Income, aspirations and the Hedonic Treadmill in a poor society

I

Ja

b

a

ARRAA

JDOO

KAACEHHS

1

arhoi

hhn8

0d

Journal of Economic Behavior & Organization 82 (2012) 67– 81

Contents lists available at SciVerse ScienceDirect

Journal of Economic Behavior & Organization

j ourna l ho me pag e: www.elsev ier .com/ locate / j ebo

ncome, aspirations and the Hedonic Treadmill in a poor society

ohn Knighta,∗, Ramani Gunatilakab

Department of Economics, University of Oxford, Manor Road Building, Oxford OX1 3UQ, United KingdomDepartment of Econometrics and Business Statistics, Monash University, Melbourne, Australia

r t i c l e i n f o

rticle history:eceived 15 January 2010eceived in revised form 25 August 2011ccepted 1 December 2011vailable online 23 December 2011

EL classification:601215

eywords:daptationspirationshinaasterlin paradoxappinessedonic Treadmillubjective well-being

a b s t r a c t

A specially designed household survey for rural China is used to analyse the determinants ofaspirations for income, proxied by reported minimum income need, and the determinantsof subjective well-being, both satisfaction with life and satisfaction with income. It is foundthat aspiration income is a positive function of actual income and reference income, and thatsubjective well-being is raised by actual income but lowered by aspiration income. Thesefindings suggest the existence of a partial ‘Hedonic Treadmill’, and can help to explain whysubjective well-being in China appears not to have risen despite rapid economic growth.

© 2012 Elsevier B.V. All rights reserved.

. Introduction

This paper explores the relationships amongst income, aspirations, adaptation and happiness in a poor society. Howre people’s aspirations determined? For instance, do they adapt their aspirations to their income, whether absolute orelative to that of their reference group or reference time? Is happiness affected by people’s aspirations? For instance, doigh aspirations reduce happiness? We use a specially designed data set for rural China which contains information not onlyn income but also on aspirations for income (call it ‘aspiration income’), on several variables which can influence aspirationncome, and on two measures of subjective well-being.

Section 2 provides a brief literature review of this rather unexplored subject, and Section 3 develops the theoreticalypotheses to be tested. The data and variables are explained in Section 4. Section 5 outlines the strategy for testing ourypotheses. Section 6 examines empirically the determinants of aspiration income, and Section 7 the aspirational determi-
ants of two dimensions of subjective well-being: satisfaction with life and satisfaction with income. It is possible in Section
to suggest some novel and notable conclusions for the understanding of economic welfare.

∗ Corresponding author. Tel.: +44 01993 702866; fax: +44 01865 271094.E-mail addresses: [email protected] (J. Knight), [email protected] (R. Gunatilaka).

167-2681/$ – see front matter © 2012 Elsevier B.V. All rights reserved.oi:10.1016/j.jebo.2011.12.005

dx.doi.org/10.1016/j.jebo.2011.12.005

http://www.sciencedirect.com/science/journal/01672681

http://www.elsevier.com/locate/jebo

mailto:[email protected]

mailto:[email protected]

dx.doi.org/10.1016/j.jebo.2011.12.005

68 J. Knight, R. Gunatilaka / Journal of Economic Behavior & Organization 82 (2012) 67– 81

2. Literature review

There is a common view in the literature on subjective well-being that people adapt their aspirations in response tochanges in income. It was put forward by Easterlin (1974; see also 2001, 2005) as the explanation for the paradox that heobserved: in cross-section studies at the individual level, happiness is found to rise with income, whereas in time-seriesstudies at the national level, happiness does not rise as income increases.1 This explanation of the ‘Hedonic Treadmill’ interms of endogenous aspirations has been accepted by Frey and Stutzer (2001, 2002), and many others.

From a very different perspective, Sen (for instance, 1985, 1987) has argued that there can be reconciliation to poverty:the poor adapt their aspirations to their low income and poor quality of life. For instance:

A thoroughly deprived person, leading a very reduced life, might not appear to be badly off in terms of the mentalmetric of utility, if the hardship is accepted with non-grumbling resignation. In situations of longstanding deprivation,the victims do not go on weeping all the time, and very often make great efforts to take pleasure in small mercies andcut down personal desires to modest – ‘realistic’ – proportions. The person’s deprivation then, may not at all showup in the metrics of pleasure, desire fulfilment, etc., even though he or she may be quite unable to be adequatelynourished, decently clothed, minimally educated and so on (Sen, 1990, p. 45).

Sen and Easterlin have something in common. In both cases, the hypothesis is that people’s aspirations adapt to theirsituation, so reducing the sensitivity of subjective well-being to income. In Easterlin’s case people adjust to upward and inSen’s case to downward movement in income. The notion that aspirations adjust to what is perceived to be possible is wellestablished and much discussed from various perspectives, for instance by Elster (1982) from a philosophical, Frederick andLoewenstein (1999) from a psychological, Gurr (1970) from a political, and Van de Stadt et al. (1985) from an economicperspective.

The inurement of the poor to their poverty leads Sen (for instance, 1985, 1987) to reject the metric of utility as thecriterion for the evaluation of economic welfare, and to replace it with the concept of the ‘capabilities to function’, i.e.people’s capabilities to be and to do things of intrinsic worth. This separation only works, however, if the evaluation ofcapabilities involves an externally imposed value judgement rather than one which emerges from individual preferences.Kingdon and Knight (2006) have argued that, provided that the aggregation of expressed individual preferences is acceptedas the criterion for evaluation, the estimation of subjective well-being functions offers an encompassing framework whichyields weights for the valuation of income and various capabilities in a social welfare function; moreover, the estimationcan provide a measure of ‘subjective well-being poverty’ and of its determinants.

Another strand of literature implicitly approaches aspiration income by examining subjective poverty (for example, VanPraag and Kapteyn, 1973; Goedhart et al., 1977; Colasanto et al., 1984; Pradhan and Ravallion, 2000 with regard to developingcountries; Gustafsson et al., 2004 with regard to urban China). Minimum income functions generally show that the minimumincome that a household perceives itself to need rises with household income. The standard practice is to treat the level ofincome at which the function predicting minimum income equals actual income as the subjective poverty line: below thatlevel minimum income exceeds, and above it falls short of, actual income.

The reason for choosing the intersection rather than the average minimum income is ‘preference drift’, i.e. people’sperceived minimum income adjusts to their actual income. It is argued that the true minimum income is misperceived,and that only when minimum and actual income are equal is such distortion not present (Van Praag and Kapteyn, 1973;Goedhart et al., 1977).2 This view requires the value judgement that subjective well-being as a policy objective shouldbe measured after eliminating the adjustment of the minimum income that people feel they need to the income that theyexperience, i.e. after standardising for endogenous aspirations. Other value judgements are possible. For instance, Gustafssonet al. (2004) find a political explanation for the Chinese government’s setting of high real poverty lines in cities where theaverage minimum real income aspired to is high.

There is a substantial literature on the causes of a quite different dimension of satisfaction or dissatisfaction: politicaldiscontent and its expression. For instance, Gurr (1970) argued that political discontent is felt, and expressed, only whenpeople perceive that there is the prospect of change. Again, there is the notion that the downtrodden grow reconciled totheir political lot: aspirations adapt to the prevailing conditions but they can rise, and even outpace the actual improvement,if conditions begin to improve.

Hirschman (1973) depicted the effect of changing expectations on perceptions of well-being using the analogy of atunnel. Two lanes of cars are jammed in a tunnel. Eventually one lane begins to move forward. The satisfaction of those inthe stationary lane initially rises, anticipating that they will soon follow, but subsequently this feeling turns to one of despair
when nothing happens. Hirschman had in mind the effect of rising inequality in the process of economic development. Thus,it is possible that, as the income of people’s reference group rises, so their expectations of income, and their happiness,initially rise in anticipation but, unless their income follows, subsequently fall.
1 He showed that in both the United States and Japan, income per capita had risen greatly over the post-war decades but measures of happiness had notrisen at all over the same period (Easterlin, 1974).

2 However, a subjective poverty line defined as the intersection does not fully eliminate aspirations insofar as a rise in the income of the communityover time raises individual aspirations and thus the minimum income function, so increasing the intersection income.

rt1c

(tBwawa

gMcfpo

iHmaeiai

3

f

wp

T

THn

iitwi

w

wci

J. Knight, R. Gunatilaka / Journal of Economic Behavior & Organization 82 (2012) 67– 81 69

There is a literature on the relationships that aspirations bear to past income and to expected future income. Psychologicalesearch has found that people tend to judge the past and the future by reference to current aspirations: it is easier for peopleo recall or project their income than to recall or project their aspirations (Easterlin, 2001; Kahneman and Snell, 1992; Rabin,998). Clark (1999), using a British panel of employees, found that job satisfaction is strongly positively correlated with thehange in pay between waves of the panel but is unrelated to the current level of pay.

The responses to a rise and a fall in income may differ. Analysing the aggregate consumption function, Duesenberry1949) predicted asymmetry in the response of consumption to deviations of income from the highest level so far achieved,he reasoning being that people are willing and able to adapt their consumption aspirations upwards but not downwards.urchardt (2005) used a ten-year panel of the British Household Panel Survey to analyse the relationship between satisfactionith income and both the level of income and its change. She found that the effect of current income was significantly positive

nd the effect of a fall in income from its previous level was significantly negative. By contrast, the effect of a rise in incomeas not significantly different from that of static income (the omitted category), at least in the long term. These results imply

symmetry: people adapt their aspirations to rising income but less so to falling income.There is much economic literature providing evidence for many countries that relative income affects happiness, and

enerally inferring that endogenous aspirations are the explanation for this effect (well surveyed by Clark et al., 2008).oreover, the same results have been obtained, and the same inference drawn, for rural China (Knight et al., 2009). In some

ases, however, more evident proxies for aspirations have been used. For instance, in the case of the US Easterlin (2005)ound that there was a positive and commensurate relationship between the number of consumer durables desired andossessed by a cohort over time, and that the proportion of a cohort reporting to have ‘a lot of money’ stayed fairly constantver the life cycle despite rising income.

There has been little attempt so far to tackle our question directly by measuring aspiration income and estimatingts determinants, and then including both aspiration income and actual income as determinants of subjective well-being.owever, pioneering research by Stutzer (2004) answered the same question that is posed in this paper, and with a similarethodology, but did so for a rich society, Switzerland. It was found that aspiration income reduces subjective well-being,

nd that aspiration income is a positive function of lagged income. Thus, in response to an increase in income, the negativeffect of the consequent increase in aspiration income tends to offset the direct positive effect on subjective well-being of thencrease in actual income. Nevertheless, the cross-section relationship between current income and subjective well-being,lthough muted by endogenous aspiration income, is positive. Aspiration income is found to be a positive function of currentncome, previous income, and community income.

. Hypotheses

This section is influenced by Stutzer (2004). We hypothesise that happiness is a positive function of income and a negativeunction of aspiration income. A simplified linear happiness function might thus be expressed as:

H = aY − bA (1)

here H is a score of happiness, Y is income, A is aspiration income, and a and b are parameters (a, b > 0). In principle, it isossible that b = a. If aspiration income is a positive and linear function of income, we have an adaptation function:

A = cY (2)

he parameter c indicates the extent to which aspirations rise with income. Combining Eqs. (1) and (2), we have:

H = (a − bc)Y (3)

he term (a − bc) represents the extent to which happiness changes with income. Thus, ıH/ıY > 0 provided that a > bc. Theedonic Treadmill is represented by a = bc, i.e. the positive effect of a rise in income on happiness is precisely offset by theegative effect of the consequent rise in aspiration income.

Current income is not the only potential determinant of aspiration income. For instance, past income may also have annfluence. We hypothesise that a recent rise in income raises happiness but a recent fall in income lowers it. This is plausiblef people have not yet fully adapted to their recent change in income. With Y given, we expect Yp, representing past income,o have a positive sign – the higher is Yp the higher is aspiration income, A. Another potential influence is the reference groupith which people compare themselves. Our hypothesis is that the higher the income of the reference group, Yr, the higher

s A.Our full specification for the equation predicting aspiration income is therefore:

A = A(Y, Yp, Yr, . . .) (4)

here ıA/ıY > 0, ıA/ıYp > 0, ıA/ıYr > 0. Our specification of the happiness function is:

H = H(Y, A, . . .) (5)

here ıH/ıY > 0, ıH/ıA < 0. However, our measure of A might be imperfect, in which case other proxies for aspiration incomean also affect happiness. Moreover, aspirations might have an effect on happiness independently of their effect on aspirationncome, A. It would accordingly be appropriate to include such variables as Yp and Yr in Eq. (5) as well as in Eq. (4).


These equations look too simple: there are several potential econometric problems to be solved. With regard to Eq. (4),it is possible that Y is a function of A, or that both Y and A are influenced by unobserved variables. Similarly both Y and Amight be endogenous in Eq. (5).

4. Data, variables and background

The data come from the national household survey for China, conducted in 2003 and relating mainly to 2002, of theInstitute of Economics, Chinese Academy of Social Sciences (IE, CASS). The survey was conducted by the National Bureauof Statistics (NBS) and was based on a representative sub-sample of the annual official household survey. It was designedby a research ream of Chinese and foreign scholars, including one of the authors, with research hypotheses in mind, andthus the questionnaires contained many more questions than did the official survey. There were separate sub-samples andquestionnaires for rural and for urban residents, and for rural–urban migrants. In this paper we concentrate on the ruralsurvey, for which we designed a special module of the questionnaire in order to investigate subjective well-being. We arethus able to combine the information on subjective well-being with a wealth of socioeconomic data on rural communities,households and their individual members.

The rural survey covers 22 of China’s 31 provinces and is intended to be nationally representative. Within each provinceon average 5.5 counties were sampled, and within each county on average 7.9 villages; selection was on the basis of incomeper capita ranking with a random starting point. Within each village normally 10 households were randomly sampled. Therural sample thus contains roughly 9500 households.

In each household one member – normally the household head – was asked a set of attitudinal questions, including someabout subjective well-being. A novelty of this data set is that it contains information on different dimensions of subjectivewell-being. In this paper we analyse two dimensions that were distinguished. The broader one is ‘happiness’ or ‘satisfactionwith life’ (we use the terms interchangeably) and the narrower is ‘satisfaction with income’. In each case the respondentswere offered five choices: very satisfied, satisfied, so-so, dissatisfied, and not at all satisfied. When treated as a dependentvariable, these answers can be converted into a cardinal value (with very satisfied given a score of 4, down to not at allsatisfied, given a score of 0), or into a binary variable (distinguishing very satisfied or satisfied from the three categoriesbelow them), or into a ranking of three categories (very satisfied or satisfied, so-so, and dissatisfied or very dissatisfied). Thesethree dependent variables can then be analysed by means of OLS, probit, or ordered probit estimation models respectively. Infact our estimations produced no notable differences in the three sets of results, and we therefore report only the estimatesfor the happiness score.3

Several other attitudinal or subjective questions were asked, including the minimum income needed by the household.This we shall take as a potential proxy for aspiration income, to be explained by a set of demographic, economic, socialand aspirational variables. Respondents were asked about the change in their income over the previous five years, theirexpectations for their income over the following five years, the people with whom they compared themselves, and abouttheir perceived position in the village income distribution. We shall introduce these variables as they enter the analysis.

Table 1 reports the percentage distribution of respondents amongst the five categories of satisfaction, for both subjectivewell-being variables. We see from column 1 that many people are satisfied (rows 1 and 2), and few are dissatisfied (rows4 and 5), in the case of the more general measure (satisfaction with life), but that the division is even in the case of themore specific measure (satisfaction with income). This distinction is confirmed in the mean scores: 2.7 for life and 2.0 forincome satisfaction. When the distributions are shown by income per capita quintile, in each case there is a monotonic risein satisfaction as we move up the income quintiles: income does indeed contribute to subjective well-being. Moreover, thedifference in mean score between the highest and the lowest quintile is greater for satisfaction with income (0.7) than forhappiness (0.5). This confirms our expectation that income is more important for yielding satisfaction with income than itis for yielding satisfaction with life.

Define the ‘subjective well-being poor’ as those who are not satisfied or not at all satisfied with life, or with income(Kingdon and Knight, 2006). The proportion of households in ‘satisfaction with life poverty’ in our data set is 9 per centand the proportion in ‘satisfaction with income poverty’ is 37 per cent. Defining the ‘income poor’ as constituting the sameproportions of households in these two forms of subjective well-being poverty, 19 per cent of the income poor are satisfactionwith life poor, and 48 per cent are satisfaction with income poor. At the lower tails of the three distributions there is some,but very incomplete, overlap. You can have low income and yet be happy, and you can have low income and yet be satisfiedwith it. The level of aspirations needs to brought into the story.

5. Testing strategy

We follow Stutzer (2004) in our strategy for testing the hypotheses. His measures of aspiration income were the income
that respondents reported as being either ‘sufficient’ or as the ‘minimum required’; the two measures produced very similarresults. His aspiration income equation included household income in the previous year, a report of financial situation in thepast, and conditioning variables. Aspiration income was a positive function of household income (the coefficient being about
3 This was also found to be the case in a methodological paper by Ferrer-i-Carbonnel and Frijters (2004).


Table 1Percentage distribution of subjective well-being, overall and by income per capita quintile, rural China.

Overall Income per capita quintile

1st 2nd 3rd 4th 5th 5th–1st

HappinessVery happy 15.8 10.3 13.3 14.1 19.3 21.7 11.4Happy 46.5 36.8 42.6 49.2 50.4 53.4 16.6So-so 28.9 36.2 33.7 28.6 25.1 20.8 −15.4Not happy 7.9 14.3 9.0 7.5 4.9 3.9 −10.5Not happy at all 0.9 2.3 1.3 0.6 0.3 0.2 −2.2Total (number) 8996 1799 1799 1799 1799 1800Mean happiness 2.7 2.4 2.6 2.7 2.8 2.9 0.5Satisfaction with incomeVery satisfied 10.1 6.1 7.0 9.5 12.0 15.9 9.8Satisfied 29.2 19.6 25.5 27.9 34.4 38.4 18.8So-so 24.0 22.9 23.8 25.3 25.1 22.9 0.0Not satisfied 23.7 30.1 30.2 24.1 19.4 14.6 −15.5Not satisfied at all 13.1 21.3 13.6 13.2 9.2 8.2 −13.1Total (number) 8996 1799 1799 1799 1799 1800Mean satisfaction 2.0 1.6 1.8 2.0 2.2 2.4 0.7

Notes: Level of happiness is based on cardinal values that are assigned to qualitative assessments as follows: very happy = 4; happy = 3; so-so = 2; nothappy = 1 and not at all happy = 0.Level of satisfaction with income is based on cardinal values assigned to qualitative assessments as follows: very satisfied = 4; satisfied = 3; so-so = 2; notsatisfied = 1 and not at all satisfied = 0.Data for this table and for all subsequent tables are derived from the Rural Household Survey (2002).Tl

0(tcceiasa

iio

6

itoow

mr0

ioswcd

he income variable is household income per capita. Incomes in this table, and all subsequent tables, are corrected for spatial differences in the cost ofiving by means of the spatial price indices calculated by Brandt and Holz (2006).

.4, with the income variables in logarithmic form) and of past financial situation. Another equation included relative incomeaverage income in the community, there being about 500 communities). There was a significantly positive coefficient onhe community variable. Because this might arise if the cost of living is positively correlated with the prosperity of theommunity, a proxy for the strength of social comparison – a report of contact with neighbours – was interacted withommunity income. The interaction term also had a positive coefficient: more social interaction strengthens the positiveffect of community income on aspiration income. Turning to the happiness function: when (the logarithm of) aspirationncome was excluded, the household income variable had a significantly positive effect (about 0.25) in the equation. Whenspiration income was also included, the income coefficient was raised (to about 0.40) and the aspiration coefficient wasignificantly negative (about −0.35). The results were found to be robust when aspiration income was instrumented to guardgainst the possibility of unobserved personal traits (such as neuroses) lowering happiness but raising aspirations.

We shall similarly estimate aspiration income functions that include current income, previous income, and relativencome amongst the arguments. These aspiration income estimates will lead on to subjective well-being functions thatnclude both aspiration income and actual income as explanatory variables. It will be necessary to test for the endogeneityf both aspiration income and actual income and to correct for this by means of instrumenting.

. The determinants of aspiration income

Aspiration income (the variable A in Eqs. (1)–(5)) is easier to conceptualise than to quantify. Our potential proxy for As the respondent’s answer to the question: what is the minimum income needed to sustain the household for a year? Ifhis is found to be based only on demographic and physiological factors, then it should be viewed simply as a measure ofbjective need. However, if it depends also on characteristics such as income – absolute or relative to the reference groupr the reference time – then there is a subjective element and it might well be viewed as aspiration income. That is indeedhat we find.

We begin with some descriptive statistics. We calculate the mean household income per capita and the mean reportedinimum household income needed per capita, by income per capita decile. Income need rises across the deciles but less

apidly than does actual income. The ratio of needed to actual income falls monotonically. It is 1.77 in the first decile and.31 in the tenth, the crossover point being between the third decile (1.12) and the fourth (0.84).

Table 2 reports our estimates of the determinants of minimum income need. The dependent variable takes two forms:ncome need in levels and in logs. In each case we use a full set of explanatory variables, including potential determinantsf basic income need as well as current income and several other potential determinants of aspiration income. Becausepatial differences in the cost of living could result in a positive association between actual income and aspiration income,
e attempt to correct for this by deflating both income measures using the spatial price differences for China that were
alculated by Brandt and Holz (2006). Indeed, all our tables are adjusted in this way. We tried two alternative ways ofealing with household size effects. One includes ln of household size and the other the number of household members in


Table 2Determinants of income need: OLS estimation.

Mean/proportion

Absolute Logarithm

(1) (2) (3) (4)

Coefficient se Coefficient se Coefficient se Coefficient se

Real total household income(Yuan)

9757.57 0.12*** 0.02 0.10*** 0.02

Log of real total householdincome

8.98 0.210*** 0.024 0.187*** 0.025

Other aspiration variablesLog of financial assets 0.97 −68.63 86.39 −0.008 0.011Log of household debt 0.22 872.55*** 181.57 0.082*** 0.016Current living standards better

than 5 years ago0.60 −243.41 231.65 0.000 0.029

Current living standards worsethan 5 years ago

0.05 952.02** 432.70 0.190*** 0.049

Agree that money is important 2.35 203.95 144.57 0.045*** 0.016Education (years) 7.18 160.09*** 42.41 0.025*** 0.005Phone 0.40 348.38 245.54 0.109*** 0.030Main reference group beyond

village0.12 789.20** 398.91 0.103** 0.045

Household income much abovevillage average

0.02 204.97 511.45 0.066 0.082

Household income abovevillage average

0.18 407.43* 237.92 0.047 0.029

Household income belowvillage average

0.19 177.54 210.25 0.044 0.027

Household income muchbelow village average

0.03 1483.95* 814.04 0.169** 0.082

Conditioning variablesAge 45.13 238.94*** 52.55 208.79*** 50.36 0.028*** 0.007 0.023*** 0.007Age squared 2146.64 −2.68*** 0.57 −2.23*** 0.54 −0.0003*** 0.0000 −0.0003*** 0.0000Male 0.75 −163.49 188.17 −336.30* 187.54 −0.032 0.027 −0.054** 0.026Married 0.96 754.20* 393.87 867.80** 412.27 0.176* 0.093 0.184** 0.090Divorced 0.00 539.42 930.58 761.84 930.06 0.229 0.184 0.259 0.173Widowed 0.02 305.49 545.71 526.83 555.43 0.105 0.115 0.135 0.111In good health 0.75 −598.94*** 217.82 −508.77** 209.13 −0.091*** 0.028 −0.080*** 0.028Hilly terrain 0.30 568.62* 303.53 567.62* 299.43 0.065 0.046 0.072 0.044Mountainous terrain 0.26 1340.30*** 502.74 1329.48*** 489.81 0.073 0.059 0.077 0.056Distance from county seat

(100 km)0.24 201.48 803.24 125.21 808.57 0.115 0.128 0.107 0.129

Satisfaction with clinic 2.34 −301.35*** 105.66 −294.34*** 103.05 −0.049*** 0.015 −0.049*** 0.015Ethnic minority dummy 0.14 −540.50 406.26 −411.02 401.25 −0.041 0.063 −0.019 0.062Senior citizens, male, aged 65+ 0.10 1203.69*** 297.63 1194.58*** 287.00 0.174*** 0.034 0.174*** 0.033Senior citizens, female, aged

65+0.11 724.43*** 243.59 685.52*** 238.05 0.116*** 0.029 0.111*** 0.029

Adult males of 18–64 years 1.47 668.77*** 127.80 656.95*** 126.44 0.120*** 0.018 0.123*** 0.017Adult females of 18–64 years 1.39 898.66*** 134.94 875.46*** 135.69 0.143*** 0.015 0.140*** 0.015Teenaged males of 11–17 years 0.36 324.87** 150.54 314.59** 150.32 0.102*** 0.019 0.105*** 0.019Teenaged females of 11–17

years0.31 443.13*** 133.24 419.94*** 128.41 0.105*** 0.017 0.103*** 0.016

Children, males less than 11years

0.27 38.88 171.94 79.32 171.72 0.014 0.022 0.024 0.021

Children, females less than 11years

0.22 306.67 187.20 298.13 185.03 0.029 0.023 0.032 0.022

Constant −2826.01** 1208.91 −4379.50*** 1360.49 5.545*** 0.247 5.426*** 0.256

R-squared 0.064 0.089 0.117 0.148Number of observations 6384 6384 6192 6192

Notes: Dependent variables: minimum income needed (Yuan) (mean 6031, standard deviation 6059); logarithm of minimum income needed (mean 8.458,standard deviation 0.730).Independent variables with cardinal values assigned to qualitative assessments so that greater intensity is represented by a higher value are: satisfactionwith clinic, agreement that money is important.The omitted categories in the dummy variable analyses are: female sex; married; plains; not healthy; in normal or worse than normal mood; current livingstandard the same as five years ago; main reference group within village; household at average village income.

* Statistical significance at the ten per cent level.** Statistical significance at the five per cent level.

*** Statistical significance at the one per cent level.The models have been clustered at village level for robust standard errors.


Table 3Determinants of income need with income instrumented.

Absolute Logarithm

(1) (2) (3) (4)

Coefficient se Coefficient se Coefficient se Coefficient se

Real total household income(Yuan)

0.23*** 0.05 0.20*** 0.07

Log of real total householdincome

0.572*** 0.120 0.707** 0.275

Log of financial assets −194.11** 97.19 −0.067** 0.034Log of household debt 812.65*** 109.49 0.056*** 0.018Household income below

village average373.07* 211.76 0.125*** 0.042


2043.51*** 485.42 0.373*** 0.102

Hilly terrain 578.08*** 184.87 486.70*** 180.93 0.053** 0.024 0.024 0.029Mountainous terrain 1526.89*** 228.74 1325.63*** 214.78 0.127*** 0.038 0.094** 0.038Number of observations 5503 5503 5317 5317

Significance of exclusion restrictions in first stage equationFather’s years of education *** ns *** **

Productive assets *** *** *** ***

F-test of excluding instrumentsF statistic 140.53 80.73 48.46 11.55P-value 0.0000 0.0000 0.0000 0.0000Hansen J test for

overidentification of allinstruments (P-value)

0.8136 0.9357 0.8585 0.6133

Anderson Rubin test of jointsignificance of endogenousregressors in main equation,F test (P-value)

0.0000 0.0108 0.0000 0.0198

Notes: The notes to Table 2 apply.All the independent variables shown in Table 2 are included in the equations but only those referred to in the text are reported.Instrumented variable regression results are generated using Baum et al.’s (2007), ivreg2.ado programme for extended instrumental variables/2SLS, GMMand AC/HAC, LIML, and k-class regression. See, Baum, C.F., Schaffer, M.E., Stillman, S., 2007. Enhanced routines for instrumental variables/generalizedm

eo

ibipseq

ctmytrpg

ciiipT

ethod of moments estimation and testing. Stata Journal 7, 465–506.

ach age–sex group. As there are no notable differences in the estimated coefficients of all the other variables, we presentnly one set of results, choosing the latter specification as it proves to be more informative.

We start with the variables that might reflect basic income need. The coefficients of both age and age squared are signif-cantly different from zero. In the levels equation, the income needed by the household rises with the age of the respondentut at a diminishing rate, and peaks at the age of 47 (column 2). Men perceive that the household needs significantly less

ncome than do women, but only in the full specifications (columns 2 and 4). This might reflect a greater concern on theart of women for the care of children and other household members. By comparison with (uncommon) single status (andtandardising for household composition as well as income), married respondents perceive higher income need. As is to bexpected, being in good health has a significant negative coefficient. The degree of satisfaction with the village clinic (theuality of local health facilities) significantly reduces income need, possibly by reducing perceived insecurity.

The household composition variables have significantly positive coefficients, as expected, except in the case of younghildren. Representing the additional income needed on account of there being one more person in each age–sex group,hey are highest for men over 64 and lowest for boys under 11. In the levels equation (column 2) this set of coefficients

ight be viewed as providing a set of weights for the income needs of different types of household member. They effectivelyield subjectively determined ‘adult equivalent’ scales (an approach pioneered by Kapteyn and Van Praag, 1976). Thus, ifhe need of an adult male is 100, then that of an adult female is 132, those of senior citizen males and females 182 and 104espectively, those of male and female teenagers 48 and 64, and those of male and female children 12 and 45. This surprisingattern, mirrored in the logarithmic equation, suggests that, except in the case of old people, females are perceived to havereater income needs than males.

We turn to the variables which might influence aspirations for income, concentrating on those that have significantoefficients. The income per capita variable has a coefficient that is both positive and significant. In levels form, the coefficients 0.10 in the full specification (column 2), i.e. a rise in household income per capita increases the income that the households perceived to need by a tenth of that amount. In logarithmic form, the elasticity of needed income with respect to actual
ncome has a value of 0.19 in the full specification (column 4). These results suggest that aspirations for income adapt inart to current income. We include the gross values of the household’s financial assets and debts, both in logarithmic form.he debts term is positive and significantly so (column 4): indebtedness raises income need.


By comparison with those who had experienced the same living standards (the reference dummy category), incomeneed is significantly higher if the respondent had experienced a fall in living standards over the previous five years – evenafter standardising for current income level. This suggests that aspiration income is influenced by previous income andthat aspirations do not fully adapt to a fall in income. The cardinal variable denoting the degree of importance attached tomoney has a significantly positive coefficient in the logarithmic equation. This might result either from having an objectiveneed for higher income or from having a materialistic personality. We hypothesised that those who reported that moneyis important would have a higher coefficient on the income variable than those who did not, but the difference was notsignificant (equations not reported).

Some variables that are not necessarily economic also appear to affect aspiration income. The more years of educationthe respondent has received, the higher the income needed: the coefficient is positive, and significantly so. Householdpossession of a phone raises the income needed substantially and significantly in the logarithmic equation, a result whichmight reflect having broader and richer reference groups or – reverse causation – people with high aspirations accumulatemore household possessions.

Column 4 adds variables to the logarithmic equation that are intended to show the effect of comparator income. Over two-thirds of the sample report that their main reference group is within the village. Having the main reference group beyond thevillage – where incomes are likely to be higher – raises the perceived income needed substantially and significantly. Whenthis variable is interacted with the log of household per capita income, the coefficient is negative (equation not reported),suggesting that the effect of making comparisons beyond the village on aspirations falls as the income disparity declines.

Respondents were asked whether their income was much above, above, at, below, or much below the average income ofthe village. Four dummy variables were created, with income at the village average being the reference category. When theperceived income rank within the village is included, we see that those who have income much below the village averagehave aspiration income substantially and significantly higher than those at the village average. Attitudes matter: when thesample was divided into those apparently with materialistic attitudes (‘agree that money is important or very important’)and those without, the most notable difference was that the perception of having income well below the village averageraised the aspiration income only of the more materialistic group (results not shown).

Data on actual position in the village income distribution were not available because only ten households were sampledin each village. There is anyway a case for using perceived rather than actual position because it is perception that gives riseto feelings of relative deprivation. When we substituted the mean income per capita of the other nine sampled householdsin the village for the dummy variables, the coefficient was never significant (equations not shown). Although this lack ofsignificance may be due to the small number of households that were sampled in each village, the possibility thus remainsthat aspirations affect perceived position.4

Table 3 shows the result of instrumenting household income. It is difficult to find variables in the data set which determineincome but do not plausibly affect aspiration income as well. We eventually chose a combination of father’s years of educationand productive assets; each is not at all significant if included in the OLS equations for aspiration income. The instrumentingpasses the standard statistical tests, as reported at the end of the table: the instruments are neither weak nor invalid, andthey are needed. However, the main test of good instruments is their theoretical and contextual plausibility. Bearing in mindthat socialisation effects are unlikely to be relevant when fathers average less than three years of education, our view is thatfather’s education does not affect income need. Similarly, we do not expect productive assets (comprising farm machineryand off-farm equipment and excluding houses and financial assets) to have an influence.

The equations contain all the explanatory variables of Table 2 but we report only those on which we wish to comment:there is little notable change in the coefficients of the other explanatory variables. The income coefficient remains signifi-cantly positive in both equations, but it is roughly doubled in the equations in levels and more or less tripled in the equationsin logs. The latter results imply that the elasticity of income need to current income exceeds one half. Since unobserved vari-ables which might plausibly raise income (e.g. ambition or ability) would be likely also to raise, rather than lower, aspirations,so producing upward bias, the increase in coefficients came as a surprise; it might reflect correction for downward attenua-tion bias. However, caution is required: the standard errors of the income coefficient are higher than in the OLS case: the 5per cent significance ranges of the elasticities are 0.327–0.807 (column 3) and 0.168–1.246 (column 4). Also in Table 3, wesee that gross financial assets have a significant negative effect on aspiration income. Respondents with household incomebelow the village average now aspire to significantly more income than those at the average, but the effect is weaker thanfor those much below the average. Hilly and, especially, mountainous terrains generally raise the income needed by thehousehold, suggesting that physical needs or unobserved aspects of the cost of living are greater in those environments.

Our estimate of the effect of income on aspiration income might still be too low because the cross-section nature of thedata set cannot account for lags in the adjustment of aspiration income. Ceteris paribus, insofar as those with current income
above their normal income report lower minimum income needs, and vice versa, the coefficient on income in the aspirationincome equation is likely to be biased downwards.5
4 When the equations of Table 3 were re-estimated with province dummies added (results not presented), none of the province coefficients was significant,and the other coefficients (reflecting within-province relationships) were generally slightly closer to zero but there was no notable change in patterns orsignificance. A proxy for the remoteness, distance (in kilometres) from the county seat, proved not to be significant.

5 This argument is analogous to that of Friedman (1957) in explaining the cross-section consumption function.

7

–av

satiTa

locncwi−

dbgssc

aadpthtm

wobtc

smcsaiob

hi

gw


. The aspirational determinants of happiness

Do aspirations affect happiness and, if so, how and to what extent? Table 4 reports the OLS subjective well-being functions first predicting satisfaction with life and then satisfaction with income. Column 1 presents the basic variables, column 2dds only our central aspiration variable, ln of minimum income needed, and column 3 includes a full set of aspirationariables.

Insofar as the specifications overlap, the coefficients on the basic variables are similar to those obtained in our previoustudy of satisfaction with life in rural China (Knight et al., 2009). For instance, we find the familiar U-shaped relationship withge; being male, being divorced or widowed (instead of married or unmarried), working longer hours, and living further fromhe county seat reduce satisfaction; and education, gross financial assets, and being in good health raise satisfaction. Beingn a better than normal mood – significantly positive – is introduced to reduce the unexplained variance of the estimates.he satisfaction with income equation is less well determined. Of the variables noted above, only good health, good mood,nd gross financial assets are significant (and positive), and household debts now have a significant negative effect.

The first two rows of results are the most important. Predictably, the coefficient on ln of household income per capita isarger in the case of satisfaction with income than in the case of satisfaction with life. Our interest lies in the introductionf the aspiration variables. The introduction of ln of per capita minimum income needed raises the coefficient on ln of perapita income in the happiness equation, from 0.18 in column 1 to 0.20 in column 2, but ln of per capita minimum incomeeeded itself has a negative coefficient (−0.08). The same pattern is found for satisfaction with income (the correspondingoefficients being 0.33, 0.35 and −0.09 respectively). When other variables that might influence aspirations are entered asell (column 3), the effect is to reduce the size of the coefficients but to retain the pattern. Thus, the coefficients on the

ncome and income need variables become 0.12 and −0.06 respectively in the satisfaction with life equation, and 0.24 and0.07 in the satisfaction with income equation. All of these coefficients are highly significant.

The other variables that might influence aspirations generally also have significant coefficients. One exception is theummy variable indicating that the respondent had lived outside the township for at least a year. This was introducedecause it could represent possession of more information about life beyond the village and the broadening of the referenceroup to include generally richer people. In both equations the coefficient is negative, as expected, but it is significantlyo only in the prediction of satisfaction with income. The cardinal variable indicating the strength of agreement with thetatement that money is important has a significant negative coefficient only in the satisfaction with life equation. Althoughausation might run either way, this is consistent with possession of a materialistic personality reducing happiness.6

Both dependent variables are highly sensitive to perceived position in the village income distribution; the relationshipsre monotonic. Aspirations are framed by the reference group, and feelings of relative deprivation give rise to unfulfilledspirations and thus to dissatisfaction. Not only the reference group but also the reference time can have a bearing. Respon-ents were asked whether their current living standard was better or worse than, or the same as, five years previously. Theositive coefficient on the dummy variable indicating a rise and the negative coefficient on a fall over that period suggesthat aspirations are partly set in the past. Table 2 provides evidence of that aspiration income was higher if the respondentad suffered a fall in income over the past five years. It seems that aspirations for income (as we measure them) are nothe only way in which aspirations can affect subjective well-being. Even standardising for aspiration income, we see that

emories of time past influence subjective well-being present.Respondents were also asked how they expected their income to change over the next five years. In both equations

e see a monotonic and powerful rise in coefficient values as the expected increase in income rises (no change being themitted category). It seems that people internalise their expected future income into their current subjective well-being,ut also that they cannot predict the rise in aspirations which is likely to accompany that rise in income. This squares withhe psychological evidence, referred to above, that people are less good at predicting their aspirations than their objectiveonditions.7

The OLS estimates of Table 4 are open to the criticism that the key variables might be endogenous. For instance, posses-ion of a certain type of personality might both raise income and lower happiness, and dissatisfaction with life or incomeight cause people to report that they need more income. Accordingly we re-estimated the equations with both ln of per

apita income and ln of per capita income need instrumented. This involved a systematic search for instruments that weretrong, were valid, and were unlikely directly to influence subjective well-being. The statistical test results are reportedt the end of Table 5: in each case the excluded instruments are not weak, they pass the Sargan test of validity wherever
t can be applied, and they are generally needed. The instruments that were eventually chosen for ln of income were twor more of father’s years of education, productive assets, and the eight household composition variables. Various com-inations of father’s and spouse’s education and the eight household composition variables were used for ln of income
6 When the sample was divided into those with materialistic attitudes (‘agree that money is important or very important’) and those without, theappiness of the more materialistic group was associated with greater sensitivity to household income per capita, to perceived position in the village

ncome distribution, to change in income over the past five years, and to expected change in income over the next five years (equation not reported).7 When Table 5 was re-estimated with province dummy variables added (results not reported), the province coefficients in the happiness equation were

enerally significantly different from zero (the omitted province being Beijing). The numerical value of most of the other (within-province) coefficientsas slightly reduced but there was no notable change in the pattern or significance of any of the important coefficients.


Table 4Determinants of two different concepts of subjective well-being: satisfaction with life and satisfaction with income: OLS estimation.

Mean/proportion (1) (2) (3)

Coefficient se Coefficient se Coefficient se

Satisfaction with lifeLog of per capita household

income (‘000 Yuan)7.58 0.179*** 0.023 0.197*** 0.023 0.122*** 0.021

Aspiration variablesLog of per capita minimum

income needed7.07 −0.076*** 0.020 −0.058*** 0.019

Lived outside township for atleast a year

0.13 −0.045 0.032

Expect big increase inincome over next 5 years

0.10 0.187*** 0.042

Expect small increase inincome over next 5 years

0.68 0.056* 0.030

Expect decrease in incomeover next 5 years

0.04 −0.135** 0.054

Agreement with statementthat money is important

2.35 −0.034** 0.016

Main reference groupbeyond village

0.11 −0.038 0.041

Household income muchabove village average

0.02 0.263*** 0.070


0.18 0.089*** 0.028


0.20 −0.314*** 0.030


0.03 −0.794*** 0.075

Current living standardsbetter than 5 years ago

0.60 0.204*** 0.029

Current living standardsworse than 5 years ago

0.05 −0.100* 0.058

Basic variablesAge (years) 45.12 −0.014** 0.007 −0.013* 0.007 −0.016** 0.007Age squared 2146.07 0.0002*** 0.0001 0.0002*** 0.0001 0.0002*** 0.0001Male 0.75 −0.066** 0.027 −0.072*** 0.027 −0.070*** 0.026Married 0.96 0.133 0.111 0.139 0.107 0.123 0.104Divorced 0.00 −0.468* 0.248 −0.447* 0.251 −0.507** 0.237Widowed 0.02 −0.291** 0.137 −0.286** 0.134 −0.208* 0.126Ethnic minority dummy 0.14 0.058 0.051 0.055 0.050 0.072 0.048Education (years) 7.14 0.010** 0.005 0.012*** 0.005 0.007* 0.004Working hours (’00 per year) 17.01 −0.003** 0.001 −0.003* 0.001 −0.002* 0.001Log of financial assets 0.95 0.042*** 0.009 0.041*** 0.009 0.033*** 0.008Log of household debt 0.22 0.002 0.015 0.008 0.015 0.026* 0.014In good health 0.74 0.419*** 0.029 0.410*** 0.028 0.338*** 0.027In a good mood 0.65 0.569*** 0.028 0.566*** 0.028 0.431*** 0.028Distance from county seat

(100 km)0.24 −0.262*** 0.096 −0.251*** 0.096 −0.255*** 0.090

Constant 0.711*** 0.226 1.057*** 0.244 1.730*** 0.240

R-squared 0.234 0.238 0.314Number of observations 6497 6497 6497

Satisfaction with incomeLog of per capita household

income (‘000 Yuan)7.58 0.329*** 0.033 0.350*** 0.034 0.236*** 0.031

Aspiration variablesLog of per capita minimum

income needed7.07 −0.088*** 0.028 −0.068*** 0.026

Lived outside township for atleast a year

0.13 −0.107** 0.051

Expect big increase inincome over next 5 years

0.10 0.452*** 0.069

Expect small increase inincome over next 5 years

0.68 0.112*** 0.042


Table 4 (Continued)

Mean/proportion (1) (2) (3)


Expect decrease in incomeover next 5 years

0.04 −0.254*** 0.078

Agreement with statementthat money is important

2.35 0.033 0.023

Main reference groupbeyond village

0.11 −0.145** 0.058

Household income muchabove village average

0.02 0.404*** 0.124


0.18 0.345*** 0.041


0.20 −0.477*** 0.043


0.03 −0.809*** 0.095

Current living standardsbetter than 5 years ago

0.60 0.261*** 0.044

Current living standardsworse than 5 years ago

0.05 −0.212*** 0.066

Basic variablesAge (years) 45.12 0.001 0.011 0.003 0.011 −0.006 0.010Age squared 2146.07 0.0001 0.0001 0.0001 0.0001 0.0002* 0.0001Male 0.75 −0.049 0.043 −0.056 0.043 −0.047 0.040Married 0.96 0.145 0.104 0.151 0.106 0.127 0.098Divorced 0.00 0.185 0.295 0.209 0.295 0.113 0.312Widowed 0.02 0.023 0.150 0.028 0.151 0.143 0.142Ethnic minority dummy 0.14 −0.030 0.072 −0.033 0.072 −0.008 0.068Education (years) 7.14 0.006 0.007 0.008 0.007 0.003 0.007Working hours (’00 per year) 17.01 0.000 0.002 0.000 0.002 0.000 0.002Log of financial assets 0.95 0.031** 0.013 0.030** 0.013 0.019 0.012Log of household debt 0.22 −0.121*** 0.022 −0.114*** 0.021 −0.091*** 0.020In good health 0.74 0.397*** 0.043 0.387*** 0.043 0.275*** 0.040In a good mood 0.65 0.398*** 0.038 0.394*** 0.038 0.197*** 0.037Distance from county seat

(100 km)0.24 −0.183 0.147 −0.170 0.149 −0.202 0.133

Constant −1.498*** 0.320 −1.100*** 0.344 −0.201 0.328

R-squared 0.124 0.127 0.224Number of observations 6497 6497 6497

Notes: Dependent variables: score of happiness based on cardinal values assigned to qualitative assessments as follows: very happy = 4; happy = 3; so-so = 2; not happy = 1 and not at all happy = 0. Score of satisfaction with income based on cardinal values assigned to qualitative assessments as follows:very satisfied = 4; satisfied = 3; so-so = 2; not satisfied = 1 and not at all satisfied = 0. The mean values of happiness and satisfaction with income are 2.64 and1.93, their standard deviations 0.88 and 1.22 and their coefficients of variation 0.33 and 0.63 respectively.Independent variables with cardinal values assigned to qualitative assessments so that a higher value denotes greater intensity: agreement that money isimportant.The omitted categories in the dummy variable analyses are: female sex; married; not healthy; in normal or worse than normal mood; household at averagevillage income; current living standard the same as five years ago.

* Statistical significance at the ten per cent level.** Statistical significance at the five per cent level.

*** Statistical significance at the one per cent level.

nodc

niTrihwe

eeded. Assets that are not normally held as a form of wealth, the details of household composition, and the education ofther persons are unlikely in themselves to influence subjective well-being. Nevertheless, the validity of the instrumentingepends on the plausibility of the instruments, and ours – as those in many other cross-section studies – are at least open tohallenge.

The precision of the estimated coefficients in Table 5 is generally lower than in Table 4, but the pattern of results is nototably affected, e.g. there are no statistically significant changes of sign. Accordingly, we report only the coefficients of the

nstrumented variables. The instruments are not weak, they are needed, and the Sargan test is passed in each possible case.he effect of instrumenting income is to raise the coefficient, to 0.55 and 0.59 for satisfaction with life and with incomeespectively (column 1). When aspiration income is introduced and instrumented as well as income, the pattern of Table 4s again observed (columns 2 and 3). The coefficient of the aspiration income variable is not only negative but in each case
as a larger absolute value (although these differences could have arisen by chance). However, despite much searching, itas not possible to find instruments which could generate statistically significant coefficients for these variables. The causal
ffect of aspiration income on subjective well-being requires further research.


Table 5Determinants of subjective well-being with income and income need instrumented.

(1) (2) (3)


Satisfaction with lifeLog of per capita household income

(‘000 Yuan)0.546*** 0.135 1.352 1.858 0.311 0.214

Log of per capita minimum incomeneeded

−0.617 1.615 −0.099 0.281

Number of observations 5879 3921 5620

Significance of exclusion restrictions for ln of per capita household income in first stage equationFather’s years of education ** * **

Spouse’s years of education ***

Productive assets ***

Senior citizens, male, aged 65+ ***

Senior citizens, female, aged 65+ ***

Adult males of 18–64 years ***

Adult females of 18–64 years ***

Teenaged males of 11–17 years ***

Teenaged females of 11–17 years ***

Children, males less than 11 years ***


***

F-test of excluding instrumentsF statistic 46.61 23.49 55.99P-value 0.0000 0.0000 0.0000Significance of exclusion restrictions for ln of need minimum income in first stage equationFather’s years of education ns nsSpouse’s years of education ***

Productive assetsSenior citizens, male, aged 65+ nsSenior citizens, female, aged 65+ **

Adult males of 18–64 years ***

Adult females of 18–64 years ***

Teenaged males of 11–17 years ***

Teenaged females of 11–17 years ***

Children, males less than 11 years ***


***

F-test of excluding instrumentsF statistic 22.82 22.72P-value 0.0000 0.0000Sargan test for overidentification of

all instruments (P-value)0.8043 0.2177

Anderson Rubin test of jointsignificance of endogenousregressors in main equation, Ftest (P-value)

0.0002 0.0013 0.0011

Satisfaction with incomeLog of per capita household income

(‘000 Yuan)0.586*** 0.196 0.574* 0.328 0.512 0.332

Log of per capita minimum incomeneeded

−0.384 0.460 −0.354 0.447

Number of observations 5879 6617 6538

Significance of exclusion restrictions for ln of per capita household income in first stage equationFather’s years of education **

Productive assets ***

Senior citizens, male, aged 65+ *** ***

Senior citizens, female, aged 65+ *** ***

Adult males of 18–64 years *** ***

Adult females of 18–64 years *** ***

Teenaged males of 11–17 years *** ***

Teenaged females of 11–17 years *** ***

Children, males less than 11 years *** ***


*** ***

F-test of excluding instrumentsF statistic 46.61 79.24 75.86P-value 0.0000 0.0000 0.0000


Table 5 (Continued)

(1) (2) (3)


Significance of exclusion restrictions for ln of need minimum income in first stage equationFather’s years of educationSenior citizens, male, aged 65+ ns nsSenior citizens, female, aged 65+ *** ***

Adult males of 18–64 years *** ***

Adult females of 18–64 years *** ***

Teenaged males of 11–17 years *** ***

Teenaged females of 11–17 years *** ***

Children, males less than 11 years *** ***


*** ***

F-test of excluding instrumentsF statistic 29.39 29.47P-value 0.0000 0.0000Sargan test for overidentification of

all instruments (P-value)0.7140 0.2752 0.1696

Anderson Rubin test of jointsignificance of endogenousregressors in main equation, Ftest (P-value)

0.0105 0.0023 0.0061

Notes: The notes to Table 4 apply.All the explanatory variables included in Table 4 are included in the equations but only the variables of interest are reported in the table.The number of observations falls when spouse’s education is used as an exclusion restriction.Iam

8

1eawwtow

eabgso

noicihht

aist

a

nstrumented variable regression results are generated using Baum et al.’s (2007), ivreg2.ado programme for extended instrumental variables/2SLS, GMMnd AC/HAC, LIML, and k-class regression. See, Baum, C.F., Schaffer, M.E., Stillman, S., 2007. Enhanced routines for instrumental variables/generalizedethod of moments estimation and testing. Stata Journal 7, 465–506.

. Concluding comments

Economists have long recognised that economic aspirations and well-being are contextual (Smith, 1776, p. 466; Marx,849, p. 163). Knight (1922) argued that wants and aspirations are malleable – they change easily and unpredictably and arendogenous to income and consumption – and for this reason can undermine the notion of utility as a criterion for normativenalysis. Robinson (1956, p. 15) noted that ‘there are two ways of satisfying desires: one is to get more and the other is toant less’. Kahneman et al. (1998) distinguished ‘expected utility’ and ‘experienced utility’. Economists focus on the former,hich is relevant to economic decision-making, and tend to ignore the latter. This focus might be sufficient and justified if

here were not systematic differences between the two. Systematic differences would not arise if aspirations were tethered,r were accurately predicted. However, our evidence suggests that aspirations tend to follow reference income, possiblyith a lag, and that people cannot predict their aspirations well.

The interplay between income and aspirations for income can give rise to a Hedonic Treadmill and provides a possiblexplanation for the Easterlin Paradox. This is what motivated the paper: our objective was to measure the determinants ofspirations for income, in particular the influence of actual income, and to measure the determinants of subjective well-eing, in particular the influence of aspirations for income. The hypothesis was that aspirations need to be understood, as auide to both positive and normative economic analysis. We found that people do indeed adjust their aspirations, and theirubjective well-being, to their reference income. Their reference income includes own income, own income relative to thatf relevant others, and own income in the past.

Tables 2 and 3 indicate that the level of income which people regard as the minimum necessary for their household isot just a matter of objective need but also a reflection of their subjective aspirations. Current income raises our measuref aspiration income – a result which is merely reinforced by our attempt to correct for the possible endogeneity of currentncome. Moreover, insofar as this cross-section data set does not allow us to examine lags in the adjustment of aspirations tohanges in income, our measured coefficient may well understate the full effect of income on aspiration income. Aspirationncome is also greater in response to higher income of reference people and reference times – the former in the sense thataving lower income than others in the village involves greater perceived income need, and the latter in the sense thataving lower income than in the past has the same effect. This evidence of relative deprivation suggests that people frameheir aspirations according to the comparisons that they make.

Aspiration income in turn reduces satisfaction with life and satisfaction with income, although the positive coefficient onctual income is larger (Tables 4 and 5). There is indeed a Hedonic Treadmill but it is only a partial one. Subjective well-beings also directly sensitive to certain variables which represent relative deprivation in relation to other people or other times,
o generating aspirations to match other people or other times. Having a broader reference group, change in income overhe past five years, and perceived rank in the village income distribution all affect happiness.
These results are not weakened by our attempts to establish causal relationships by means of instrumenting actual incomend aspiration income, although the instrumenting cannot be regarded as conclusive. Better data sets, in particular panel


data, are needed in order to verify the causal relationships suggested by our cross-section analysis. We found evidence whichis consistent with own income having a causal effect on aspiration income: a proportionate rise of own income appears toproduce a proportionate rise in aspiration income that is more than half as large (Table 3). Nevertheless, it is possible thatthe association between actual income and aspiration income also contains some reverse causation, i.e. people with higheraspirations apply more effort in order to raise their incomes. This effect might be viewed as making an important contributionto economic development. However, it might instead be viewed as keeping people on the straight and narrow and circular(to use the topology of a treadmill). Given that the Hedonic Treadmill is only a partial one, both views are simultaneouslytenable.

The results of our study for a poor society (rural China) are similar to those obtained by Stutzer for a rich society(Switzerland). However, his estimates suggest a more powerful Hedonic Treadmill effect, with the elasticity of aspirationincome to actual income being more than 0.4 and the negative effect of aspiration income on happiness being 80 per centor more of the positive effect of actual income (Stutzer, 2004; Tables 3 and 5). This difference between the two studiesmight arise because subjective well-being can be expected to depend less on the fulfilment of basic needs and more on thefulfilment of aspirations in a rich society than in a poor one. In China the village is an important social unit as it is normallyresidentially compact and households from outside rarely join the village. Our relative income results, in their effect onaspiration income, might therefore be stronger than in other poor societies.

Easterlin and Sawangfa (2010) provide evidence of recent trends in measured life satisfaction in China, drawn from threeavailable sources: the World Values Survey, the Gallup Survey, and the Asiabarometer Survey. The sources had different unitsof measurement but in each case the average score fell: from 6.83 to 6.76 between 1995 and 2007, from 2.82 to 2.67 between1995 and 2007, and from 3.73 to 3.68 between 2000 and 2007, respectively. Our evidence for rural China of a partial HedonicTreadmill based on own income contributes to the explanation of this declining happiness in China as a whole but cannotfully explain it. Another contribution comes from the evidence, in our data set as in many others, that subjective well-beingis much influenced by relative income. Even rapid income growth need have little effect on perceived satisfaction if relativeincome remains constant. Given that our study highlights the importance of relative income in the village and does notrelate to the population as a whole, this argument helps to explain why subjective well-being in rural China might not haverisen as village incomes rose. More broadly, the socioeconomic changes occurring in the process of China’s rapid economicgrowth, reform and marketisation are likely to have played a part, including rapid urbanisation, changing reference groups,increasing urban economic insecurity, and rising inequality (Knight et al., 2009; Knight and Gunatilaka, 2010a,b).

There is a value judgement, discussed above, that counterfactual subjective welfare – absent aspirations – should guidepolicy. According to a second value judgement, due to Sen, the metric of utility should be entirely cast aside in favour ofpeople’s capabilities to be and to do things of intrinsic worth. Yet another value judgement is that people’s perceptions oftheir own well-being, however formed, should be the criterion. In any case, the existence and extent of adaptive aspirationsis an important issue for both the utility-based approach and the capabilities-based approach to social evaluation. The resultsof this paper – more striking because they relate to a poor society – add another tiny step towards what might eventuallybe a paradigm shift in economists’ conventional notions of economic welfare.

Acknowledgements

We are grateful to the Nuffield Foundation Small Grant Scheme for financial support, and to the editor and three refereesfor helpful comments.

References

Brandt, L., Holz, C., 2006. Spatial price differences in China: estimates and implications. Economic Development and Cultural Change 55 (1), 43–86.Burchardt, T., 2005. The role of income aspirations in individual happiness. Social Indicators Research 74, 57–104.Clark, A., 1999. Are wages habit-forming? Evidence from microdata. Journal of Economic Behavior and Organization 39, 179–200.Clark, A., Frijters, P., Shields, M., 2008. Relative income, happiness and utility: an explanation for the Easterlin paradox and other puzzles. Journal of Economic

Literature 46 (1), 95–144.Colasanto, D., Kapteyn, A., van der Gaag, J., 1984. Two subjective definitions of poverty: results from the Wisconsin Basic Needs study. Journal of Human

Resources 19 (1), 127–138.Duesenberry, J.S., 1949. Income, Saving and the Theory of Consumer Behavior. Harvard University Press, Harvard.Easterlin, R.A., 1974. Does economic growth improve the human lot? Some empirical evidence. In: David, P.A., Reder, M.W. (Eds.), Nations and Households

in Economic Growth. Essays in Honor of Moses Abramovitz. Academic Press, New York, pp. 89–125.Easterlin, R.A., 2001. Income and happiness: towards a unified theory. Economic Journal 111, 465–484.Easterlin, R.A., 2005. A puzzle for adaptive theory. Journal of Economic Behavior and Organization 56, 513–521.Easterlin, R.A., Sawangfa, O., 2010. Happiness and economic growth: does the cross section predict time trends? Evidence from developing countries. In:

Diener, E., Kahneman, D., Helliwell, J. (Eds.), International Differences in Well-being. Oxford University Press, New York.Elster, J., 1982. Sour grapes – utilitarianism and the genesis of wants. In: Sen, A.K., Williams, B. (Eds.), Utilitarianism and Beyond. Cambridge University

Press, Cambridge, pp. 219–238.Ferrer-i-Carbonnel, A., Frijters, P., 2004. How important is methodology for estimates of the determinants of happiness? Economic Journal 114 (497),

640–659.
Frederick, S., Loewenstein, G., 1999. Hedonic adaptation. In: Kahneman, D., Diener, E., Schwarz, N. (Eds.), Well-being: The Foundations of Hedonic Psychology.
Russell Sage Foundation, New York, pp. 302–373.Frey, B.S., Stutzer, A., 2001. Happiness and Economics. Princeton University Press, Princeton.Frey, B.S., Stutzer, A., 2002. What can economists learn from happiness research? Journal of Economic Literature 40, 402–435.Friedman, M., 1957. A Theory of the Consumption Function. Princeton University Press, Princeton.

GGG

HKKKK

KKK

KMP

RRSSS

SSVV


oedhart, T., Halberstadt, V., Kapteyn, A., Van Praag, B., 1977. The poverty line: concept and measurement. Journal of Human Resources 12 (4), 503–520.urr, T.R., 1970. Why Men Rebel. Princeton University Press, Princeton.ustafsson, B., Li, S., Sato, H., 2004. Can a subjective poverty line be applied to China? Assessing poverty among urban residents in 1999. Journal of

International Development 16 (8), 1089–1107.irschman, A., 1973. The changing tolerance for income inequality in the course of economic development. Quarterly Journal of Economics 87 (4), 544–566.ahneman, D., Wakker, P., Sarin, R., 1998. Back to Bentham? Explorations of experienced utility. Quarterly Journal of Economics 112 (2), 375–406.ahneman, D., Snell, J., 1992. Predicting taste change: do people know what they will like? Journal of Behavioral Decision-Making 5, 187–200.apteyn, A., Van Praag, B., 1976. A new approach to the construction of family equivalence scales. European Economic Review 4, 33–72.ingdon, G., Knight, J., 2006. Subjective well-being poverty versus income poverty and capabilities poverty? Journal of Development Studies 42 (7),

1199–1225.night, F., 1922. Ethics and the economic interpretation. Quarterly Journal of Economics 36, 454–481.night, J., Song, L., Gunatilaka, R., 2009. The determinants of subjective well-being in rural China. China Economic Review 20 (December (4)), 635–649.night, J., Gunatilaka, R., 2010a. Great expectations? The subjective well-being of rural–urban migrants in China. World Development 38 (January (1)),

113–124.night, J., Gunatilaka, R., 2010b. The rural–urban divide in China: income but not happiness? Journal of Development Studies 46 (March (3)), 506–534.arx, K., 1849. Wage labour and capital. In: Marx, K., Engel, F. (Eds.), Selected Works. Progress Publishers, Moscow [1969].

radhan, M., Ravallion, M., 2000. Measuring poverty using qualitative perceptions of consumption adequacy. Review of Economics and Statistics 82 (3),462–471.

abin, M., 1998. Psychology and economics. Journal of Economic Literature 36 (March), 11–46.obinson, J., 1956. The Accumulation of Capital. Macmillan, London.en, A., 1985. Commodities and Capabilities. Elsevier Science Publishers, Oxford.en, A., 1987. The Standard of Living. The Tanner Lectures. Cambridge University Press, Cambridge.en, A., 1990. Development as capability expansion. In: Griffin, K., Knight, J. (Eds.), Human Development and the International Development Strategy for

the 1990s. Macmillan, London, pp. 41–58.
mith, A., 1776. An Inquiry into the Nature and Causes of the Wealth of Nations. Clarendon Press, Oxford [1976].tutzer, A., 2004. The role of income aspirations in individual happiness. Journal of Economic Behavior and Organization 54, 84–109.an de Stadt, H., Kapteyn, A., van der Geer, S., 1985. The relativity of utility: evidence from panel data. Review of Economics and Statistics 67 (2), 178–187.an Praag, B., Kapteyn, A., 1973. Further evidence on the individual welfare function of income: an empirical investigation in the Netherlands. European
Economic Review 7, 313–335.

Date post:	21-Oct-2016
Category:	Documents
Upload:	john-knight
View:	216 times
Download:	2 times

Income, aspirations and the Hedonic Treadmill in a poor society

Documents