Transcript Aucun titre de diapositive - Paris School of Economics

Plan of the course
Income and subjective well-being
•
I) Absolute income and the Easterlin paradox
•
II) Relative income (comparisons)
•
III) Adaptation- Expectations
•
IV) Inequality
1
I) The Easterlin paradox
Will raising the incomes of all raise the happiness
of all ?
• GDP is the objective of economic policy and development policy
• But the ultimate goal of public policy is welfare, not income
• So, does GDP proxy the different dimensions of welfare?
 Health, life expectancy, security, capacities, freedom of
choice, human rights, quality of life…
• Answer this question  using Subjective Well-Being data
• This meets the discussion about the measures of well-being (e.g.
the Sen, Stiglitz Report for President Sarkoy)
A) The Easterlin paradox: stylized facts
• Paradoxical relationship between income growth and
subjective happiness:
Within country
Across countries
Over time at the individual level
Over time in average across countries
1) In a given country, richer individuals are
happier and more satisfied with their lives
Source : WVS, China 2007.
OLS Esstimate of Happiness in USA
(General Social Survey, 2006)
Happiness = 0,2 log(income). Individuals aged 25-65, earning more than 5000$.
Happiness and log household income
American General Social Survey (Stevenson and
Wolfers, 2008)
Source : WVS, China 2007.
Sacks, Stevenson and Wolfers (2010)
7
“In every representative national survey ever done, a
significant bivariate relationship between happiness and
income has been found” (Easterlin 2005)
• Western developed countries: German Socio-Economic Panel (GSOEP), British
Household Panel Survey (BHPS), Swiss household panel, Australian household
survey (HILDA), General Social Survey (America), Netherlands, Denmark….
• Transition countries: Albania, Bulgaria, Latvia, Romania Russia, Estonia,
Lithuania, Hungary, Belarus, Poland, Ukraine, Life in Transition Survey (LITS,
2006)....
• Asia: China, India, Shanghai, South Korea.
• Africa and Middle-East: Ethiopia, Nigeria, Peru, South-Africa (SALDRU),
Tanzania, Turkey
• Latin America: Argentina, Brazil, Chile Mexico, Venezuela.
• International surveys: World Values Survey (1981- 2008, 5 waves, 105
countries), International Social Survey Program (101 countries), Gallup World
Poll (2006, 105 countries), Latino Barometer (18 countries), European Social
Survey (25 countries), European Values Survey, Euro-barometer.
2) People are happier and more satisfied with
their lives in higher-income countries
Source :
Inglehart,
Foa,
Peterson,
Welzel
(2008)
GDP per Capita and Life Satisfaction in the
World. Cross-sectional evidence
Deaton
(2008, p57)
GDP and Average Life Satisfaction in the
World in the 2000’s
WVS in the 2000’s. Last available year for each country.
Sacks, Stevenson et Wolfers (2010)
12
3) Individuals become happier as they grow
richer
• Dynamic relationship
• Individual Panel Data in Developed Countries
GSOEP, BHPS, HILDA, Netherlands, Denmark
• Individual Panel data in Low Income Countries
RLMS (Russia), ULMS (Ukraine), Peru, LSMS
(Tajikistan)
4) But in average, on the long run, people do not
become happier over time, when national income
increases
Happiness and Real GDP per Capita, United States,
1972-2002 (Easterlin and Angelescu, 2007)
Japan
15
USA
Europe
16
In summary
• Cross-country analysis: positive relation
between Life Satisfaction and GDP per capita.
• Cross-section and panel analysis based on individual data:
strong relation between individual income and well-being.
• Aggregate time-series: no correlation between Life
Satisfaction and GDP per capita.
• The Easterlin Paradox is related to the time-dimension.
17
B) Maybe GDP growth increases happiness only
in poor countries?
• The happiness returns to income growth are decreasing
• But is there an upper bound to happiness?
“Once a country has over $15,000 per head, its level
of happiness appears to be independent of its income
per head” (Layard, 2005).
“Survival Societies” versus “Modern Societies”
(Inglehart et al. 2009)
 Satiation points (Adam Smith)?
But the happiness-log GDP per capita
gradient does not tend to zero
• New lessons from Gallup world poll 2006
Deaton (2008)
Wolfers and Stevenson (2008, 2010)
• “In logarithmic terms, there is no evidence that
cross-country effects of greater income fade out or
vanish as countries increase their income” (Deaton,
p 3).
19
The happiness-log GDP per capita gradient
does not tend to zero
• Stevenson and Wolfers (2008, p11-12): the well-being-GDP
gradient is about twice as steep for poor countries as for rich
countries. That is […] a rise in income of $100 is associated with
a rise in well-being for poor countries that is about twice as large
as for rich countries”.
• However, the marginal utility of GDP growth is still positive in
developed countries. “the Gallup results suggest that a 1 percent
rise in GDP per capita would have about three times as large an
effect on measured well-being in rich as in poor nations.
• Of course, a 1 percent rise in U.S. GDP per capita is about ten
times as large as a 1 percent rise in Jamaican GDP per capita”.
20
Stevenson and Wolfers. “Subjective Well-Being and Income: Is
There Any Evidence of Satiation?” (2013) AER, P&P
21
22
The happiness-log GDP per capita gradient
does not tend to zero
• Deaton (2008): “the relationship between log per capita
income and life satisfaction is close to linear. The coefficient
is 0.838, with a small standard error. A quadratic term in the
log of income has a positive coefficient: confirming that the
slope is higher in the richer countries! […]
•
If there is any evidence for a deviation, it is small and is
probably in the direction of the slope being higher in the highincome countries”.
23
Deaton, 2008
Life satisfaction = 0.838 log GDP per capita
24
Easterlin and Angelescu, 2007: “Rather than
diminishing marginal utility of income, there is a
zero marginal utility of income”
Easterlin and Angelescu, 2007, p 24:
“The usual constancy of subjective well-being in the face of rising
GDP per capita has typically been reconciled with the crosssectional evidence on the grounds that the time series
observations for developed nations correspond to the upper
income range of the cross-sectional studies, where happiness
changes little or not al all as real income rises.”
But “the income change over time within the income range used in
the point-of-time studies do not generate the change in happiness
implied by the cross-sectional pattern”.
25
Easterlin and Angelescu (2007, p 24):
“In 1972, the cohort of 1941-1950 had a mean per
capita income of about 12000$ (expressed in 1994
constant prices).
By the year 2000, the cohort’s average income had
more than doubled, rising to almost 27000$.
According to the cross-sectional relation, this increase
should have raised the cohort’s mean happiness from
2.17 to 2.27. `
In reality, the actual happiness of the cohort did not
change”.
26
Misleading cross-sections.
Actual versus predicted happiness in Japan, 1958-1987.
Easterlin, 2005, Easterlin and Sawangfa 2005
C) The Evidence about the Long-Run IncomeHappiness Nexus is Mixed and Still Controversial
• Income growth does not increase happiness over time:
Easterlin (2005a), Easterlin and Sawangfa (2005,
2009), Easterlin and Angelescu (2007), Easterlin
(2009), Layard; Brockmann, Delhey, Welzel, Yuan
(2009)
• Income growth does increase happiness over time:
Helliwell (2002), Stevenson and Wolfers (2008,
2010), Deaton (2008), Blanchflower (2008)
• Income growth does increase happiness over time but
not always and weakly:
Using GSS, Euro-barometer survey series, WVS,
Gallup World Poll, BHPS
o Hagerty and Veenhoven (2000, 2003, 2006): positive and
statistically significant coefficient, but not in all countries.
o Inglehart, Peterson and Welzel (2008); Kenny (2005):
positive and statistically significant coefficient, but not in
all countries.
o Layard, Mayraz and Nickell (2010): positive coefficient
but not always statistically significant.
o Oswald (1997): positive coefficient but not always
statistically significant.
o Di Tella and MacCulloch (2008): positive coefficient but
weak statistical significance.
Is the dynamic correlation “small enough to
ignore”? (Hagerty and Veenhoven, 2000)
• Still controversial
• Statistical dispute: even when the data do not
allow to establish a relation, this does not
mean that it can be rejected.
• Less statistical power in the long run series of
well-being than in the cross-section, because
of the smaller variance
=> The evidence is mixed.
The Evidence about the Long-Run IncomeHappiness Nexus is Mixed
Inglehart, Foa,
Peterson,
Welzel (2008).
The debate continues
• Betsey Stevenson et Justin Wolfers (Wharton,
University of Pennsylvania)
“Economic growth and subjective well-being:
reassessing the Easterlin Paradox”, Brookings Papers
on Economic Activity, 2008 (see also Sacks, Stevenson
and Wolfers, 2010, IZA DP n°5230).
32
Stevenson and Wolfers (2008)
• Re-assess the paradox analyzing multiple rich datasets spanning
many decades. Using recent data on a broader array of countries.
• Establish a clear positive link between average levels of
subjective well-being and GDP per capita across countries,
• find no evidence of a satiation point beyond which wealthier
countries have no further increases in subjective well-being.
• show that the estimated relationship is consistent across many
datasets and is similar to the relationship between subject wellbeing and income observed within countries.
• Finally, find that economic growth is associated with rising
happiness.
• Together these findings indicate a clear role for absolute income.
33
Stevenson and Wolfers (2008) - cntd
34
Stevenson and Wolfers (2008) – cntd 2
35
Sacks, Stevenson and Wolfers (2010)
At least 10 years between 2 dates
36
W&S revisit Easterlin:
Japan, 1958-1986, Average satisfaction score
37
Taking into account the discontinuity in
Japanese series used by Easterlin
38
D) GDP Fluctuations are strongly
correlated with average happiness
• Recession makes people unhappy
• Macroeconomic movements exert strong effects on
the happiness of nations:
unemployment,
inflation,
the volatility of output
o Di Tella, MacCulloch and Oswald (2003) , Wolfers
(2003).
40
Source: Stevenson and Wolfers (2008)
Life Satisfaction Follows the Business Cycle
during the Russian Transition
Guriev and Zhuravskaya (2008)
Happiness and GDP Growth during
Transition
Easterlin (2009)
Angus Deaton (PNAS, 2011)
The Financial Crisis and the Well-Being of Americans
43
Graham, Chattopadhyay and Picon (2010)
44
A misleading short-term association?
• “One should avoid confusing a short-term positive
happiness-income association, due to fluctuations in
macroeconomic conditions, with the long-term
relationship”.
• “this disparity between the short and long-term
association is due to the psychological phenomenon of
loss aversion”. (Easterlin, 2009).
A Short-Term Positive Happiness-Income
Association?
• Is Transition a short-term phenomenon? Or a regime
change?
Transition shares the essential features of development
o take-off period
o profound qualitative and institutional changes
o restructuring
Development should increase happiness… in the “short
run”
It is only with the passage of time, that one will be able
to observe whether the increase in subjective well-being
continues with GDP growth, stagnates at a certain point,
or goes down to the initial (1990) level.
E) A measurement problem?
• Artefact linked to a limited scale?
 better question « do you find that your income has
increased as compared to 10 years ago? »
or extendable satisfaction scale.
47
• Deaton (2008, p12-13): “The « best possible » life for you » is a
shifting standard that will move upwards with rising living
standards, so that we might expect the Danes to continue to
report 8 out of 10 as national income rises, provided they stay in
the same position in the global income rankings.
• Indeed it is hard to see how they could do differently faced with a
scale that has a maximum of 10.
• According to this view, average national life satisfaction will be a
useful measure in the cross-section, but not over time.”
48
Bounded Happiness Scales
• Satisfaction judgements are expressed on an ordinal
bounded-scale.
• They express relative judgements, i.e. the relation
between individuals’ attainments and the existing of
possibilities (represented by the scale).
• Framing effect
Hedonic treadmill
"Real" adaptation
Satisfaction treadmill
(nominal adaptation)
Measures of Quality of Life in Asian Countries
during Positive Episodes of GDP Growth
Cardinal measures 
Ordinal measures 
World Bank data, 1980-2007
Measures of Quality of Life in Rich Western
Countries during Positive Episodes of GDP
Growth
Cardinal measures 
Ordinal measures 
World Bank data
F) SPECIFICATIONS OF THE ESTIMATES
• Co-movements between growth and quality of life
indicators
Life expectancy
Child mortality
School enrollment
…
• These channels from GDP growth to happiness
should not be neutralized (controlled for) in
statistical estimates.
NEGATIVE OMITTED VARIABLES
• Pollution, income inequality, work stress, extended working hours,
unemployment, environmental degradation, fat intake (obesity and
blood pressure), sub-urbanization…
• The influence of these “omitted variables” could hide the positive
influence of GDP growth on subjective well-being in econometric
estimates (Di Tella and MacCulloch, 2008).
• Many of the negative externalities of growth increase in the initial
stages of development and decrease in later stages.
WHAT IS LEFT?
• The Easterlin paradox is about the TREND in life satisfaction over
time, on the long run
 not about cross-sectional relations
 not about happiness over the business cycle
• The question is about the size of the correlation coefficient
 Is is significantly different from zero? Too small to matter?
• Another formulation of the question would be: “what difference
does it make in terms of well-being, whether GDP grows at a rate
of 3% versus 1%?
What about the distribution of happiness?
Clark, Flèche and Senik (2012)
the Great Happiness Moderation
• As GDP rises, average happiness becomes
more equally distributed
• Not a negligible gain for risk-averse agents
• Attention, remettre les bons formats
56
What do we do?
• We uncover a declining spread in happiness over time during
periods of positive income growth
 In cross-sections of countries
 In panels of countries
 In time series within-country
• An “augmented Easterlin paradox” (mean-preserving declining
spread in happiness)
A Mean-Preserving Decline in Happiness
Spread
 This applies to life satisfaction and to all the domains of satisfaction
available in the data.
 The fall in happiness variance comes about because there are fewer
responses at both ends of the subjective well-being scale.
 We find no such movement in the countries that have had zero growth,
and the opposite effect in countries that have undergone recessions.
 This does not reflect demographics
show that, as expected, the percentage of people declaring either very low or very high happiness
The extremes of the happiness scale shrink
shrinks, whereas the weight of intermediate happiness categories rises.
&
4.A&Great&Britain&(BHPS)&
40
&
&
&&&&&&&& &&4.B&Germany&(GSOEP)&
35
1996
2008
35
&
1984
2009
30
30
&
25
25
20
20
15
15
10
10
5
5
0
0
1
2
3
4
&
4.C&Australia&(HILDA)&
&
5
6
&
7
0
&
&&&&&&
1
2
3
4
5
6
7
8
9
10
&&&4.D&United&States&(GSS)&
8
Data
(Respondents 18-65 years old)
• World Values Survey: 5 waves, 1981-2008. Life satisfaction: 1-
10
Periods of positive income growth, at least 5 years  60
countries
• Great-Britain: BHPS: 1996-2008.
Life satisfaction: 1-7
• Germany:
Life satisfaction: 0-10
• Australia:
GSOEP: 1984-2009.
HILDA: 2001-2009.
Life satisfaction: 0-10
The American General Social Survey (19722009)
• The only long run survey containing a happiness or life
satisfaction question in the United-States.
• But only 3 modalities: very happy, pretty happy, not too
happy
• Obviously not fit to the analysis of the variance (although
W&S, Dutta and Foster)
• We consider the results with greater caution.
Household Income
• Ideally, use the net disposable income after tax and transfers.
• GSOEP and HILDA: measure of the annual disposable net
combined income after taxes and public transfers
(Government pensions and benefits).
• BHPS: combination of labor income, non-labor income and
pensions for all household members, in the previous year,
before taxes.
• GSS: “total family income”, i.e. all types of income from all
sources, for all members of the household, before taxes, in
the previous year.
Standard deviation / mean happiness
• Self-declared happiness is a choice on a proposed scale
Equality  all respondents choose the same rating
• Standard deviation of self-declared happiness / mean
happiness for each country*year
No scale dependence
Cardinalization
Index of Ordinal Variance
Wolfers and Stevenson (2008), Dutta and Foster (2011)
device sophisticated measures of happiness inequality
o which lead to exactly same results as ours.
Happiness inequality and GDP per capita crosscountries,
WVS, last available year (2000s)
A doubling of
GDP per
capita is
associated
with a 10%
reduction in
happiness
inequality.
Same result with
sd(happy)/ mean(happy)
Happiness Inequality Over Time in Growing
Countries, Western Countries (WVS)
Happiness Inequality Over Time with Some
Periods of Negative or Zero Growth (WVS)
Happiness Inequality Over Time during Periods
of Decreasing GDP (WVS)
Happiness Inequality Over Time during Periods
of Decreasing GDP (Eurobarometer)
Happiness Inequality Over Time during Periods
of Negative or Zero Growth (WVS)
Happiness Inequality Over Time during Periods
of Decreasing GDP (Eurobarometer)
Trends in income growth, average happiness
and happiness inequality
Great Britain (BHPS)
Same result with sd(happy)instead of sd/mean; with or
without controls
Trends in income growth, average happiness
and happiness inequality
West Germany (GSOEP)
Same result with sd(happy)instead of sd/mean; with or
without controls
Trends in income growth, average happiness
and happiness inequality
Australia (HILDA)
Same result with sd(happy)instead of sd/mean; with or
without controls
Trends in income growth, average happiness
and happiness inequality
USA (GSS)
Same result with sd(happy)instead of sd/mean; with or
without controls
Trends in income growth, average happiness
and happiness inequality
Other OECD countries
(10 years, positive growth), flat happiness trend)
Trends in income growth, average happiness
and happiness inequality
Other OECD countries
(10 years, positive growth), flat happiness trend)
The Extreme Levels of Happiness are
Vanishing
The Extreme Levels of Happiness are
Vanishing
Trends in average Satisfaction and Satisfaction
Inequality by Domain, Great Britain
Happiness Inequality and Income Inequality
• Starting in the 1980s, in Australia, Great-Britain, Germany,
United-States, general rise in income inequality
Dustmann, Ludsteck and Schönberg (2008)
Atkinson, Piketty and Saez (2011)
• General fall in the spread of happiness
 although in Germany and the US, this trend breaks in the
1990s.
82
Happiness inequality and social expenditures
Two-Periods in Germany and USA
• In the late 1990’s happiness inequality rises again in Germany
and in the USA, probably due to the rise in income inequality
Interpretation
• Explain: (1) the rise in average income per capita over
time, (2) the stability of average happiness over time, (3)
the fall in happiness inequality over time.
• Not the evolution of income inequality
• Not the effect of demographic change
Two Lines of Interpretation
• Actual Homogeneization of Happiness
• Rescaling
In both cases: qualification of the Easterlin paradox
1) Social harmonization, public goods and
modern growth
• Externalities of economic growth and modernization
Welfare system
Improvement in education, health, life expectancy, child
mortality
Infrastructure
Political rights, private liberties, gender equality,
capabilities
2) Income as Buffer Stock
• Higher revenue allows buffering income shocks and other
various shocks of life, such as job loss, divorce etc…
Happiness dispersion is lower within richer groups
Rich deciles of the population experience less volatility
of happiness over time
The spread of Happiness is lower within higher
deciles of the income distribution
Individuals are
assigned to the
average decile
they belong to
over the whole
period
The spread of happiness over time is lower in
higher deciles of the income distribution
!
Great&Britain&&
&
&
&
&
&
!!!!!!!!!!
Australia&(HILDA)&&
&
&
&
West&Germany&(GSOEP)&
!
&
Individuals are assigned to
the average decile they
belong to over the whole
period.
!
!
3) Rescaling
• Satisfaction treadmill rather than hedonic treadmill
Relationship between latent happiness and self-declared
happiness
o “The ‘best possible life for you’ is a shifting standard that will
move upwards with rising living standards” (Deaton, 2008)
Unequal growth: people “rescale” more at the top of the
ladder than at the bottom, because their world of
opportunities expands more than that of less fortunate
people.
This would create a convergence movement whereby the
Some Usual Interpetations or Discussions of
the Easterlin Paradox that Do Not Hold
1. Happiness as a log function of (absolute)
income and nothing else? (W&S)
• Possible if income distribution concentrates around the
mean or median
Not really what’s happening
• Test:
Estimate happiness function in first year:
o Happiness= a0 + a1 age + a2 age2 + a3 log income + a4
gender + e
Predict distribution of happiness in end year
Actual versus predicted distribution of
happiness
BHPS
2. Social comparisons
• A priori, in the presence of rising income inequality, income
comparisons should lead to an increase in the standard
deviation in happiness, not to a fall.
• Comparisons are mostly upward
• Top income are diverging
• Not easy to imagine how comparisons could lead to a
reduction in the happiness spread.
3. Adaptation
• Simple negative influence of lagged income?
• Bliss point, satiation point?
• Would not lead to the fall of the happiness spread
• More sophisticated notions of Adaptation
Maslow
Rescaling
Adaptation à la Maslow (1943, 1954)
• Higher needs may be much more difficult to fulfill than basic
needs.
• Survival versus life; Survival societies versus modern
societies
•
“Economic development increases people’s sense of existential security, leading them
to shift their emphasis from survival values towards self-expression values and free
choice…” (Inglehart 2010).
• Development  the share of the population that feels totally
deprived (the bottom of the scale) and totally satisfied (the top
of the scale) both shrink.
Conclusions
• Actual concentration of happiness:
Positive message to developing countries: promise of
greater social homogeneity
More optimistic interpretation of the Easterlin paradox
• Rescaling:
Questions the pessimistic message of Easterlin
Appendix
Index of Ordinal Variance vs Standard deviation
Index of Ordinal Variance vs Standard deviation
Income inequality and happiness inequality in
Germany
Sd(income) and sd(happiness)
Average happiness by quintile
Income level by quintile
Sd(happiness) by quintile
Happiness inequality between and within
income, groups, Germany: 2 periods
Average happiness by quintile
Sd(happiness) by quintile
Income inequality and happiness inequality in
Australia
Sd(income) and sd(happiness)
Average happiness by quintile
Income level by quintile
Sd(happiness) by quintile
Income inequality and happiness inequality in
the U.S.
Sd(income) and sd(happiness)
Income level by quintile
Happiness inequality between and within
income, groups, USA: 2 periods
Average happiness by quintile
Happiness equalization is not due to
demographic change
• RIF regressions of the variance in happiness (based on the
WVS) show that GDP per capita and income inequality affect
happiness inequality
beyond the impact of demographic change
both in cross-section estimates (controlling for year fixedeffects) and over time (controlling for country fixed-effects).
• Common trend: happiness inequality declines in all countries
within age, education, gender, marital status and employment
status categories,
•
RIF: Recentered Influence Function regressions (Firpo , Fortin and Lemieux, 2009)