomic shocks create stagnation points in income development, which are especially detrimental to 
cohorts who are transitioning into labor markets. Additionally, middle-income attainment has not 
changed due to periodic shocks but rather is related to inter-cohort inequalities and relative income 
differences, where the baby boomer generation is a clear winner.
JEL Codes: C31, D31
Keywords: age-period-cohort effects, income distribution data, income inequality, inter-cohort 
differences
1. introduction
Research on economic inequality usually focuses on income dynamics ana-
lyzed through the lenses of age and period characteristics. Cohorts are often 
ignored because of methodological limitations and the lack of appropriate data. 
Although there is a wide range of evidence for age and period effects on income 
dynamics, there is less information on the generational pattern of the relationship 
between income and cohorts. We argue that shifting our focus to cohort differences 
in relation to age and period may reveal significant differences in income dynamics 
across households (see also Lim and Zeng, 2016). The questions are whether there 
are inter-cohort inequalities and whether some cohorts hold a better economic 
position than others.
In Finland, the trend of income inequality since the 1960s can be divided into 
five periods (Blomgren et al., 2014). First, the era of welfare state expansion in the 
1960s and 1970s decreased income inequality regardless of the income concept. 
Notes: This research was supported by the Strategic Research Council of the Academy of Finland 
(decision numbers: 293103 and 314250).
*Correspondence to: Esa Karonen, University of Turku, Assistentinkatu 7, 20014 Turku, Finland 
(eokaro@utu.fi).
Review of Income and Wealth
Series 66, Number 2, June 2020
DOI: 10.1111/roiw.12409
This is an open access article under the terms of the Creative Commons Attribution License, 
which permits use, distribution and reproduction in any medium, provided the original work is 
properly cited.
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
bs_bs_banner
Review of Income and Wealth, Series 66, Number 2, June 2020
288
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
Second, from the mid-1970s to the economic recession of the early 1990s, market 
income inequality increased but, due to income transfers, gross and disposable 
income inequality remained constant. Third, the recession of the 1990s increased 
inequality in market income but not in gross and disposable income. Fourth, whereas 
market income inequality since the mid-1990s has been constant, inequality in gross 
and disposable income increased towards the early 2000s. From a comparative per-
spective, the increase in income inequality was exceptionally fast and steep in Finland 
during the period between 1995 and 2002 (OECD, 2008, 2011). Fifth, since the turn 
of the millennium, the development of income inequality has been rather stable.
The Finnish evidence suggests that since the 1990s, changes in income inequal-
ity have been associated with the economic cycle. Income inequality increased 
during periods of economic growth and decreased during economic downturns. 
However, the development of income inequality on the aggregate level cannot 
reveal the intricacies of income dynamics in terms of inter-cohort inequalities. This 
situation calls for a life course perspective that would identify—separately—the 
effects of age, time and generation. In this study, we define life course as life events, 
transitions and trajectories with cohort variation in development (Elder, 1997). 
Life course here is defined as developmental patterns that are structured by events 
and other biological and social constraints and that vary by historical time. The 
individual is defined as part of a certain historical period or an event by his/her 
birth year. The impact of a historical event is contingent on the point of intersec-
tion of the life stage of the cohort (Elder, 1997). Such a perspective emphasizes that 
economic shocks could lead to an accumulating effect for certain cohorts. Hence, 
the Finnish case, combined with the long time-series data utilized in this study, pro-
vides an ideal context in which to focus on age-period-cohort effects in income dis-
tribution and, specifically, to chart the effects of the economic cycle. Prior research 
shows that there are significant generational differences in economic measures such 
as income, consumption and wealth (Jappelli, 1999; Berloffa and Villa, 2010; Lim 
and Zeng, 2016). In addition, empirical evidence from various countries supports 
the concern that younger generations are falling behind compared to older ones in 
regard to the evolution of household income (e.g. Smeeding and Sullivan, 1998; 
Gosling et al., 2000; Beaudry and Green, 2000; Fitzenberger et al., 2001; Grenier, 
2003; Osberg, 2003; OECD, 2011; Ostry et al., 2014; Chetty et al., 2017).
We approach our research questions from the perspective of the concepts of 
relative and absolute income. In practice, this means that we analyze income distri-
bution from the “detrended” perspective, which will determine which birth-cohorts 
and age groups have benefitted the most in terms of relative income. This perspec-
tive alone does not illuminate how inter-cohort income has developed over time 
in an absolute fashion; thus, we address this question in our “trended” approach, 
which measures how absolute income has developed over time in regard to cohort 
and age groups. Finally, we put absolute and relative income into the context of the 
“optimal” income class by measuring the attainment of middle-income according 
to the dimensions of age, period and cohort. Hence, our main contribution to 
the literature is to examine the effects of economic cycles on inter-cohort income 
inequality in an absolute and relative manner, taking into account the three dimen-
sions of age, time and cohort. In contrast, previous studies have mostly included 
only two of these three factors.
Review of Income and Wealth, Series 66, Number 2, June 2020
289
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
This study aims to improve upon previous studies in multiple ways. First, this 
study will assess how income distribution has developed across periodic economic 
fluctuations in relation to cohorts and age groups. The central interest lies in how 
economic shocks and cycles affect income through inter-cohort income dynamics. 
Second, this study utilizes register-based datasets with significant periodic range, 
which are capable of revealing the long-term effects of various economic cycles. 
Cohort studies have usually been conducted on data that do not have a satisfy-
ing yield in statistical years. Finland is a unique case, with its economic history 
that contains four distinct economic shocks. Therefore, the Finnish case offers the 
opportunity to observe multiple points of economic fluctuation over time. Third, 
the main drawback of previous research on income inequality and the distribu-
tional effects of economic cycles is that this research has usually used narrower 
models, which only take two factors of the age-period-cohort triad into account, 
whereas we consider all three factors at the same time. What makes our contribu-
tion stronger is our use of a relatively new method of age-period-cohort modeling, 
which is capable of measuring relative and absolute changes in income distribution.
This paper is organized as follows. Section 2 discusses related studies. Section 3 
presents the data and descriptive facts about age-period-cohort (APC) profiles that 
motivate the specification of the statistical APC models in Section 4. Section 5 pres-
ents the empirical results of our age-period-cohort analysis of income distribution 
in Finland during the period between 1966 and 2015. Analyses provide illustrated 
graphs and tables of the estimates for both the detrended and trended APC models. 
Finally, Section 6 offers concluding remarks and discussion.
2. rElatEd studiEs and thE FraMEworK oF analysis
A generation carries a unique “scar” that it acquires through shared social-
ization, as seen in how different birth-cohorts grow up in a similar historical 
period (Mannheim, 1928). In an economic context, this can be described as 
inequality because some cohorts may have a smoother entry into the labor mar-
ket due to their specific economic situation. Thus, an economic up- or down-
turn can play an enormous role in how a given generation is able to establish 
itself during changing market situations. For example, cohorts that have become 
adults during economic booms are more likely to profit from that favorable mar-
ket situation. Conversely, cohorts who are “scarred” by an economic downturn 
may be more risk-averse and have a more disadvantaged economic trajectory 
from a life course perspective (Malmendier and Nagel, 2011). These observations 
have created building blocks for the life course hypothesis and paved the way 
for new developments of the theory whose main themes revolve around the idea 
of systematic cycles of advantaged and disadvantaged generations (Myles, 2002, 
p. 138). This approach was pioneered by many researchers (e.g. Campbell et al., 
1976; Clark and Oswald, 1994; Kahneman et al., 1999; Frey and Stutzer, 2002; 
Helliwell, 2003; Layard et al., 2012) who focused on how well-being is affected 
by other outcomes such as income, employment and educational qualifications.
Changes in income trajectories and distribution can be tracked through out-
ward stimuli, for example, through macro-economic shocks (Mayer, 2005). Periodic 
Review of Income and Wealth, Series 66, Number 2, June 2020
290
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
change indicates how socio-political reactions against economic downturns have 
changed the income dynamics between generations. Mayer (2005) argues that these 
cohort changes are key to creating a more solid picture of how different mechanics 
of inequalities across generations are constructed. Regarding the external stim-
uli that Mayer mentions, there have been various alternating phases of economic 
downturns and growth in Finnish economic history. The cycle of 1970 to 1974 was 
linked to the 1973 oil crisis, which was followed by a phase of economic growth 
from 1980 to 1990. The period from 1990 to 1995 was the most significant shock 
that Finland has ever faced and has largely been considered “the great depression”. 
These shocks offer excellent effects through which to observe the influence of eco-
nomic cycles on inter-cohort income dynamics.
Results from previous cohort studies examining income inequality can be 
summarized in three observations. First, it seems that the “baby boomer” gener-
ation has benefitted most from its birth cohort compared with other generations. 
Studies have researched the impact of cohort membership on disposable income 
in European welfare states and the United States (Chauvel, 2013; Chauvel and 
Schröder, 2015; Freedman, 2017). These studies show that there are discrepancies 
in income accumulation between cohorts, which means that having been born 
in the “baby boomer” generation seems to give an advantage. Second, economic 
shocks have been identified as a major influence on inter-cohort income inequality. 
The main hypothesis states that cohorts that enter the job market during times of 
austerity and economic downturn are—compared to cohorts born during an eco-
nomic upturn—in a more disadvantaged position with regard to attaining similar 
career options. For example, younger generations have an 8 percent lower expected 
income in Italy than older generations, which is a result of the economic situa-
tion and different socio-political reforms (Berloffa and Villa, 2010). Additionally, 
research on income and wealth inequality has reported that younger generations 
have lower living standards than their parents did at similar ages in Great Britain 
(Crawford et al., 2015), Europe (OECD, 2011), the United States (Chetty et al., 
2017), Canada (Kershaw, 2015) and Australia (Daley and Wood, 2014). Third, one 
effect is the role of educational expansion, which is linked to the profits gained 
from certain educational fields (Pekkala and Lucas, 2004). Older generations had 
the opportunity to work towards higher education but also benefitted from growing 
job markets and low competition within the same educational field. As educational 
expansion advanced, the younger generations had to compete with an ever-growing 
pool of highly educated individuals for the same jobs. Thus, according to supply 
and demand, higher degrees hold less profit-making value for younger generations 
than for older generations, who reaped the benefits of starting their careers in an 
optimal job market situation while also gaining work experience.
The Finnish context tells a similar story. Riihelä (2006) found that the younger 
generation, who faced an economic shock, attained a lower income level than what 
would be expected according to the life-cycle hypothesis. This result illustrates how 
the role of economic cycles affects income trends between generations. These cycles 
are linked to periodic changes in the market, and certain birth cohorts occur to be 
at the “right place at the wrong time”. Additionally, there are indications that inter-
generational income mobility did not increase much beyond the levels achieved 
among the “baby boomer” cohorts born in 1945–1950, although these older 
Review of Income and Wealth, Series 66, Number 2, June 2020
291
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
cohorts benefitted from educational expansion and a more advantageous labor 
market situation (Pekkala and Berman, 2002; Pekkala and Lucas, 2007). Overall, 
relative income mobility has decreased over time among young adults. However, 
the decrease is related to the rise of permanent income inequality (Suoniemi, 2012). 
Additionally, the development of pension policy has particularly strong impacts 
on both poverty cycles and economic well-being among the elderly (Jäntti et al., 
1996; Kangas and Palme, 2000).
3. data, VariablEs and stylizEd Facts
Our empirical analyses are based on the Finnish Income Distribution 
Statistics (IDS) and the Finnish Household Expenditure Surveys (HES) provided 
by Statistics Finland. Both datasets belong to the series of the Official Statistics 
of Finland (OSF) and the European Statistical System (ESS). HES provide data 
on incomes and expenditures in 1966, 1971, 1976, 1981 and 1985. The data are par-
tially derived from interviews, and since the 1971 survey, they have been derived 
from official registers. IDS provide data on incomes, and the data have been 
collected annually since 1987. Thus, the period of analysis in this study is 1966-
2015. In both cases, the income data are collected from tax and other registers 
and are generally considered to be of high quality. The data are harmonized to 
be a representative sample of the Finnish population. The basic unit of anal-
ysis is the household. The sample size varies from 4,471 households in 1966 to 
10,620 households in 2015. The mean sample size is 10,216 households. The data 
are multiplied to the level of total population by special weights included in the 
household surveys. In the household assets, the person with the highest personal 
income is chosen as the household’s reference person, which serves as a proxy for 
demographic and background status. The income variables measure the house-
hold’s income and individual income of the household’s reference person.
The descriptive statistics of the merged data are shown in Table A.1. In regard to 
dependent variables, the income metric is the annual household disposable income, 
which includes monetary income items and benefits in kind connected to employ-
ment relationships, but does not include imputed income items such as imputed 
rent (see OSF, 2015). Household income is not top coded. The dependent variable 
is the logged equivalized annual household disposable income, which is adjusted for 
inflation. As a common practice for empirical application, we have bottom-coded 
all of the negative income values as zeroes, because in equivalization it makes little 
sense to apply equivalization to negative values (see OECD, 2012, 2015).
Conceptually, we measure disposable household income in the form of relative 
income, absolute income and attainment of middle income. Relative income reveals 
how income is in relation to other households of society, weighing it against the 
standards of the given period, whereas absolute income reflects the total amount 
of a household’s earnings received in a given period. Attainment of middle income 
is annual income, which is transformed as income deciles and derived as a dummy 
variable: the first and second quintiles equal zero, and the third to fifth quintiles 
equal one. The aim is to measure the attainment of income class, which exceeds the 
threshold of the middle-income level.
Review of Income and Wealth, Series 66, Number 2, June 2020
292
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
This research uses the equivalence scale, which is a variant of the Oxford scale. 
The equivalence scale is constructed with the formula m = 1 + a(A-1) + bL, where 
A is the number of adults and L is the number of children in the household. The 
value of parameter a is 0.5, and b is 0.3. Because of the inadequate information on 
children’s ages, this paper uses the equivalence scale, where the reference is 18 years 
old or above, unlike the Oxford scale, in which the reference person of the house-
hold is 14 years old or over.
For the purposes of this study, the analyses are limited to the population aged 
20 through 70 years. The rationale for excluding the 18- and 19-year-old age groups 
is the requirement for conscription in the Finnish Defense Force (6-12 months) and 
the requirement for civilian service (12 months). For example, conscription usually 
occupies approximately 14000 people annually. Thus, excluding this age group will 
also provide a more neutrally distributed sample.
Cohort variables were constructed to be consistent with previous studies and 
with the orthogonal requirements of the age-period-cohort models (Chauvel, 
2011; Chancel, 2014; Chauvel and Schroder, 2015). We conducted sensitivity tests 
to find the optimal cut-off  values for the APC variable grouping (not shown here) 
and decided to use 5-year intervals to maintain acceptable observation size and 
Figure 1. Stylized Facts of Age-Cohort-Period Profile
Review of Income and Wealth, Series 66, Number 2, June 2020
293
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
measurement accuracy for each group. This choice was also made to mitigate 
possible volatility connected to measurement range and observation sizes within 
groups (see Luo et al., 2016). Cohorts were cut in 5-year intervals spanning the 
period from 1945 to 1995. The formed variables were constricted by the age factor 
of 20- to 70-year-olds to restrict the range to the active working population. The 
period variables were formed between 5-year intervals. The reasoning behind this 
decision was the rhythm of economic cycles.
Figure 1 illustrates the average disposable incomes by age-period-cohort pro-
files. The profile of disposable income across different age groups takes the form of 
an upside-down U-shape, although those between 30-40 years of age have a slight 
lag before their income trajectory takes off  more rapidly. The cohort profiles show 
an aggregate trend in which persons born from 1940 to 1960 share almost the same 
level of disposable income. After this point, younger cohorts fall off  the horizontal, 
as expected, because their educations are in progress and their transition to the job 
market is not yet complete. When the reference person of the birth-year is 48 years 
old, the income trajectory is uniform against the period estimates.
A limitation of the dataset is the absence of a variable for gender. During the 
measured time period, Finland saw an increase in female labor force participation, 
which not only affected the incomes of households in which women were working 
but also the (relative) position of households in which women were not working. 
Taking this trend into account would be relevant.
Previous research has indicated that education plays a tremendous role in 
income development (OECD, 2015; Psacharopoulos, 1994). Educational expan-
sion explains a large part of income development; especially important are an 
increased amount of monetary support from the state and the availability of free 
higher education. Figure A.1 shows a clear shift from lower education to higher 
education. This educational expansion is relevant because it illustrates one of the 
key aspects of Finnish socio-political mechanisms. Education has been a major 
political investment, especially since the crisis of the 1990s, and has partly hastened 
the evolution of income development. In particular, the amount of polytechnic 
and tertiary education has risen significantly, which means that almost half  of the 
population has a higher education degree (Statistics Finland, 2015).
4. statistical ModEl
4.1. Age-Period-Cohort Conundrum
Age-period-cohort models are designed to estimate inter-cohort income tra-
jectories through the effects of age, period and cohort from two contextual view-
points: the relative and absolute contexts. Cohorts are the key research unit but 
also the main object of study in research on inter-cohort differences. Although 
many papers have proposed that answers lie in the empirical estimations and 
effects of age (a), period (p) and cohort (c), there is an underlying predicament 
with regard to retrieving results. One major problem has been the “identification 
problem,” which arises from the equation c=p−a. The variables are collinear 
to each other, so when two variables of age, period or cohort are known, the 
third is also known. Collinearity between regressors indicates that the statistical 
Review of Income and Wealth, Series 66, Number 2, June 2020
294
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
model produces an infinite number of possible solutions for the least squares or 
maximum likelihood estimators (Yang et al., 2004). Thus, the main dilemma is 
that the model does not hold a unique solution, and such a solution cannot be 
identified.
Age-period-cohort models are an attempt to grasp the identification problem. 
The model aims to explain outcomes through three components: the individual’s 
age a (훼a), membership in a cohort c (훾c) and statistical period p (휋p). Thus, the 
equation can be stated as follows:
The key function of the APC model is to detect “how” an outcome is explained by 
the position in the life cycle (age), date of birth (cohort) and the time of the statis-
tical measurement (period).
Previous research has solved the identification problem by imposing special 
restrictions on the model (Mannheim, 1928; Ryder, 1965; Mason et al., 1973; 
Hobcraft et al., 1982; Yang et al., 2004; O’Brien, 2011). These restrictions are 
built to constrain the coefficients of some variables. An example of this is the 
Constrained Generalized Linear Model estimator (CGLIM), which uses a theoret-
ical foundation wherein constraints use extra information to constrain coefficients 
based on theory. The CGLIM’s reliance on external information is problematic 
because such information often does not exist and CGLIM is sensitive to the choice 
of constraints, as Glenn (1976) notes. This problem is one of the reasons why the 
intrinsic estimator model (IE) was created by Yang et al. (2004, 2008).
The IE uses—as its core—principal component analysis to reduce the collin-
ear APC dimensions to a bidimensional plane. Additionally, this solution is criti-
cized by O’Brien (2011) and Luo (2013), who note that the intrinsic constraint is 
as arbitrary as in any other CGLIM. For example, O’Brien (2011) and Chauvel 
(2013) have shown that the model fails empirical tests such as that of detecting 
educational levels (see the discussion in Pelzer et al., 2015).
Instead of utilizing the “intrinsic” models, this paper uses a variation, the 
APC “detrended” and “trended” method, which was developed by Chauvel (2011, 
2012). These models have been used in empirical research and have produced reli-
able estimates (see Chauvel, 2010, 2013; Chancel, 2014; Chauvel and Scröder, 2015; 
Freedman, 2017).
4.2. Age-Period-Cohort “Detrended” and “Trended” models
The APCD model recognizes that because of the identification problem, 
linear trends in APC models cannot be robustly attributed to age, period and 
cohort. Instead, the “detrended” approach will focus on how the effects of age, 
period and cohort fluctuate around a linear trend, which this approach absorbs. 
In a more expressive manner, APCD is a “bump” detector that shows how dif-
ferent cohorts differ from the linear trend. The APCD model uses a set of con-
straints on the zero-sum, zero-slopes and on the domain of estimation of the 
cohort effects that excludes the first and the last cohort. Examples of excluded 
observations are the oldest age group of the first period and the youngest of 
(1) yapc=휇+훼a+휋p+훾c
Review of Income and Wealth, Series 66, Number 2, June 2020
295
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
the last period that appear once in the model. This method allows the model to 
become identifiable, obtain better estimates and provide a unique solution.
The APCD model can be illustrated with OSL expressions. In this model, yapc 
stands for the dependent variable but also for the independent variables of age (a), 
period (p) and cohort (c). The APC effects are the following: period effect 휋p fits 
the categorical period, 훼a is the coefficient for the non-linear age changes, 훾c is 
the estimates for the cohort effects, 훽0 denotes the general intercept, and 훽jxj are 
coefficients for the control variables. Rescale (a) and rescale (c) are linear functions 
that rescale the indexes a and c, which transforms the coefficients 훼0 and 휋0 in a 
standardized form to a scale of −1 to +1. Both rescale (a) and rescale (b) absorb the 
linear trend. Finally, if  훾c, as the cohort effect coefficient, is zero, this means that 
the cohort does not show any unique cohort-specific behavior and that it maintains 
homogenous behavior or effect. The detrended version of the age-period-cohort 
model is stated as follows:
The APCD model splits trends into two categories: the first category is based 
on the linear trend, and the second category is based on the non-linear trend on 
the fluctuations around the linear trend. The model works around the traditional 
identification dilemma by calculating the constrained trend with zero-sum and 
zero-slope parameters that are compared against a unique decomposition of a, p 
and c, which contain the estimated fluctuations (Chauvel, 2011). Briefly, when at 
least one coefficient has a difference to zero-slope coefficient, the APCD model will 
show the variations. When choosing the appropriate model between AP and the 
APC, it is practical to compare (Raftery, 1986) BIC values between the models to 
estimate which model has better explanatory power.
With disposable income as the dependent variable, there are multiple reasons 
to analyze deviations from the linear trends. The APCD distinguishes the relative 
share of period variations between cohorts but cannot take the unconformity of 
changing living standards into account in an absolute manner. Thus, the APCT 
is developed as an instrument to measure absolute declines and progressions. The 
main modification to the APCD model is to remove the zero-slope constraint from 
the cohort coefficients and thus suppress the 훾0 rescale(c)-term.
The period fluctuations are controlled in the same way in both APCD and 
APCT; however, rather than absorbing the long-term linear trend, the parameter 
훾c acts as a trended cohort effect in the APCT-model. This trended effect denotes 
per-cohort change while being controlled. Thus, the APCT will estimate the pro-
gression that controls for period fluctuations at a given age without being affected 
by the long-run period trends.
Hence, the APCT will illustrate how inflation-adjusted disposable income 
changes between cohorts but will not show whether the change arises from cohort 
(2)
⎧⎪⎪⎨⎪⎪⎩
yapc=𝛼a+𝜋p+𝛾c+𝛼0 rescale (a)+𝜋0 rescale (p)+𝛽0+
∑
j
𝛽jxj+𝜀i
⎧⎪⎨⎪⎩
∑
a
𝛼a=
∑
p
𝜋p=
∑
c
𝛾c=0
Slopea(𝛼)=Slopep(𝜋p)=Slopec(𝛾c)=0
min (c)< c<max (c)
Review of Income and Wealth, Series 66, Number 2, June 2020
296
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
effects or period effects. The model is flawed in the sense that it cannot reveal 
whether the trend will continue on its path or whether there is a chance for a long, 
overarching period change1 (see, e.g. Chauvel, 2016; Freedman, 2017). The APCT 
formula is defined as follows:
4.3. Age-Period-Cohort Estimations on Disposable Income
First, in both cases of APC-variants, we estimate an uncontrolled model 
of logarithmic disposable income for household reference person i of a (age), c 
(cohort) and p (period):
where μ0 is the intercept for the logarithmic disposable income, 훼a is the coefficient 
on age, 휋p is the period coefficient, 훾c is the cohort coefficient and 휀i is a random 
error with E(휀i) = 0. Thus, the disposable income of each household is predicted by 
the age of the reference person of the household, the date of birth and the statisti-
cal year, including the random error.
After measuring the uncontrolled effects, we introduce the control variables of 
education level, primary economic activity, number of persons in the household and 
unemployment level in the model, where 훽j is the coefficient for control variable xj:
Additionally, we estimate the attainment of middle-income MIi by using a 
dummy variable, where disposable income is split into quintiles in which QU 1−QU 2
equals 0 and QU 3−QU 5 equals 1. We do not use the logit model, which has meth-
odological problems. Instead, we use a linear model with a dependent dummy vari-
able, which is equivalent in results to the marginal effects (Mood, 2010). Thus, the 
uncontrolled and controlled APC-models are stated as follows:
1An APCH is a “hysteresis” model, which aims to solve these concerns by estimating the lasting 
scarring effects (Chauvel, 2016). However, a hysteresis model estimator yields only a summary estimate 
of variation and cannot help in identifying the shape of life cycle dynamics (see, e.g. Freedman, 2017).
(3)
⎧⎪⎪⎨⎪⎪⎩
yapc=𝛼a+𝜋p+𝛾c+𝛼0 rescale (a)+𝛽0+
∑
j
𝛽jxj+𝜀i
⎧⎪⎨⎪⎩
∑
a
𝛼a=
∑
p
𝜋p=
∑
c
𝛾c=0
Slopea(𝛼)=Slopep(𝜋p)=0
min (c)< c<max (c)
(4) log (DI )i
apc
=휇0+훼a+휋p+훾c+휀i
(5) log (DI )i
apc
=휇0+훼a+휋p+훾c+
∑
j
훽jxj+휀i
(6) E(MIi)
apc
=휇0+훼a+휋p+훾c+휀i
(7) E(MIi|Xj)apc=휇0+훼a+휋p+훾c+
∑
j
훽jxj+휀i
Review of Income and Wealth, Series 66, Number 2, June 2020
297
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
In conclusion, the identification principle in our study is to observe whether 
the age, period or cohort coefficients of the models statistically significantly devi-
ate from zero, which will confirm that there is an age, period or cohort effect on 
disposable income. In the controlled model, the effects that deviate from zero indi-
cate that there are effects on the three APC dimensions that do not depend on 
Figure 2. Results of APCT Models and Periodic Linear Regression (solid lines) of the Disposable 
Household Income (Reference period is 1971 instead of 1966 because education information is 
missing from the oldest dataset due to older data collection principles.) 
Note: Controlled models—both APCT and OLS—are adjusted by education, main economic 
activity, unemployment level and number of persons in a household. In OLS-model, vertical axis 
is linear regression coefficient (Results shown as the percentage change in relation to the reference 
group, which is calculated by logarithmic transformation of coefficients: (( exp (훽)−1)∗100)), 
and horizontal axis shows range of measurement period. In APCT-models, vertical axis is APCT 
coefficient, and horizontal axis shows range of each APC component. Dotted lines represent 95 
percent confidence intervals.
Review of Income and Wealth, Series 66, Number 2, June 2020
298
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
the control variables. Last, when observing age, period and cohort effects, we can 
determine whether control variables “suppress” one or more of these dimensions 
(see, e.g. Yang and Land, 2013). In an ideal case, this would isolate the effects 
towards one or two factors of the APC triad, giving the model more explanatory 
power and also providing an estimation of which control variable explains this 
change.
5. rEsults
5.1. “Trended” age-period-cohort effects
Figure 2 shows the results of trended age, period and cohort effects on 
income, where the periodic zero-sum trend is freed from constraints to show 
the absolute income trajectory. Thus, the APCT-model will reveal how absolute 
income has developed across age, period and cohort. The APCD model shows 
the same relative periodic effects as the APCT model. Thus, it does not provide 
any new information about periodic effects. This lack of new information is a 
limitation of the APCT period effects, which remain bound to the zero-sum 
trend. Therefore, we utilize a linear regression model, which illustrates how the 
periodic development of disposable income has changed in an absolute fashion. 
The limitation of this estimation is that the model is restricted to age and period, 
which means that it does not take into account possible cohort effects. The choice 
of the AP model over the CP model was verified with the VIF-test, which mea-
sured the AP model as less collinear, with VIF values under 5 instead of 10.
The periodic OLS-model reveals that overall absolute income has increased 
over time, although the economic shocks of the 1990s and the financial crisis of 
2007/08 have created a period of income stagnation. Absolute income rose steadily 
from 1976 to 1992 until the economic shock of the early 1990s. The effect of the 
shock can be observed from 1993 to 1997, when income levels dropped in the 
uncontrolled model. Additionally, in the controlled model, income level does not 
decrease, but its linear development stops and the trend stagnates, whereas there 
is only 1 percentage point of growth between the years 1987–1997. It is also evi-
dent that income development has been halted by the financial crisis. Levels of 
disposable income rose during the 2003–2013 period but did not rise in 2014. This 
stagnation was created by the financial crisis and can be compared to the economic 
stagnation of the 1990s because both periods show a growth level of 2 percentage 
points during the period of stagnation.
When comparing the relative changes (APCD-model) to the absolute changes 
(OLS-model), the periodic stagnation points show where income development has 
been halted. These phases indicate the positions of the life course in which dif-
ferent birth cohorts have faced different economic shocks. For example, negative 
estimates in both models for the birth cohorts of 1975–1995 could explain why 
income development has become negative. The reason is economic stagnation in 
the shock points that occurred at critical phases of the life course. In addition, the 
analysis shows that absolute income development did not stagnate at the time of 
the 1973 oil crisis. This indicates that that particular economic shock did not affect 
Review of Income and Wealth, Series 66, Number 2, June 2020
299
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
the income development of the older generations even though it occurred during 
the same phase of the life course.
The results of the trended analysis show that overall income has increased 
over time. Absolute income in Finland has been rising steadily, and inequalities are 
found more clearly in relative income (see, e.g. Suoniemi, 2012).
It is apparent that the disposable income of cohorts matches relatively well 
with general increases in the disposable income. The disposable income deviates 
from the general development of the disposable income (R2 value is 0.975). The 
APCT model with control variables included shows that cohort incomes remain 
close to general increases in disposable incomes. The model shows that the level 
of absolute income has been rising steadily from generation to generation. The 
R2 between general increases in incomes and cohort-based incomes is 0.979. The 
cohort effects show the rising income from older to younger birth cohorts, although 
the detrended effects show that there are relative cohort differences. This result 
indicates that income inequality in Finland is more relative than absolute, which is 
a result of rapid economic growth, a strong union presence in policy making and 
educational expansion (see, e.g. Suoniemi, 2012).
In regard to age effects, the non-linear slope follows a classic inverted U-shape 
slope, where income rises after graduation at the beginning of the life cycle and 
falls again in the retirement phase. After introducing control variables, the age 
effects follow a linear trend, showing a steady increase throughout the life course.
In the uncontrolled model of age, the most notable differences are found at 
the extremities of the slope. The group of 20- to 24-year-olds shows clear growth in 
income compared to 25- to 29-year-olds. The age effects cross the zero coefficient 
line at the age of 43 years, which lands right on the point of the middle-income 
level by age group in this dataset. In the uncontrolled model, the age effects stop 
growing within the age group of 50- to 70-year-olds, which is a period of transition 
into retirement. The controlled age effects are revealed in the effects of the APCD-
model in the context of periodic change. The deviations in the uncontrolled model 
are explained by the addition of control variables when the age effects form an 
almost perfect line (R2 = 0.99) with the linear trend, without any anomalies. In line 
with the APCD-model above, the effect of education and main economic activity 
reduce the differences in all of the age groups (Table A.2). The income develop-
ment between age groups shows an ideal trajectory in the absolute sense, whereas 
the intergenerational distributions are identified in the cohort effects, which show 
income variance in the controlled model.
In relation to the APCT-model, it can be stated that absolute disposable 
income has, overall, been rising in the dimensions of age, period and cohort. This 
result indicates periodic development of income in the dimensions of time and 
birth cohorts, whereas individuals are expected to see rises in income through the 
life cycle. Although the development of income has stagnated due to economic 
shocks, this stagnation has not halted the overall growth trend. The trended model 
shows which age, period and cohort groups have the highest absolute income but 
does not reveal how income has changed in a relative sense, which we will address 
next with the trended model.
Review of Income and Wealth, Series 66, Number 2, June 2020
300
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
5.2. “Detrended” age-period-cohort effects
Figure 3 shows the results of the age-period-cohort “detrended” analysis. 
The analysis shows relative changes in income in relation to the linear trend, 
revealing which age, period or cohort category benefitted most compared to 
other groups. This analysis offers an answer to the research question on rela-
tive income in three different ways: which cohort group has the more privileged 
Figure 3. Results of APCD Models (solid lines) of the Disposable Household Income 
Note: (Statistical years were coded as representative 5-year periods for methodological reasons (see Section 
3). All periods are presented as actual statistical years in all figures. Chosen period years are centered on the 
technical period representative year and to match known economic cycles in Finnish history.) Controlled 
model is adjusted by education, main economic activity, unemployment level and number of persons 
in household. Vertical axis shows APCD coefficient, and horizontal axis shows range of each APC 
component. Dotted lines represent 95 percent confidence intervals.
Review of Income and Wealth, Series 66, Number 2, June 2020
301
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
position, how age variations reveal overall life course effects on income, and 
whether certain periodic shocks negatively affected the progress of incomes.
The uncontrolled period effects show the relative effects of economic fluctu-
ations. The periodic effects illuminate the dramatic effect of the 1990s economic 
downturn. Additionally, the most recent data point shows a relative decline in 
incomes under the linear trend. This same effect can be observed in the controlled 
model, in which the nonlinear slope maintains the same overall shape. The relative 
estimates show that the economic shock of the early 1990s still maintains its posi-
tion as the most severe influence on income development. In addition, the lowest 
coefficient is in 1971.
The periodic coefficients are consistent with the historical development of 
income inequality in Finland. Income inequality declined in the early 1970s, which, 
in turn, is shown as a rapid increase in periodic coefficients in the APCD-model. 
Subsequently, inequality remained almost constant until the economic shock of 
the 1990s. This effect can be observed in 1981 with no deviation from the linear 
trend and with a steady rise in incomes until the economic downturn of the early 
1990s. The periodic effects show a substantial decline in incomes to negative levels, 
under the linear trend. In the first five years after the economic downturn, income 
inequality rose but steadily recovered to present levels at the turn of the millen-
nium. Additionally, the APCD-model shows that income levels increased from 
negative to positive, crossing the linear trend. After the depression of the early 
1990s, the periodic income effects rise above the linear trend during the period of 
2003–2007 and continue their rise until the period of 2008–2013. After that, the 
periodic effects decline below the linear trend by 2014–2015.
In addition to the economic cycles, the period effects are explained by the 
educational expansion that was illustrated in Figure A.1. Both the uncontrolled 
and controlled period models show a profound rise in income during the period 
from 1971 to 1976. When control variables are absorbed, the impact of the periods 
decreases notably, which can be explained by the increased incomes of the rela-
tively small group with higher education during that time period. In contrast, the 
period between 1981 and 1992 has the most dramatic “spike” after the controls, 
which could be explained mainly by economic growth.
The results of the cohort effects show that the baby boomer generation 
indeed benefited the most from their birth cohort, whereas younger generations 
are falling behind (see also Berloffa and Villa, 2010; Chauvel and Schröder, 2015). 
Additionally, these findings are connected to periodic shock effects, which sup-
ports the hypothesis that cohorts that enter the labor market during economic 
downturns are in a more disadvantaged position than cohorts that enter the labor 
market during economic upturns.
The cohort effect without controls shows a peak in the birth years from 1945 
to 1949, which deviates by approximately 11 percent from the linear trend. The 
strongest negative effects are located at the extremes of the cohort distribution, 
on the birth cohorts of 1915–1919 (-10 percent) and 1990–1995 (-15 percent). The 
point at which the older cohorts crossover from negative to positive is seen in the 
birth cohort of 1930–1934, and the decline back to negative coefficients is seen in 
the birth cohort of 1980–1985.
Review of Income and Wealth, Series 66, Number 2, June 2020
302
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
When education, primary economic activity, household size and unemploy-
ment are controlled, the peak of income widens and is located between the birth 
cohorts of 1940–1949. The most influential change is the drop between the birth 
cohorts of 1955–1979. These cohorts more closely follow the linear trend, with an 
effect on the oldest cohort of 4 percent and an effect on the youngest cohort of -0.3 
percent. Controls also affect the youngest generation, where the strong negative 
slope is evened out towards the linear trend. For example, the effects on the birth 
cohort of 1990–1994 are halved in the controlled model. Thus, when the control 
variables are taken into account, the youngest generation does not have the max-
imum negative deviation, which is now located in the older generations, although 
the number of observations makes the estimates more unstable.
The interpretation of relative income decline in the younger generations is 
most likely connected to economic shocks. The timing of the Finnish economic 
downturn of the early 1990s and the financial crisis of 2007/08 had particular 
effects on the birth cohorts that were born in 1975 and later. For example, the 
birth cohort of 1975–1980 was at its earliest point of transition into the job market 
during the economic downturn of the 1990s. For this cohort, the financial cri-
sis—which occurred when this group was 28-33 years of age—is also an important 
factor for income development. Additionally, the younger cohorts are affected by 
the indirect effects of economic shock through their parents.
In light of these findings, it is plausible that the children of older generations 
will have lower relative income than their parents (see also Chetty et al., 2017). 
Regarding relative income, the results show that the birth cohorts of 1940 to 1949 
enjoy the most lucrative generational position. This can be explained by the period 
of high economic growth, expanding educational opportunities, abundance of 
available jobs and—foremost—the lack of periodic economic shocks, all of which 
have contributed to positive income development.
The main findings with regard to age effects are important for the APC-model 
because they reveal the effects of age and cohort as dominant components of rel-
ative income, whereas the dominant economic activity functions as a suppressive 
control mechanism (see Table A.2). This finding shows that relative income effects 
are due to periodic and inter-cohort changes, whereas age effects do not play a 
major role in this dynamic. Additionally, in contrast to what previous research has 
indicated, education, as a control variable, does not have a primary impact on the 
age effects.
The age effects show a convex life course, where middle-age acts as an income 
peak and the youngest and oldest age groups are the most disadvantaged in terms 
of income levels. Interestingly, the uncontrolled model shows a stagnation phase 
between the ages of 30 and 40. The estimates show an increase in incomes after 
entry into the labor market. After this transition, income growth slows down rela-
tive to the linear trend.
Introduction of control variables into the model creates a drastic change for 
the age effects. The main economic activity suppresses the age coefficients, such that 
they follow the linear trend. Education itself  does not possess the power to nullify 
the relative age effect: even if  the main economic activity is introduced without 
education, the coefficients follow the linear trend. Still, minor deviations remain. 
The youngest age group is below the linear trend, but the level of income increases 
Review of Income and Wealth, Series 66, Number 2, June 2020
303
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
steadily between the ages of 25 and 34. After this point, income will adhere to the 
linear trend, and later, it crosses the zero line, entering the negative region during 
the period of transition to pension. In the uncontrolled model, the relative peak of 
incomes is located in the age group of 50- to 54-year-olds. When controls are intro-
duced, it seems that young people at age 30 to 34 are to some extent in a more ben-
eficial position than other age groups, yet the estimates are modest (See Table A.2).
This suppression of age coefficients offers interesting insights because it iden-
tifies the main explanatory effect of cohort and periodic outcomes and provides an 
Figure 4. Results of APCD Models on Attainment of Middle Income (0/1). 
Note: Controlled model is adjusted by education, main economic activity, unemployment level and 
number of persons in household. Vertical axis shows APCD coefficient, and horizontal axis shows 
range of each APC component. Dotted lines represent 95 percent confidence intervals.
Review of Income and Wealth, Series 66, Number 2, June 2020
304
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
explanation of how income inequality varies in the triad of age, period and cohort. 
Regarding relative income, the analysis suggests that a person’s birth cohort and 
the effects of economic cycles during the life course—not belonging to a certain 
age group—are the factors that indicate changes in income distribution. Still, the 
detrended and trended approaches do not answer the question of whether and 
how inter-cohort inequalities change in the context of achieving average income 
class. The following section will put relative and absolute income changes into 
the context of achieving “optimal” income class by measuring the attainment of 
middle-income.
5.3. Attainment of middle-income over age, period and cohort
Figure 4 shows the results of the probability of attaining middle-income, 
which will help us put our previous results into the context of achieving “opti-
mal” and societally valued income class. The periodic results show that mid-
dle-income attainment is not strongly connected to periodic income changes. 
This finding suggests that economic cycles do not have a dramatic effect on the 
probability of achieving middle-income status, which identifies the dimensions 
of cohort and age as the main components of middle-income attainment.
The periodic changes in the uncontrolled model shows a moderate U-shape 
pattern. The minimum periodic effect is at its most intensive level during the eco-
nomic shock of the early 1990s (-2.3 percent in 1995). The most beneficial times at 
which to achieve middle income were in the early 1970s and from 2008 onwards, 
where periodic effects were approximately 3 percent. The periodic effects are rather 
weak from the year 1985 onward and closely follow the linear trend, excluding the 
slight deviation in the period 1993–1997. In conclusion, the controlled model indi-
cates that the year 1970 was the optimal time to attain middle-income, and the worst 
time was in 1980, with a difference of 8 percentage points between these two years.
We conducted robustness tests on our measure of middle-income attainment by 
calculating estimates for the upper 5th quintile and the quintile range of QU 3−QU 4 
(not reported here), which revealed that the periodic anomaly during the period 
of 1976–1985 was caused by the upper 5th quintile. By taking this deviation into 
account, the periodic effects follow the reference trend almost completely. In other 
words, these results indicate that age and cohort effects dominate the period effects 
in middle-income attainment: all of the differences in middle-income attainment 
are identified as age or cohort effects. The periodic anomaly of the 5th quintile is 
explained by low income inequality during the growth years of 1970–1984. Taking 
these deviations into account, middle-income attainment has been rather stable in 
relation to the linear trend, whereas the effects stay weak.
Middle-income attainment and cohort effects have overlapping results related 
to relative income, where higher income raises the probability of a certain birth 
cohort being among the group of individuals who achieve middle-income. Again, 
the results show that the baby boomers are the generation in the most beneficial 
position with regard to attaining middle-income. As previously stated, period 
effects are suppressed after introducing control variables, which indicates that the 
impact on the attainment of middle-income has a strong foundation in genera-
tional effects.
Review of Income and Wealth, Series 66, Number 2, June 2020
305
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
The cohort effects on the uncontrolled model have analogous peaks, as in the 
previous models that measured logarithmic income as a dependent variable (see 
Figure 3). The highest probability of reaching middle-income is found in the birth 
cohort of 1945–1949, with the deviation from the linear trend being 10 percent. 
The cohort effects do not change their intensity between the uncontrolled and con-
trolled models, although the range of the optimal birth cohort for middle-income 
attainment expands. However, the birth cohort of 1945–1949 is the most advan-
taged cohort for reaching middle-income—with a 10 percent probability—but the 
peak is greater, even where the birth cohorts of 1935–1950 are relatively close to 
each other. In contrast, the most disadvantaged cohort is, unsurprisingly, the birth 
cohort of 1990–1995, with a -10 percent deviation. In the controlled model, the 
lowest point has drastically shifted to the other end of spectrum, where the oldest 
birth cohort of 1910–1914 has a minimum coefficient of -14 percent from the linear 
trend. The birth cohort of 1970–1974 is at the zero-sum line, which marks the point 
after which all birth cohorts have a negative probability of middle-income attain-
ment; this point is 5 years earlier than in the uncontrolled model. Additionally, the 
cohorts from 1980 onwards show a major change—in a downward trend—com-
pared to the cohorts of 1955–1979.
This downward change could be explained by the economic shocks of the 
early 1990s and the financial crisis of 2007/08. First, the recession of the early 1990s 
could have an indirect impact through parental effects, and after this, the financial 
crisis is timed precisely when the young cohorts of 1985–1990 are supposed to 
enter the job market. The results also indicate that cohorts from 1975 onwards 
face more difficulties in reaching middle-income compared with their older coun-
terparts. It is to be noted that our tests indicated that there were no clear periodic 
effects on attainment of middle-income, which means that economic shocks do not 
work as a clear explanatory framework. Instead, it is likely that economic shocks 
impact income levels, which will have effects on the cohorts or age groups that are 
more likely to achieve middle-income. Attaining middle-income is associated with 
the income levels of cohorts: when a certain cohort’s income is high, this raises the 
probability that that cohort will surpass middle-income. In conclusion, it can be 
stated that individual choices about education and placement in the labor market 
dictate possible income margins, which will determine how likely it is for a certain 
age and cohort group to attain middle-income; however, all time periods are equal 
with regard to attaining middle-income status.
The age effect has an inverted U-shape, which indicates that the attainment 
of middle-income rises steadily from age 20 (-23 percent) to 35 (7 percent) and 
then rises at a more nominal level. After that, the attainment level stays the same, 
although the high point is observed among 50- to 54-year-olds, at 10 percent above 
the linear trend. In other words, middle-income attainment increases rapidly from 
a young age after graduation and remains relatively the same, with a slight rising 
trend through the later years. Finally, the transition to retirement entails dropping 
out of middle-income attainment (See also Riihelä, 2006; Nummi et al., 2012).
The controlled age effects show a moderate version of the reverse U-shape 
of the uncontrolled model. The model shows how age effects are relatively stable 
during the life course and how the deviations from the linear trend follow life cycle 
changes, such as education and retirement. Having middle-income status is more 
Review of Income and Wealth, Series 66, Number 2, June 2020
306
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
probable when one is at the active working stage of the life cycle. The effects are 
weak and follow the linear trend rather closely, which shows that middle-income 
attainment is not heavily age-related in Finland.
The overall consensus of this analysis is that middle-income attainment in 
Finland is most likely for the baby boomer generation, which is related to that 
generation’s higher relative income. Additionally, the results show that periodic 
changes do not influence the probability of achieving middle-income status, which 
is related more strongly to life cycle and cohort effects. The spikes in periodic effects 
are fully explained by the affluence of the 5th quintile, and age effects do not have a 
strong impact on middle-income attainment compared to cohort effects.
6. conclusions
The main objective of this study was to assess how economic shocks affect 
income trends and inter-cohort income dynamics. The findings suggest that 
period and cohort effects have the main effects on relative income in the con-
text of APC-method, whereas age effects have no meaningful impact when con-
trol variables are taken into account. This result reveals a clear emphasis on the 
relationship between the effects of economic shocks and cohort placement on 
labor market entry, whereas the occupational position attained explains all of 
the age variations. For example, those who were born in 1970 were entering the 
labor market at the brink of the economic shock of 1993 (at the age of 23) and 
have no income development in relation to the linear trend; in addition, all of the 
cohorts born after 1975 have a negative relative income effect. Additionally, abso-
lute income analysis reveals that economic shocks create stagnation points in 
income development, which, with regard to cohort incomes, are especially detri-
mental to cohorts who are in the process of transitioning into the labor markets. 
Whereas older cohorts experienced no such effect during the oil crisis, younger 
generations seem to suffer birth-positional disadvantages from the economic 
stagnation caused by both the periodic shock of the 1990s and the financial cri-
sis. Additionally, our research illustrates how middle-income attainment has not 
changed due to periodic shocks but rather is related to inter-cohort inequalities 
and relative income differences, where the baby boomer generation is a clear 
winner. Regarding economic shocks, the oil crisis in Finland did not have the 
same impact on cohort incomes as did the 1990s economic depression and the 
financial crisis of 2007/08.
The unique severity of the Finnish economic crisis of the 1990s highlights 
how income trajectories can change in fairly short order, especially in the inter-
national context. Compared to other Nordic countries, Finland did suffer from a 
significant drop in GDP per capita during the economic depression of the 1990s, 
whereas after the financial crisis, Finland struggled to end economic stagnation. 
This situation called for stronger economic and social policy solutions, which will 
in return impact income trends and especially those people who are located on 
distribution tails. Compared to a previous cohort study (Osberg, 2003), Finland 
seems to have a similar effect on declining incomes in Generation X as Sweden and 
Germany during the 1990s, although the data range is limited to year 1995. The 
Review of Income and Wealth, Series 66, Number 2, June 2020
307
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
study does not answer why between-cohort inequalities are rising but speculates on 
the importance of possible institutional context on market forces.
Our results show that the younger cohorts are in a more disadvantaged eco-
nomic position than the older “baby boomer” cohorts in the case of relative income 
and attainment of middle-income but not in the case of absolute income. In our 
study, the main economic activity is the key explanatory factor. The result is partic-
ularly interesting because previous studies have emphasized the role of educational 
expansion, which is associated with the profits gained from certain educational 
fields (Pekkala and Lucas, 2004). This finding supports, to some extent, a signal-
ing theory (Spence, 1973). Education plays a role in occupational attainment, but 
education itself  does not suppress the age effects in the same way that the main 
economic activity does. In other words, gained positional outcome in the labor 
market is more essential, but a solid foundation in education is needed to reach 
positions that offer higher income. These positions or “outcomes” are therefore 
more directly influential than education itself.
In addition, the time period of 1945–1980 was a time of great economic 
growth. The cohorts of 1930–1950 entered the labor market during the 1960s and 
1970s, which was a time of economic prosperity and high employment. These 
labor market changes and increased educational opportunities formed an optimal 
platform for social mobility: higher education offered high returns in a growing 
labor market, where the supply and demand ratio for the educated workforce was 
in favor of educated workers. The younger generations, in turn, have been facing 
the economic recession of the early 1990s and the financial crisis of 2007/08. The 
scarring effects on these cohorts are visible in the cohort effects of lower relative 
income and middle-income attainment, where those born later than the 1970s do 
not share similar positions and opportunities as earlier cohorts.
Methodologically, APC-modeling offers a unique perspective on the life 
course theory of Meyer (2005), which emphasizes the importance of cohorts as an 
object of research. Our research suggests that the assertion is valid: in every model, 
cohort variations are strong, which suggests the relevance of inter-cohort effects 
on income that are affected through periodic macro-economic shocks. The APC 
framework in our study takes a different methodological perspective compared 
to previous research. APC-models are useful when measuring the “dominance” 
effects among age, period and cohort when different control variables are taken 
into account (see, e.g. Yang and Land, 2013). Previous studies based on Chauvel’s 
(2011) APC-model have not been utilized to identify which combinations of age, 
period or cohort are meaningful to the dependent variable or to uncover the main 
independent variable that is responsible for this “suppression” effect (e.g. Chauvel, 
2013; Chauvel and Schröder, 2015; Freedman, 2017). From a theoretical per-
spective, by identifying the “suppression” effect on cohorts, our finding supports 
Meyers (2005) notion about the importance of cohorts in life course research. Our 
study used this point of reference to gain access to the substance of economic 
cycles and to see whether income fluctuations are due to shock effects. Thus, taking 
into account modeling limitations, this methodological approach offered a major 
advantage in terms of untangling the age-period-cohort conundrum with regard to 
income, whereas traditional models only utilize two factors out of the three APC 
components.
Review of Income and Wealth, Series 66, Number 2, June 2020
308
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
In terms of policy implications, the paper has emphasized the importance of 
inequalities between cohorts in the context of economic cycles. We suggest that 
policy instruments should be aimed at younger generations—especially those who 
are entering in labor markets—to alleviate negative effects of economic shocks. 
Potential solutions could include taxation reforms where income transfers are 
aimed at younger age groups to support employment, similar to what Atkinson 
(2015) suggests to control economic outcomes. Additionally, governments could 
support employers for hiring early-career workers by implementing market support 
mechanism to finance a part of wage expenditures. In further research, it would be 
fruitful to use individual-level register-based datasets in a panel framework and to 
use microsimulation models to study how different policy models affect inter-cohort 
income distribution, which could offer some societal implications on how welfare 
policies could be optimized to also combat intergenerational income erosion.
rEFErEncEs
Atkinson, T., Inequality: What Can Be Done? Harvard University Press, Cambridge, MA, 2015.
Beaudry, P. and D. Green, “Cohort Patterns in Canadian Earnings: Assessing the Role of Skill 
Premia in Inequality Trends,” Canadian Journal of Economics, 33, 907–36, 2000.
Berloffa, G. and P. Villa, “Differences in Equivalent Income Across Cohorts of Households: 
Evidence From Italy,” Review of Income and Wealth, 56, 693–714, 2010.
Blomgren, J., H. Hiilamo, O. Kangas, and M. Niemelä, “Finland Growing Inequality with Contested 
Consequences,” in B. Nolan, W. Salverda, D. Checchi, I. Marx, A. McKnight, G. Tóth, and H. 
G. V. D. Werfhorst (eds), Changing Inequalities & Societal Impacts in Rich Countries. Thirty 
Countries’ Experiences, Oxford University Press, Oxford, 222–47, 2014.
Campbell, A., P. E. Converse, and W. L. Rodgers, The Quality of American Life. Perceptions, 
Evaluations, and Satisfactions, Russel Sage Foundation, New York, 1976.
Chancel, L., “Are Younger Generations Higher Carbon Emitters than their Elders? Inequalities, 
Generations and CO2 Emissions in France and in the USA,” Ecological Economics, 100, 195–
207, 2014.
Chauvel, L., Age-Period-Cohort Detrended APC-D Model, Institut d’Etudes Politiques, Paris, 2011.
   , APCD: Stata Module for Estimating Age Period Cohort Effects with Detrended Coefficients, 
Age–Period–Cohort Detrended APC-D Model, Statistical Software Components, College 
Department of Economics, Boston, MA, 2012.
   , Age-Period-Cohort with Hysteresis APC-H Model / A Method, University of Luxembourg, 
Luxembourg, 2016.
   , Le Destin des générations: structure sociale et cohortes en France du XX siècle aux années 
2010, PUF, Paris, 2010.
Chauvel, L. and M. Schroeder, “The Impact of Cohort Membership on Disposable Incomes in West 
Germany, France, and the United States,” European Sociological Review, 31, 298–311, 2015.
Chauvel, L., “Welfare Regimes, Cohorts and the Middle Classes,” in J. C. Gornick and M. Jäntti 
(eds), Income Inequality: Economic Disparities and the Middle Class in Affluent Countries 
(Studies in Social Inequality), Stanford University Press, Palo Alto, CA, 115–41, 2013.
Chetty, R., D. Grunsky, M. Hell, N. Hendren, R. Manduca, and J. Narag, “The Fading American 
Dream: Trends in Absolute Income Mobility Since 1940,” Science, 356, 398–406, 2017.
Clark, A. E. and A. Oswald, “Unhappiness and Unemployment,” Economic Journal, 104, 648–59, 
1994.
Crawford, R., D. Innes, and C. O’Dea, The Evolution of Wealth in Great Britain: 2006–08 to 2010–
12, Report No. R109, Institute for Fiscal Studies, London, 2015.
Daley, J. and D. Wood, The Wealth of Generations, Grattan Institute, Carlton, 14–5, 2014.
Elder, G. H., Jr., “The Life Course and Human Development,” in R. M. Lerner (ed.), Handbook of 
Child Psychology, Vol. 1: Theoretical Models of Human Development, John Wiley, New York, 
939–91, 1997.
Fitzenberger, B., R. Hujer, T. MaCurdy, and R. Schnabel, “Testing for Uniform Wage Trends in 
West-Germany: A Cohort Analysis Using Quintile Regressions for Censored Data,” Empirical 
Economics, 26, 41–86, 2001.
Review of Income and Wealth, Series 66, Number 2, June 2020
309
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
Freedman, M., Are Recent Generations Catching Up or Falling Behind? Trends in Inter-Generational 
Inequality, LIS Working Papers No. 689, LIS Cross-National Data Center in Luxembourg, 2017.
Frey, B. and A. Stutzer, “The Economics of Happiness,” World Economics, 3, 1–17, 2002.
Glenn, N. D., “Cohort Analysts’ Futile Quest: Statistical Attempts to Separate Age, Period, and 
Cohort Effects,” American Sociological Review, 41, 900–4, 1976.
Gosling, A., S. Machin, and C. Meghir, “The Changing Distribution of Male Wages in the U.K,” 
Review of Economic Studies, 67, 635–66, 2000.
Grenier, G., In Which Year Is It Better to Be Born? The Earnings of Men and Women in Canada 
During a Quarter of a Century, Department of Economics, University of Ottawa, Working 
Paper No. 0102E, 2003.
Helliwell, J. F., “How’s Life? Combining Individual and National Variables to Explain Subjective 
Well-Being,” Economic Modeling, 20, 331–60, 2003.
Hobcraft, J. N., J. Menken, and S. H. Preston, “Age, Period, and Cohort Effects in Demography: A 
Review,” Population Index, 48, 4–43, 1982.
Jäntti, M., O. Kangas, and V.-M. Ritakallio, “From Marginalism to Institutionalism: Distributional 
Consequences of the Transformation of the Finnish Pension Regime,” Review of Income and 
Wealth, 42, 478–91, 1996.
Jappelli, T., “The Age-Wealth Profile and The Life-Cycle Hypothesis: A Cohort Analysis with a 
Time Series of Cross-Sections of Italian Households,” Review of Income and Wealth, 45, 57–76, 
1999.
Kahneman, D., E. Diener, and N. Schwarz (eds), Well-Being: The Foundations of Hedonic Psychology. 
Russell Sage Foundation, New York, 1999.
Kangas, O. and J. Palme, “Does Social Policy Matter? Poverty Cycles in OECD Countries,” 
International Journal of Health Services, 30, 335–52, 2000.
Kershaw, P., “Population Aging, Generational Equity & The Middle Class,” Generation Squeeze, 
Childcare Resource and Research Unit, 4–20, 2015.
Layard, R., A. E. Clark, and C. Senik, “The Causes of Happiness and Misery,” in J. Helliwell, R. 
Layard, and J. Sachs (eds), World Happiness Report, The Earth Institute, Columbia University, 
CIFAR, and CEP, 58–89, 2012.
Lim, G. C. and Q. Zeng, “Consumption, Income, and Wealth: Evidence from Age, Cohort, and 
Period Elasticities,” Review of Income and Wealth, 62, 489–508, 2016.
Luo, L., “Assessing Validity and Application Scope of the Intrinsic Estimator Approach to the Age-
Period-Cohort Problem,” Demography, 50, 1945–67, 2013.
Luo, L., J. Hodges, C. Winship, and D. Powers, “The Sensitivity of the Intrinsic Estimator to Coding 
Schemes: Comment on Yang, Schulhofer-Wohl, Fu, and Land,” American Journal of Sociology, 
122, 930–61, 2016.
Malmendier, U. and G. Nagel, “Depression Babies: Do Macroeconomic Experiences Affect Risk 
Taking?” Quarterly Journal of Economics, 126, 373–416, 2011.
Mannheim, K., “Das Problem der Generationen,” Kölner Vierteljahreshefte für Soziologie, 7, 157–
85, 309–30, 1928.
Mason, K., W. Mason, H. Winsborough, and W. Poole, “Some Methodological Issues in Cohort 
Analysis of Archival Data,” American Sociological Review, 26, 67–82, 1973.
Mayer, K. U., “Life Courses and Life Chances in a Comparative Perspective,” in S. Svallfors (ed.), 
Analyzing Inequality: Life Chances and Social Mobility in Comparative Perspective, Stanford 
University Press, Palo Alto, CA, 17–55, 2005.
Mood, C., “Logistic Regression: Why We Cannot Do What We Think We Can Do and What We 
Can Do About It,” European Sociological Review, 26, 67–82, 2010.
Myles, J., “A New Social Contract for the Elderly?,” in Gosta Esping-Andersen (ed.), Why We Need 
A New Welfare State, Oxford University Press, 130–72, 2002.
Nummi, T., J. Salonen, and L. Koskinen, Earnings Profiles of Finnish Wage Earners in 2000–2010, 
Finnish Centre for Pensions, Working Paper 5, 2012.
O’Brien, R., “Constrained Estimators and Age Period Cohorts Models,” Sociological Methods and 
Research, 40, 419–52, 2011.
OECD, Growing Inequal: Income Distribution and Poverty in OECD Countries, OECD Publishing, 
Paris, 2008.
   , Divided We Stand: Why Inequality Keeps Rising, OECD Publishing, Paris, 2011.
   , Quality Review of the OECD Database on Household Incomes and Poverty and the OECD 
Earnings Database Part I, OECD publishing, Paris, 2012.
   , Education at a Glance 2015: OECD Indicators, OECD Publishing, Paris, 2015.
Official Statistics of Finland, Statistical Yearbook of Finland 1952–2015, Statistics Finland, 
Helsinki, 2015.
Review of Income and Wealth, Series 66, Number 2, June 2020
310
© 2019 The Authors Review of Income and Wealth published by John Wiley & Sons Ltd on behalf 
of International Association for Research in Income and Wealth
Osberg, L., “Long Run Trends in Income Inequality in the United States, UK, Sweden, Germany, 
and Canada: A Birth Cohort View,” Eastern Economic Journal, 29, 121–41, 2003.
Ostry, J., A. Berg, G. Tsangarides, Redistribution, Inequality, and Growth, IMF, Washington, D.C., 
2014.
Pekkala, S. and E. Berman, Natural Born Winners: Rising Returns to Education and Background, 
Boston University, Mimeo, Boston, MA, 2002.
Pekkala, S. and R. E. B. Lucas, On the Importance of Finishing School: Half a Century of Inter-
generational Economic Mobility in Finland, Discussion Papers 141, Institute for Economic 
Development, Boston University, 2004.
   , “Differences Across Cohorts in Finnish Intergenerational Income Mobility,” Industrial 
Relations, 46, 81–111, 2007.
Pelzer, B., M. te Grontenhuis, R. Eisinga, and A. W. Schmidt-Catran, “The Non-Uniqueness 
Property of the Intrinsic Estimator in APC Models,” Demography, 52, 315–27, 2015.
Psacharopoulos, G., “Returns to Investment in Education: A Global Update,” World Development, 
22, 1325–43, 1994.
Raftery, R. E., “Choosing Models for Cross-Classifications,” American Sociological Review, 51, 
145–6, 1986.
Riihelä, M., Kotitalouksien kulutus ja säästäminen: ikäprofiilien ja kohorttien kuvaus, Goverment 
Institute for Economic Research, 386, 2006.
Ryder, N., “The Cohort as a Concept in the Study of Social Change,” American Sociological Review, 
30, 843–61, 1965.
Smeeding, T. and D. Sullivan, “Generations and the Distribution of Economic Well Being: A Cross- 
National View,” American Economic Review, 88, 254–8, 1998.
Spence, M., “Job Market Signaling,” Quarterly Journal of Economics, 87, 355–74, 1973.
Suoniemi, I., Income Mobility, Income Risk and Age, Finnish Centre for Pensions, Working Papers, 
Labour Institute for Economic Research, Working Papers, 2012.
Yang, Y. and K. C. Land, Age-Period-Cohort Analysis: New Models, Methods, and Empirical 
Applications, Chapman & Hall/CRC, Boca Raton, FL, 2013.
Yang, Y., W. J. Fu, and K. C. Land, “A Methodological Comparison of Age–Period–Cohort Models: 
Intrinsic Estimator and Conventional Generalized Linear Models,” Sociological Methodology, 
34, 75–110, 2004.
Yang, Y., S. Schulhofer-Wohl, F. J. Wenjiang, and K. C. Land, “The Intrinsic Estimator for Age–
Period–Cohort Analysis: What It Is and How to Use It,” American Journal of Sociology, 113, 
1697–736, 2008.
Supporting Information
Additional supporting information may be found in the online version of 
this article at the publisher’s web site:
Figure A.1: Share of Individuals Between Periods by Educational Level in 
the Data Sample
Table A.1: Descriptive Statistics
Table A.2: Trended and Detrended Age-Period-Cohort Effects; 
AP-OLS-Models