Income Comparisons in China Andrew E. Clark (Paris School of - - PowerPoint PPT Presentation

Income Comparisons in China

Andrew E. Clark (Paris School of Economics and IZA) Cl di S ik (P i S h l f E i d P i IV) Claudia Senik (Paris School of Economics and Paris IV) Happiness and Economic Growth: Lessons from Developing Countries March 9th 2012 March 9 2012

One of the reasons that we are (or should be) One of the reasons that we are (or should be) interested in China is that it is a very large country so that it “matters” in a global sense country, so that it “matters” in a global sense. Another is it’s fantastic recent level of economic growth growth.

Chinese growth rates are (arguably) unprecedented

Chinese Real GDP Growth Rate 14 16 8 10 12 2 4 6 8 2 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 1 1 1 1 1 1 1 1 2 2 2 2 2

Leading to a rise in real GDP of an

• rder of magnitude

What might be the Subjective Well-Being consequences g j g q

• f such growth?

While the Chinese case is impressive, there is something

• f a parallel with Japan
• f a parallel with Japan.

Japan was a poor country in the 1950s/early 1960s, but then experienced unprecedented growth. Easterlin, R. A. (2005). “Diminishing Marginal Utility Easterlin, R. A. (2005). Diminishing Marginal Utility

• f Income? Caveat Emptor”. Social Indicators
• Research. pp. 243-255.

esea ch. pp. 3 55.

Fact 1. Richer countries are happier countries.

The blue lines show the estimated relationship between income and between income and happiness Japan Japan was in the middle of the income distribution in the early 1960s and had Japan was in the middle of the income distribution in the early 1960s, and had a middling level of happiness

So what happened as Japan became richer? So what happened as Japan became richer? Look at annual indices (1962=100) of life satisfaction and real GNP per capita for Japan satisfaction and real GNP per capita for Japan, 1958-1987 B 1962 d 1987 J i d Between 1962 and 1987 Japan experienced unprecedented economic growth, with GNP i (i l ) i i 3 5 f ld per capita (in real terms) rising 3.5-fold: growing from 22 to 77 percent of the United States level in 1962 We might then imagine that Japan would follow g g p the blue regression tramlines above: as Japan became richer, it would become happier. , pp

In fact, the trend in happiness was flat, despite Japan’s remarkable economic growth

What “should” have happened What did happen

Are the Japanese strange? N t il th thi b b d i 30 f A i d t Not necessarily: the same thing can be observed in 30 years of American data

2.5 3

30000 40000

2000 US$) 1.5 2

Happiness 20000 30000

Per Capita (2 0.5 1

Average H 10000

Real Income P

Happiness Real Income Per Capita 1973 1977 1981 1985 1989 1993 1998 2003 Year

R

Year

This is the Easterlin Paradox. Richer individuals are happier, but as countries become richer over time they do not become happier. The challenge to social science is to try to explain why.

One candidate explanation of the Easterlin paradox is via p p comparisons of my own income or consumption to:

• The consumption of “others”

The consumption of others

• My own consumption in the past

This is a great research topic. But, unfortunately, not always so easy to carry out convincing tests convincing tests.

We would like to estimate a “well-being” function of the form: W = W(Y, Y*, ....) The key variable Y* here is “comparison income”: the income to which we compare/income of the reference group reference group.

Imagine that we can measure well-being “W” in Imagine that we can measure well being, W , in surveys via job satisfaction, life satisfaction, happiness mental stress/depression etc happiness, mental stress/depression etc. These measures have by now been reasonably- widely validated by widely validated by physiological/neurological studies, third- t t d ( t i t tl ) f t party raters, and (most importantly) future behaviours such as divorce, unemployment duration, quitting one’s job, and morbidity and mortality. y

But there remains the problem of measuring Y*: But there remains the problem of measuring Y : to whom do we compare?

• Peer group/people like me
• Others in the same household
• Spouse/partner

Spouse/pa e

• Myself in the past
• Friends
• Friends
• Neighbours
• Work colleagues
• “Expectations”

p

We very rarely have direct information about to whom people compare.

• And if we do have that information, we do not know

how much individuals in the reference group earn. A d h d h i f ti it l

• And when we do have information, it rarely covers

everyone in the reference group. The data we use here help us to overcome these drawbacks.

“IFPRI Public Policy and Rural Poverty” Program Survey Jointly conducted by IFPRI the China Academy of Jointly conducted by IFPRI, the China Academy of Agricultural Sciences, and Guizhou University.

Guizhou Province

About.com does not seem to have been sponsored by the local tourist office… L ti G i h (贵州) P i i i Location: Guizhou (贵州) Province is in China's south. It is one of China's less- developed provinces. Capital city: Guiyang (贵阳) is the capital and largest city in Guizhou and largest city in Guizhou. Attractions: there isn't a big list of famous attractions in Guizhou.

The Guizhou data is very interesting for a y g number of reasons.

• It is panel with waves in 2004 2007 and

It is panel, with waves in 2004, 2007 and 2010.

• At each wave individuals are interviewed
• At each wave, individuals are interviewed

within households. S

• Serious attempts are made to interview all

households within 26 different villages in Guizhou.

• In 2010, the head of household provided

, p information on various measures of satisfaction, including satisfaction with satisfaction, including satisfaction with income.

In 2010, average household income was , g 16157.75 Renminbi (about 2500 Dollars, using an exchange rate of 6 3) using an exchange rate of 6.3). The distribution of income is very unequal. The Gini coefficient in 2010 in the Guizhou The Gini coefficient in 2010 in the Guizhou data was 0.547. Th hi h t OECD fi d 0 5 The highest OECD figures are around 0.5 (Chile and Mexico) The D9/D1 figure is 14 3; D9/D5 is 3 6 The D9/D1 figure is 14.3; D9/D5 is 3.6.

In this context of fast income growth, but also considerable inequality, we would like to see whether there is any evidence of income i i Chi comparisons in China. If there are, then the effect of income growth on ell being ill be dampened well-being will be dampened. The natural reference group in this dataset is the village. There are 25 villages in the 2010 wave. The smallest village has 36 households interviewed; the largest has 256 households interviewed.

• No. of

HHs % Cum. % 36 1 1 1 1

There are 25 villages in the

36 1.1 1.1 40 1.3 2.4 45 1.4 3.8 48 1.5 5.4

There are 25 villages in the 2010 wave.

56 1.8 7.1 61 1.9 9.1 63 2.0 11.1 72 2.3 13.4

The smallest village has 36 households interviewed; the largest has 256

86 2.7 16.1 89 2.8 18.9 102 3.2 22.2 112 3 6 25 7

the largest has 256 households interviewed.

112 3.6 25.7 123 3.9 29.6 125 4.0 33.6 129 4.1 37.7 147 4.7 42.3 161 5.1 47.4 167 5.3 52.7 168 5 3 58 1 168 5.3 58.1 184 5.8 63.9 213 6.8 70.7 218 6.9 77.6 221 7.0 84.6 230 7.3 91.9 256 8.1 100.0

Our measure of well-being here is ti f ti ith i satisfaction with income.

“Are you satisfied with your current income?” Freq. Percent Cum. Freq. Percent Cum. Not satisfied 1,295 42.13 42.13 Not too satisfied 896 29 15 71 28 Not too satisfied 896 29.15 71.28 Average 481 15.65 86.92 Rather satisfied 374 12.17 99.09 Very satisfied 28 0 91 100 00 Very satisfied 28 0.91 100.00 Total 3,074 100.00

Our Guizhou villagers are not very satisfied. Littl id f S ’ H Sl h Little evidence of Sen’s Happy Slaves here

The completeness of information regarding p g g the reference group means that we can not

• nly appeal to standard (estimated)
• nly appeal to standard (estimated)

measures of centrality such as the mean, b t l id th di but we can also consider the median. We also know the household’s rank within the village: is it the top earner? In the top the village: is it the top earner? In the top 10%? The top 25%?

Empirical questions: Empirical questions:

• Does own household income predict

income satisfaction? income satisfaction?

• What aspects of the distribution of

h h ld i i th ill tt household income in the village matter, conditional on the household’s own income?

• Here are some of our preliminary results

Here are some of our preliminary results.

• None of these are contractually-binding

Basic Specification (these are all linear regressions, with clustering at the village level):

Income Satisfaction Log Household Income 0.071 (0.018)*** Male

• 0.006

(0.043) Age

• 0.002

(0.007) Age-Squared/1000 0.050 (0 065) (0.065) Married 0.004 (0.059) Did not graduate from primary school 0.103 (0 062)* (0.062)* Primary School Graduate 0.206 (0.080)** Did not graduate from junior high 0.138 (0 077)* (0.077)* Junior high graduate or above 0.125 (0.072)*

• no. of children
• 0.040

(0 016)** (0.016)

Income satisfaction rises with own income (Phew!), sort of rises with education, and falls with number of children. Log income is preferred to the level of income by the regression

Including Village Income:

Income Satisfaction Log Household Income 0.057 g (0.040) Log Median Village Household Income 0.233 (0.120)*

Income satisfaction RISES with village median income, conditional on own income. The Chinese seem to be similar to the Danish (if h I ) you see what I mean)

Is it the level of Village Income or rank that matters?

Log Household Income

• 0.033

Log Household Income 0.033 (0.080) Log Median Village Household Income 0 300 Log Median Village Household Income 0.300 (0.136)** Vill N li d I R k 0 368 Village Normalised Income Rank 0.368 (0.365)

Rank within the village attracts a positive coefficient, but in no way significant but in no way significant.

Are income comparisons therefore a rich country phenomenon? This is a critical question to answer. If they are then the sauce for the Goose (OECD countries) is not the sauce for the Gander (everyone else) In the results so far Chinese income growth In the results so far, Chinese income growth seems to be a good thing for all concerned: there are no externalities/spillovers from others’ higher are no externalities/spillovers from others higher income here

But…What if rank is not cardinal?

Going from 2nd decile to 1st decile more important than going g p g g from 7th decile to 6th decile? (like Wimbledon)

Log Household Income 0.033 0.051

• 0.040
• 0.044

(0.042) (0.040) (0.061) (0.060) Log Median Village Household Income 0.250 0.236 0.307 0.310 (0.121)** (0.119)* (0.136)** (0.135)** Top 10% Village Normalised Income Rank 0.155 0.238 0.220 (0.082)* (0.101)** (0.105)** Top Income Rank in Village 0 372 0 342 Top Income Rank in Village 0.372 0.342 (0.269) (0.292) Bottom 25% Village Normalised Income Rank

• 0.216
• 0.222

(0.122)* (0.121)*

• Being in the top decile in the village attracts a positive

significant coefficient; being in the bottom 25% is associated with

(0.122) (0.121)

significant coefficient; being in the bottom 25% is associated with lower satisfaction (conditional on own income)

• Being top dog is good (only 25 observations, so estimate is

Being top dog is good (only 25 observations, so estimate is insignificant here)

Heterogeneity: Immigrants are unhappy about being low rank Immigrants are unhappy about being low rank

Log Household Income

• 0.040
• 0.018

(0.061) (0.062) Log Median Village Household Income 0.307 0.271 (0.136)** (0.137)* Top 10% Village Normalised Income Rank 0.238 0.242 (0.101)** (0.105)** Bottom 25% Village Normalised Income Rank

• 0.216
• 0.159

(0.122)* (0.129) ( ) ( ) Immigrant 0.004

• 5.429

(0.093) (2.836)* Immigrant*Log Household Income

• 0.501

Immigrant Log Household Income 0.501 (0.258)* Immigrant*Log Median Village Household Income 1.135 (0 214)*** (0.214) Immigrant*Top 10% Village Normalised Income Rank

• 0.013

(0.541) Immigrant*Bottom 25% Village Normalised Income Rank 1 233 Immigrant*Bottom 25% Village Normalised Income Rank

• 1.233

(0.304)***

Heterogeneity: Men are particularly rank-sensitive Men are particularly rank sensitive

Log Household Income 0.011 (0 066) (0.066) Log Median Village Household Income 0.326 (0.128)** Top 10% Village Normalised Income Rank 0.084 (0.138) Bottom 25% Village Normalised Income Rank 0 166 Bottom 25% Village Normalised Income Rank

• 0.166

(0.124) Male*Log Household Income

• 0.097

(0.037)** Male*Log Median Village Household Income

• 0.037

(0 064) (0.064) Male*Top 10% Village Normalised Income Rank 0.291 (0.133)** ( ) Male*Bottom 25% Village Normalised Income Rank

• 0.089

(0.078)

Heterogeneity: Positive effect of median income found only for the lowest- educated (Illiterate/ did not graduate from primary school) educated (Illiterate/ did not graduate from primary school)

Log Household Income 0.011 (0.066) Log Median Village Household Income 0.326 (0.128)** (0.128) Top 10% Village Normalised Income Rank 0.084 (0.138) ill li d k Bottom 25% Village Normalised Income Rank

• 0.166

(0.124) Educated*Log Household Income 0.040 ducated

• g
• use o d

co e 0.0 0 (0.072) Educated*Log Median Village Household Income

• 0.422

(0 185)** (0.185)** Educated*Top 10% Village Normalised Income Rank

• 0.041

(0.187) ( ) Educated*Bottom 25% Village Normalised Income Rank 0.257 (0.160)

Panel Data: Past Income is POSITIVELY associated with income satisfaction satisfaction

Log Household Income

• 0.013

(0.076) Log Median Village Household Income 0.108 (0 137) (0.137) Top 10% Village Normalised Income Rank 0.191 (0 126) (0.126) Bottom 25% Village Normalised Income Rank

• 0.089

(0 144) (0.144) Lagged Log Household Income 0.217 (0 045)*** (0.045)

So no habituation effects here So no habituation effects here

Panel Data: Same thing for past Income rank Same thing for past Income rank

Log Household Income 0.004 (0 079) (0.079) Log Median Village Household Income 0.208 (0.167) ( ) Top 10% Village Normalised Income Rank 0.189 (0.126) Bottom 25% Village Normalised Income Rank

• 0.064

(0.144) Lagged Top 10% Village Normalised Income Rank 0 130 Lagged Top 10% Village Normalised Income Rank 0.130 (0.171) Lagged Bottom 25% Village Normalised Income Rank

• 0.272

(0.094)***

Panel Data: But do you get used to your Income rank? But do you get used to your Income rank?

Log Household Income

• 0.002

(0.080) Log Median Village Household Income 0.204 (0.167) Top 10% Village Normalised Income Rank 0.310 p g (0.165)* Bottom 25% Village Normalised Income Rank

• 0.204

(0 168) (0.168) Lagged Top 10% Village Normalised Income Rank 0.215 (0.170) Lagged Bottom 25% Village Normalised Income Rank 0 411 Lagged Bottom 25% Village Normalised Income Rank

• 0.411

(0.116)*** Top 10% Village Normalised Income Rank Both Now and Past

• 0.457

(0 256)* (0.256)* Bottom 25% Village Normalised Income Rank Both Now and Past 0.448 (0.188)**

Habituation not to income, but to income rank

We also have information on subjective income rank in 2010

Within your village what socioeconomic level does your family’s income place you? y y Freq. Percent Much higher than average 9 0.29 Higher than average 261 8.38 Average 1,335 42.88 Below average 1,088 34.95 Much below average 420 13.49 Total 3,113 100.00

Little evidence that growth has led these villagers to consider themselves as high income rank consider themselves as high income rank

Subjective income rank predicts income satisfaction L H h ld I 0 035 Log Household Income 0.035 (0.037) inclevel2010==Higher than average 0.864 (0.222)*** inclevel2010==Average 0.530 (0.137)*** ( ) inclevel2010==Below average 0.082 (0.145) O itt d i k t i “ h b l ” (0.145) Omitted income rank category is “much below average”

Subjective income rank predicts income satisfaction, E h t l f bj ti i k Even when we control for objective income rank

Log Household Income

• 0.051

(0 060) (0.060) inclevel2010==Higher than average 0.862 (0.218)*** inclevel2010==Average 0.517 (0.135)*** inclevel2010==Below average 0.087 g (0.146) Log Median Village Household Income 0.260 (0 136)* (0.136)* Top 10% Village Normalised Income Rank 0.150 (0.108)

S bj ti i k i th f h th

Bottom 25% Village Normalised Income Rank

• 0.187

(0.115)

Subjective income rank is therefore much more than where you are in the village income distribution…

Even though subjective and objective income rank in th ill i d d l t d the village are indeed correlated

“Within your village what socioeconomic level does your Within your village what socioeconomic level does your family’s income place you” Mean(normalised income rank) Higher than average .565 Average .551 Below average .491 Much below average .391

One research task would therefore be to understand why higher relative income levels do not seem to do a why higher relative income levels do not seem to do a good job of making people feel high rank.

Conclusions

• Much work has been carried out on income

comparisons (either to others or to oneself in the past) as a potential brake on the ability of higher income to produce higher well-being.

• We know much less about such comparisons

in non-OECD countries in non OECD countries

• The Guizhou data is very useful here,

containing information on all households within 26 villages in rural China, including past household income.

• The initial results suggest that income

ti f ti i ith b th th ’ i d satisfaction rises with both others’ income and with own past income.

• So Chinese income growth involves no well-

being externalities? being externalities?

• Not so fast: there are comparisons, but in

t f i k Chi i terms of income rank. Chinese comparisons seem to be ordinal, rather than cardinal

• Being low rank in the village (CONDITIONAL
• n own income) is particularly harmful
• n own income) is particularly harmful
• But there is also habituation to both low and

hi h k high ranks