Professional Documents
Culture Documents
DOI 10.1007/s11205-009-9552-4
Abstract Among the surprising results of research on the relation between a persons
material circumstances and his or her subjective well-being was the finding that this
relationship appears to be rather weak (throughout this paper the terms (general) life
satisfaction, (subjective) satisfaction, happiness and subjective well-being will be
used interchangeably. The same applies to the terms material circumstances, material
conditions, material situation and material well-being). However, more recently
authors began to ask the question, whether this might at least in part be explained by the
insufficiencies of income as an indicator for the material situation. Building on this idea,
they have shown that the inclusion of alternative measures for the respondents material
situationsuch as wealth measures in particularreveals that the relationship between a
persons material well-being and his or her subjective well-being might just be somewhat
stronger than researchers thought before. The paper will follow this lead but will go beyond
current research by first, systematically reviewing the various approaches available for
measuring the material situation and second, by proposing the use of a so-called deprivation index, an alternative measure of material well-being, which is frequently used in the
context of poverty research (compare e.g. Townsend in Poverty in the United Kingdom,
Penguin Books, Harmondsworth, 1979; Hallerod in J Eur Soc Policy 5:111129, 1995;
Nolan and Whelan in J Eur Soc Policy 6:225240, 1996). It will be argued, that such a
deprivation based measure will perform better than indicators like income or wealth when
analyzing the relationship between material conditions and subjective well-being. This
hypothesis will be tested using three different German datasets. Based on this data it will
be shown that in all cases deprivation measures perform better in explaining differences in
subjective well-being than the alternatives. However, both types of measures seem to
capture slightly different aspects of the material situation, a result which has also been
found in the poverty literature cited above. Thus using a combination of both seems to be
the best alternative.
Keywords
B. Christoph (&)
Institute for Employment Research (IAB), Nuremberg, Germany
e-mail: bernhard.christoph@iab.de
123
476
B. Christoph
123
477
will, on the other hand, have a much smaller impact, which is why starting at a certain
threshold, the positive effect of having additional material resources should decrease.
The core idea of the relative standards model is that people might apply various
standards in order to judge their current situation, such as the circumstances they lived in
before or the current situation of relevant others. This is to say that it is not a currently high
level of income per se, which will make people more satisfied but what is important is the
difference of current income levels to various standards that might be applied. Less formally stated: what makes people really happy is not meeting an absolute standard, but
rather that their current income is higher than it was some time before or that they are
doing better than their neighbors or colleagues do. A good example for this kind of
theoretical approach is provided by Easterlin (2001, 2002), who tries to explain the finding,
that a persons happiness (on average) will remain largely unaltered over the life cycle,
even though his or her income will increase during the same period. The reason for this is,
according to Easterlin, that this persons aspirations will equally increase over time, thus
compensating for the material gains. Another example which includes quite a bunch of
different comparison processes is Michalos (1985) Multiple Discrepancies Theory.
Michalos argues that net satisfaction is the result of adding up several discrepancies
experienced as a result of multiple comparison processes, which have before been
described only separately by different theoretical approaches. The main dimensions are
given by contrasting the actual situation, i.e. what one actually has, with (a) ones personal
aspirations, (b) the situation of relevant others, (c) what one deserves, (d) ones basic needs
(a theoretical strain that is not to far from Veenhovens livability theory discussed above),
and (e) intertemporal comparisons as contrasting ones current situation with the past, with
ones past expectations about today or the expectations one has for the future.
A core aspect of the relative standards approach is that personal standards might be
changed as ones personal situation improves or deteriorates, i.e. processes of adaptation
are taking place. Thus, increases in the material situation might be answered by raised
standards instead of growing satisfaction. This might be one theoretical explanation why
the correlation between income and well-being is actually so small.
The cultural approach argues that people are socialized to follow (society specific)
social norms and that displaying behaviors that are in accordance with these norms will
increase subjective well-being (cf. Markus and Kitayama 1994; Oishi 2000). In modern
industrial societies, being in paid employment is one of the most important behaviors one
might participate in and therefore having a paid job will increase well-being. Thus
according to this theoretical approach, having more income appears to be merely a byproduct of following this generally accepted cultural norm. This also implies that people
with a comparably low income might still display high subjective well-being, since despite
the low financial returns they still fulfill the norm by being involved in a productive
activity (compare Diener and Biswas-Diener 2002: 151). If this is true and poorer people
show well-being levels that are higher than one would expect given their income, this
could explain why the correlation between income and well-being is rather low.
Set-point theoriy (e.g. Headey and Wearing 1989) in general will not assume a link
between income and subjective well-being. As a specific type of trait theories (for a critical
overview of trait theories compare e.g. Veenhoven 1994), just as the former they are based
on the core idea that happiness is a fixed personality trait, which means it might not easily
be altered by external conditions. Nevertheless, these theories allow actual satisfaction
levels to vary, e.g. due to the influence of specific life-events. Yet ultimately, satisfaction is
assumed to return to its original, person-specific state, the so-called set-point.
123
478
B. Christoph
2 Material Situation and Happiness: The Use of Alternative Methods and Indicators
It has been only recently, that researchers have argued that the link between the material
situation and subjective well-being might in fact not be as weak as was previously thought.
In essence, there have been two general approaches to make this point: the first is calculating more elaborate models and the second is to apply more appropriate measures for
the material situation.
Researchers trying to apply more elaborate models for most of the part tried to tackle
different types of heterogeneity. One such example is provided by Frijters and his
co-authors (Frijters et al. 2004), who try to control for heterogeneity of individual characteristics. In their paper they examine the development of life-satisfaction
among East Germans during the decade following German Reunification (19912001).
Applying a fixed-effects ordered logit model for ordinal dependent variables developed
1
This appears to be a problem since it is just the fraction of happiness which is not influenced by
personality but by external factors, which is of interest here. This is not to say that personal characteristics
are unimportant for a persons happiness. Quite to the contrary, they should be considered a central
influence-factor (for an overview compare e.g. Diener and Lucas 1999). However, using well-being as an
output indicator in analyses on social policies or as proxy for preferences in economic analyses would be
inadequate, should it entirely be determined by personality.
123
479
by Ferrer-i-Carbonell and Frijters (2004), they are able to show that particularly in the
years immediately following unification (i.e. until about the mid 1990s) there is a much
stronger influence of income on life satisfaction than is usually found.
Other researchers put their focus not on individual heterogeneity, but rather try to
model differences in response behavior emerging between different groups of respondents or at different stages of the satisfaction scale. An example for this is provided by
Clark et al. (2005). Using a latent class model they can demonstrate that the relation
between well-being and income is different for the four latent groups they identified.2 A
somewhat related idea can be found in two papers by Boes and Winkelmann (2006,
2009). They are applying a particular type of ordered probit model (a so-called generalized ordered probit model), which allows the effect of income on satisfaction to vary
for each stage of the dependent variable.3 Using this modeling approach they are able to
show that increases in income have stronger effects among respondents that have low
satisfaction scores, i.e. who are rather dissatisfied, than among those whose satisfaction
scores are high.
Another topic that has been approached using more elaborate models is the interrelations between the individual and the national level. This was for example an issue in
the above cited paper by Clark et al. (2005), who showed that the latent classes of
respondents identified by their analysis did also differ with regard to their national
composition. A more complex approach to deal with this problem has been provided by
Schyns (2002, 2003). She applied a multilevel model to simultaneously control differences at the national (GDP) and the individual level. Doing so, she was able to show that
there is an interaction between these two levels and that the relationship between income
and satisfaction is stronger in poor countries than it is in affluent ones, which is mainly
due to the fact that poorer people in less affluent nations score particularly low on wellbeing.
Research on providing better measures for the material situation is also relatively
scarce. So it was not without reason that Diener and Biswas-Diener as late as in 2002
called for improving research on better measures for the material situation, which go
beyond mere indicators of income (Diener and Biswas-Diener 2002: 355). An early
attempt to improve on material measures was brought forward by Mullis (1992). In fact
his approach incorporated many new ideas, which have been developed further in recent
years. In his paper, Mullis proposed a composite index for which he used permanent
income, operationlized as average income over several years, and annualized net worth,
which set the balance of a persons savings and debts in proportion to his or her
remaining life expectancy. This material measure was divided by a normatively fixed
basic income level, thus resulting in a relative measure for economic well-being. Mullis
could show, that this index was a better predictor of subjective well-being than traditional measures.
One of the measures that had already been introduced by Mullis but has been used more
frequently in recent years is wealth. The most notable example for this is probably the
2
Due to the lacking information on general life satisfaction in the data used by Clark et al. they can show
this only for the relation between income satisfaction and actual income.
This means that there is not only a different intercept for each stage of the satisfaction-variable, while
effects of independent variables are assumed to be the same at all stagesas is the case in standard ordered
probit or logit modelsbut coefficients for the independent variables might vary, too. In contrast to a
multinomial model, which would also allow calculating different coefficients for the independent variables
at different stages of the dependent variable, the generalized ordered probit model will consider the ordering
information of the dependent variable when calculating the estimates.
123
480
B. Christoph
work of Headey and his collaborators (Headey and Wooden 2004; Headey et al. 2005,
2008). They explicitly argued that income is not the only or necessarily the best indicator
of material standard of living (Headey et al. 2008: 66). In order to provide a better
alternative they supplemented income by measures of wealth and by basic consumption
indicators. Doing so, they were able to show that the overall influence of a persons
material situation on his or her subjective well-being is considerably larger than it was
previously assumed.
A further example for the use of wealth indicators is provided by DAmbrosio et al.
(2009). In addition to using a detailed account on wealth they also improved on another
idea brought forward by Mullis, which is using a measure for long-term income (calculated
as average over several years) instead of only relying on current values. They showed that
both, the use of long term income as well as additionally controlling for wealth brought a
significant improvement when analyzing subjective well-being.
Another more complex indicator for material well-being, which is commonly applied in
poverty research and might be useful, when analyzing the relation between subjective wellbeing and the material situation is a so-called deprivation index.4 A deprivation index
usually consists of an extended list of items a household may or may not posses, which are
combined into a single measure.5 In the remainder of the paper it will be argued, that it
indeed is a useful tool for analyzing subjective well-being and that applying such an index
will be an improvement compared to using indicators like income or wealth.
Up until now such an instrument has not been commonly used in satisfaction research.
There are some examples where single deprivation items or even a combined index have
been used in descriptive analyses of subjective well-being (e.g. Bohnke and Delhey 1999a;
Bohnke 2005; Mller 2007). There are also examples where smaller lists of items have
been used as proxies for wealth (Howell et al. 2005).6 In fact there is one example (Bohnke
and Delhey 1999b), where an indicator for deprivation based poverty and income are both
included in a multivariate model explaining satisfaction with the standard of living as well
as satisfaction with some other life domains. However, in this article the deprivation and
the income measure are merely used as control variables by which the authors try to
explain differences in satisfaction between East and West Germany. The differences
between the two predictor variables were not an issue discussed by Bohnke and Delhey.
Hence, to the knowledge of the author, this is the first time that a systematic comparison of
a detailed deprivation index and income as predictors of subjective well-being is made.
4
It should be noted that the term deprivation in this context does not refer to a subjective concept as it e.g.
does in relative deprivation (Stouffer 1949). Since it was designed as a poverty measure it explicitly aims at
identifying those who actually are poor and not those who feel that way (in comparison to some reference
group) but probably are not (compare e.g. Hallerod 1995: 115). Instead, the term deprivation refers to living
in material circumstances which are lower than a commonly shared and/or socially agreed upon standard,
i.e. a person is considered to be deprived if he or she does lack a certain amount of items, which are regarded
to be part of this standard. Usually the items that actually make up the standard are identified by either
asking the survey-respondents, which among the items presented to them should be considered a necessity
or by identifying those items, which are held by a majority.
For a more detailed description compare the paragraph on measures for the material situation below.
Howell and his collaborators examine the relationship between subjective well-being and material living
conditions for an indigenous ethnic minority in Malaysia. Since many respondents in this population group
have no or only irregular income, collecting adequate income information was not possible. Instead, they
presented heads of households with a list of 13 consumer items, asking which items their households
possessed. However, instead of generating a deprivation index based on these items, they were used as
proxies for monetary wealth by substituting market prices for used goods for each item possessed and
combining the resulting sum with information on savings into a single wealth indicator.
123
481
Measures of expenditure and consumption are related but they are not the same. The main difference is
that measures of expenditure only focus on actual spending, while consumption measures will also account
for the consumption of durables (like owned housing or cars). They do so by adding an appropriate amount
for their usage/consumption to actual spending (while not fully considering such one-time investments in
case they should fall in the reporting period). Moreover, certain kinds of expenditures like e.g. cash
payments to family members are not considered (for a more detailed description compare for example
Meyer and Sullivan 2003, pp. 1188 et seq.). In addition, consumption might also include goods which are
not bought, e.g. benefits in kind (Headey 2008), foodstuffs from subsistence farming or other home-made
products (although the latter should not be too common in developed countries).
123
482
B. Christoph
exceeds the one found for the consumption variables among respondents with low consumption levels (Meyer and Sullivan 2003). The downside of consumption measures is that
they are complex and time consuming to administer in a survey, since they either require
respondents to keep a book of household accounts for a certain period of time8 or to report
(and remember) the amount of various items on a very extensive list they bought during
this term.
The deprivation approach, on the other hand, is much easier to administer than
expenditure data, but is still somewhat time consuming, when compared to the survey time
required for obtaining income data. Originally developed by Townsend (1979) for his
landmark study on poverty in the United Kingdom, this approach essentially consists in
checking for a selected list of goods and activitieswhich are considered central for an
appropriate standard of livingwhether respondents own these items or participate in
these activities, respectively.
Since the original introduction of the approach, quite a number of problems linked to its
practical implementation have been tackled. The first is the question of consumer preferences: the non-availability of an item or the non-participation in an activity should not
necessarily be considered an indication of deprivation but might merely result from a
particular taste or distaste of the respondent. Should the latter be the case, it surely would
be problematic to consider the fact that this item is lacking to be a sign of material
hardship. For this reason, nowadays items are only considered to indicate deprivation if
they are lacking for financial reasons.
A second problem is how to combine the items on the list to a common index. The most
basic approach is to calculate an unweighted index by simply adding up all items that are
lacking (for financial reasons). However, when looking at a typical list of items used to
generate a deprivation index (compare e.g. Table 7 in the Appendix) it is easy to see that
some itemslike for example a home without damp wallsshould be considered more
important for an appropriate standard of living than otherssay like owning a DVDPlayer. For this reason, weighting the items used to build the index seems appropriate.9 In
fact, there are several approaches to achieve this. The three central ones are probably: first,
weighting an item by the percentage of respondents considering it to be important for an
acceptable living standarda procedure which requires to obtain additional information on
this importance during the survey (as done e.g. by Hallerod 1994, 1995, when constructing
his so-called Proportional Deprivation Index). The second approach is to weight each item
by the percentage of the population that actually owns this item. This latter approach draws
on relative deprivations idea, that the lack of a particular item might be considered much
more problematic if it is something which everybody owns and rightfully should own than
if this item is some kind of rarely owned luxury good (compare Desai and Shah 1988). An
advantage of the latter approach is that it might be applied without administering any
additional questions. The last one is to select only such items to be part of the index, which
are considered to be highly relevant. This might be achieved for example by selecting only
those items that are owned by at least 50% of the population (as done e.g. by Mack and
Lansley 1985). Another method which has been applied by Callan et al. (1993) and Nolan
8
This period is often 1 or 2 weeks. It might, however, be considerably longer. In the German income and
expenditure survey (EVS) for example, respondents have to keep a book of household accounts for 3 months
in total and at least part of them has to keep a detailed list of all expenditure on food, drinks and tobacco for
1 month.
Even though systematic comparisons between different weighting approaches have shown, that applying
either of these or none at all will only generate minor differences between the resulting indices (Lipsmeier
1999), weighting an index still seems reasonable for conceptual reasons.
123
483
and Whelan (1996) is to use factor analysis to identify the basic items.10 While building the
index on highly relevant items only might not be considered a weighting scheme in the
narrow sense of the word, this procedure should guarantee that only items that are strongly
related to deprivation are taken into account, while others are discarded.
A third problem which has been somewhat more difficult to solve is how to define a
threshold for the number of items lacking that is considered necessary, to hold a person to
be poor. A useful differentiation of criteria to define such a poverty line is provided by
Andress and Lipsmeier (2001) who distinguish between statistical and non-statistical criteria. Among the statistical criteria is defining a certain percentage of the mean or median
value to draw a line between the poor and the remainder of the population, just the way it is
done when defining income-poverty. The problem with this is that deprivation indices have
a strongly skewed distribution, often with a modal value of zero, i.e. a majority of
respondents does not lack any of the items on the list. Therefore the core argument of the
aforementioned definition as it is usually applied in the case of incomethat the mean or
median represents an average and thus a normal income levelseems not to be applicable in case of a deprivation index. A second option is to look for the number of items by
which a certain percentage of the population will be classified as poor (e.g. Hallerod 1995).
This means for example that we will use a poverty threshold of three if the least well off
10% of the population lack three items or more. The obvious weakness of this approach is
that doing so would be somehow begging the question, since the matter of interesti.e. the
proportion of poor personswould thus be answered by definition instead of being
identified empirically.11 The major non-statistical criterion is to define a particular number
of items, the lack of which is considered to classify a person as poor either by means of an
outside criterion or simply by setting a particular cut-off point (e.g. Mack and Lansley
1985). In any case, properly defining deprivation based poverty still remains somewhat of a
problem, which is not solved satisfactorily by any of the approaches described.12
3.2 Income, Consumption or Deprivation: Advantages and Drawbacks
It has sometimes been argued, that deprivation measures (among others) should be best
viewed as attempts to make do in the absence of valid consumption measures (Headey
2008: 26). However, this position seems to be less convincing when considering an
argument made by Andress and Lipsmeier (2001) and Andress et al. (2004) who hold that
direct and indirect measures for the material situation should be understood against the
background of a simple three stage action-model, in which people (a) hold certain
resources, (b) use these resources in accordance with their preferences in such a way as to
(c) obtain certain, desired results (compare Fig. 1).
Keeping this model in mind it becomes clear, that resource, expenditure/consumption
and deprivation indicators try to measure a persons material situation at the three different
stages of this process: indicators like income or wealth try to measure the resources
available for achieving a proper standard of living, consumption indicators measure the
10
For details of the procedure compare Callan et al. (1993): 150 et seq. and Nolan and Whelan (1996): 228
et seq.
11
Even though in our example it is somewhat unlikely that a threshold of three items will identify exactly
ten percent of the population to be poor, the actual percentage value should be close.
12
For the current paper, the fact that defining a poverty line based on deprivation measures is a somewhat
difficult task seems to be of minor importance, since it merely focuses on the relationship between wellbeing and measures for the material situation but does not explicitly focus on the relation between wellbeing and poverty.
123
484
B. Christoph
Fig. 1 Overview of different approaches to measuring the material situation. Adapted from: Andress and
Lipsmeier (2001)
actions taken to attain this goal and deprivation indicators try to measure the material
situation on the level of the results obtained.
Nevertheless, consumption and deprivation indicators have a lot in common, and these
shared characteristics are in fact what would make one assume a much closer relation to
subjective well-being than is found for income or other indirect indicators. The first shared
advantage is a theoretical one which is based on need theory. At least if this theory should
hold, it should not be the theoretical potential to have ones needs fulfilled (as it is
measured by indirect indicators), which should make one happy, but much more the actual
fulfillment of those needs (which is the focus of direct indicators).
A second shared advantage is that as direct measures for a persons material circumstances, both indicators are much less vulnerable to short-term fluctuations than income is.
In many cases, these fluctuations do not happen accidentally but are anticipated, e.g.
among self-employed persons whose income is subject to a certain amount of variation. As
long as the persons concerned take proper measures to compensate for this kind of variation in such a way as to keep their actual standard of living constant, a measure doing
likewise should much better grasp the connection between their well-being and their
material situation.
Moreover, it has been shown for both indicators that they capture a somewhat different
aspect of the material situation than income does. The most obvious indication of this is
that the groups of persons which are considered to be poor based on income and consumption or income and deprivation measures, respectively are quite distinct (compare e.g.
Brewer et al. 2006 or Noll and Weick 2007 for consumption and Hallerod 1995 or Bohnke
and Delhey 1999a for deprivation).13 This should not be considered an advantage per se,
but chances are that the aspects of the material situation captured by those indicators are
also the ones that are more strongly connected to well-being. So in sum, from a conceptual
point of view we could expect both types of direct indicators to be much more strongly
related to subjective well-being than indirect indicators like income and wealth are.
13
This is why poverty researchers often argue to combine direct and indirect measures to identify the truly
poor, as e.g. Hallerod (1995) calls the group of persons that is deprived according to both measures.
123
485
In contrast to that, to make a theory-based decision between consumption and deprivation indicators seems to be a much more difficult task. As has been shown, both types of
indicators share quite a lot of common advantages that seem to make both of them a good
choice for analyzing the material situation/subjective well-being relation. One difference
between both is that deprivation indicators are usually constructed to capture the availability or non-availability of quite basic items. They usually draw a distinction between
those with a (more or less strongly) restricted standard of living and all the other
respondentsand do so in quite some detail. The downside of this approach is that unlike
consumption indicators they are unable to distinguish between the well off and those that
do even better, because basically all of them can afford all the items on the list.14 This is
why deprivation indices are usually strongly skewed, with sometimes more than 50% of
respondents lacking no item at all. However, what might be a disadvantage for answering
some questions could in fact be useful for answering the current one: one of the most stable
results of previous research is that it is particularly the less well off, for whom material
improvements lead to increased subjective well-being. Thus an indicator focusing more
strongly on the lower end of the material spectrum might in fact do a much better job. Be
that as it may, since in Germany consumption measures are largely unavailable outside
official statistics15 and official data usually do not include subjective measures like wellbeing indicators, at this point the only comparison which can be made is that between
indirect measures (income and wealth) and deprivation indicators.16
The fact that a deprivation indicator is particularly apt to identify people at the lower end of the material
distribution but does not so good a job in distinguishing differences among the better-off has been made e.g.
by Andress et al. (2001).
15
The most notable official data being the EVS described above.
16
As mentioned above, Headey et al. (2005, 2008) have also examined the influence of consumption
indicators using the British household panel survey (BHPS) and the Hungarian Tarki Panel. At least in the
case of general well-being, results were mixed, showing a strong influence for the Hungarian data but no
significant effects for the BHPS. To what extend these results might be due to the somewhat reduced set of
consumption questions one is restricted to in a multi-topic household survey, i.e. whether more detailed data
would show a stronger effect, might be an interesting question for further research. In any case, since to the
knowledge of the author even a reduced set of consumption data is unavailable in any German dataset
covering subjective well-being as well as deprivation items, an empirical comparison for consumption and
deprivation indicators and their influence on subjective well-being is not possible using German data.
123
486
B. Christoph
record of personal wealth and debts, which is not available in the two other datasets. The
downside of the GSOEP is, that it includes only a very restricted set of deprivation items
and moreover, that it does not include any information on the differences in importance,
respondents might assume the items on the list to have (and which could be used as
weighting factors when computing the index).17
The second dataset is the German Welfare Survey 1998 (WS98, compare Thum et al.
1999). At first, the most obvious downside of this dataset seems to be that it is somewhat
aged. However, since we assume the correlation between subjective well-being and
material living conditions not to vary significantly over time, the age of the data should
constitute no problem for the validity of the results presented below. A somewhat more
serious shortcoming of the data is the almost complete lack of information on wealth. An
advantage compared to the GSOEP is the very detailed list of deprivation items and the
availability of information on the relative importance of these items.
The third dataset to be used is the second wave of the Panel Study Labour Market and
Social Security (PASS, compare Trappmann et al. 2009).18 Like the Welfare Survey this
dataset includes a very detailed list of deprivation items and also the corresponding
importance information. In addition, there is at least summary information on wealth,
which is, however, far from reaching the degree of detail the corresponding information in
the GSOEP provides. The downside of this dataset is the comparably low response rate,
particularly in the general population sample. In this sample the response rate on the
household level for the first wave was merely 26.6% in the general population sample and
35.1% in the sample of benefit recipients.19
Summing things up, each of the datasets used has certain advantages and disadvantages.
Table 1 gives a summary of these. So not all questions of interest might be answered by all
of the datasets and results obtained with only one of these datasets might be considered
17
For more details on this problem compare the theoretical discussion of deprivation indices above and/or
the description of index construction below.
18
The primary goal of PASS is to provide information on recipients of the reformed unemployment
assistance scheme in Germany, the so-called Arbeitslosengeld II (Unemployment Benefit II). In order to
achieve this goal, the study comprises two different samples, a sample consisting only of households,
receiving this benefit at the date the sample was drawn and a sample of the general population, which might
also include benefit recipients, but only to the degree usually found in the population. As of January 2009
this was 10.1% of all applicable persons, i.e. those between 0 and 65 years of age (STBA 2009: 12. This
value equates to 8.1% of the entire population. Note, however, that calculating such a percentage for the
entire population might be somewhat problematic, since this figure does not represent all recipients of means
tested benefits. This is so since people of age 65 and older would have to apply for a different type of benefit,
if they are in need. This benefit will result in quite comparable financial circumstances of the needy person,
but is governed by a different part of the German Social Code, namely social assistance legislation/SGB
XII). For details on the sampling design compare Rudolph and Trappmann (2007). When calculating models
that should provide representative information on the entire population, appropriate weighting factors are
available, which allow to adequately combine the two subsamples.
19
This results in an overall response-rate of 30.5%. Response rate within households varied between 84.3
and 85.6% resulting in person-level response rates of 30.0% for the recipient sample, 22.4% for the
population sample and 25.9% for the combined samples (for wave 1 figures compare Christoph et al. 2008).
Overall re-response rates for wave two were 62.4% on the household and 53.1% on the person level
(compare Bungeler et al. 2009). The welfare survey, which is a cross-sectional survey, had a response rate of
56.1%. Due to the complex sample structure and up to now more than 25 panel waves, it is not possible to
present detailed figures for the GSOEP here. Original response rates on the household level are between
60.6% in sample A (1984) and 40.2% in sample H (2006; compare Haisken-DeNew and Frick 2005; von
Rosenbladt et al. 2007). For an overview of attrition between panel waves in the GSOEP compare Kroh and
Spiess (2008).
123
487
Year
Welfare Survey
GSOEP
PASS
1998
2007/2008 (wave 2)
11,424
9,408
Positive
Detailed deprivation
information
Detailed deprivation
information
Negative
somewhat weak. If, however, all datasets show comparable results, it seems to be most
likely that these findings are reliable despite the weaknesses of any single dataset.
4.2 Measures and Harmonization
As indicator for subjective well-being, all three surveys used a single measure with
comparable wording. The question-stimulus was How satisfied are you currently with
your life as a whole in general? and responses could be given on an 11-point scale ranging
from 0 (completely dissatisfied) to 10 (completely satisfied).
For income, equivalized household income was used. Equivalence weights were applied
according to the modified OECD-Scale (Hagenaars et al. 1994), giving the first person
older than 14 years of age a weight of 1, additional persons over 14 received a weight of
0.5 and persons aged 14 or younger received a weight of 0.3. Since previous research has
shown that the relationship between subjective well-being and income is non-linear and it
has been exemplified that these non-linearities might be properly covered by logging (e.g.
Duncan 1975; Easterlin 2002), logged income has been used. For the Welfare Survey,
income information was still in German Marks (DM). In order to allow for more easy
comparisons of descriptive statistics this was converted to Euro using the official Exchange
rate (1 = 1.95583 DM). After that, income was adjusted to 2007 values using the
consumer price index (StaBA 2009).20
Wealth has been operationalized quite differently in the three datasets. As has been
mentioned before, there was no information on wealth available in the Welfare Survey. For
the other two surveys, separate variables were used for savings and debts. The main reason
for this was that separate variables allowed for logging,21 which would be difficult for an
integrated variable, due to the negative values resulting in cases where debts exceeded
savings. The PASS included two items asking respondents for categorized information on
the households savings and debts, the values of which were replaced by the theoretical
category means.22 The 2007 wave of the GSOEP was the first replication of a very detailed
wealth module that had been originally developed for the 2002 wave. This module
included information on each respondents possessions as well as on debts for the following domains: real estate (incl. loans), investments (e.g. stocks) and savings, insurances
20
The consumer price index information standardized to 2005 values in StaBA (2009) was re-standardized
to 1998 values. Prices in the year 2007, which was used as the reference year for price adjustment, were
equal to 114.3% of 1998 prices.
21
The actual values logged were savings or debts plus one, in order to be able to calculate a logged value
for respondents with debts or savings of zero.
22
For the highest category, indicating 50 000 Euro or more, a value of 75,000 Euro was imputed.
123
488
B. Christoph
Table 2 Descriptive statistics for subjective well-being, income, wealth and deprivation
Mean
Min
Max
7.6
1,346
2.2
11.2
0
78
10
2,130
21,428
2,130
19.4 (19.4/22)
2,130
100.0
2,130
10
GSOEP 2007
Satisfaction w. life
6.7
1,452
11,424
156
23,333
11,424
9.2
1,071.5
11,424
1.5
151.9
11,424
1.2
9.7 (13.7/15)
11,424
8.6
71.0
11,424
Satisfaction w. life
7.1
10
9,408
1,561
66,667
9,408
1.5
7.5
9,408
0.6
7.5
9,408
2.0
19.9 (23.8/26)
9,408
8.4
83.4
9,408
PASS 2007/2008
(e.g. life or private pension insurance), business property, private possessions (e.g. jewelry,
art etc.) and consumer credits.
The integration of the several variables in each of these modules was done following the
general procedures described by Schafer and Schupp (2006) for the 2002 wave. However,
in contrast to the former, no integrated wealth variable was generated, but wealth and debts
(where applicable) from all the components were added up separately. As a result of this
much more detailed account, respondents in the GSOEP have much higher values in the
wealth variables than their PASS counterparts, with maximum values of both, debts and
wealth, exceeding a million Euros (compare descriptive statistics in Table 2). A significant
part of this is made up by business property, possessions, or real estate as well as (in case of
debts) housing loans, which are all not accounted for in PASS. For both datasets the debts
and wealth indicator on the household level was transformed into available wealth per
capita before being logged, which means that for the GSOEP the personal information was
aggregated to the household level and for both datasets, household level information was
divided by the number of household members.
Information on deprivation varied strongly between the three surveys. The general
strategy applied here was to take whatever was available in the respective survey.
Therefore the relative effect of the deprivation index is difficult to compare between
surveys, even if problems of scale are accounted for (compare below). This, however,
seems not to be too much of a drawback, since the difference that is really of importance in
the current context is that between income and the deprivation index within each survey.
The Welfare Survey included a list of 22 items23 that might be lacking due to financial or
23
For a detailed list of items available in each survey compare Table 7 in the appendix.
123
489
123
490
B. Christoph
Table 3 Basic ordered logit models predicting subjective well-being (0 = completely dissatisfied;
10 = completely satisfied)
WS 1998
M1 (Inc)
Deprivation index
Ln HH-Inc. (n. OECD)
GSOEP 2007
M2 (DI)
M1 (Inc)
-0.058***
1.363***
PASS 2007/2008
M2 (DI)
M1 (Inc)
-0.049***
0.960***
M2 (DI)
-0.085***
1.036***
LL (intercept)
3,794.250
-3,794.250
-22,195.709
-22,195.709
-17,953.866
-17,953.866
LL (model)
-3,671.930
-3,563.522
-21,797.247
-21,580.564
-17,447.159
-16,742.129
AIC
3.458
3.356
3.818
3.780
3.711
3.561
0.032
0.061
0.018
0.028
0.028
0.068
2,130
2,130
11,424
11,424
9,408
9,408
higher level of satisfaction as will lacking less items contained in the deprivation index.
A second consistent finding is that in all datasets the relation between the deprivation
index and well-being is considerably stronger than that between income and the latter,
which can be seen from the Pseudo R2 and AIC values of the models. The question is,
whether this result will hold, when additional variables are controlled for and whether
both variables, income and deprivation, will contribute a substantial share to improving
the model independent from one another. We will first try to answer this question using
the Welfare Survey data.
In Table 4, three different models are compared, the first model including only income
and the control variables, the second the deprivation index and the third takes account of
both these variables. As can be seen from the Pseudo R2 values, the difference between the
models one and two might be somewhat reduced by adding the controls, but it still remains
strong.
Looking at Model 3 we find that combining both indicators in a common model will
improve results only to a minor extent. Nevertheless, income still displays a separate effect
when deprivation is controlled for. However, looking at the standardized coefficients
displayed in the second column of the model table it is obvious, that the contribution made
by income is the smaller of the two.
Considering the results for the control variables, they show by and large the figures that
were to be expected. Looking at regional differences, East Germans are significantly less
satisfied than their West German counterparts. A significant gender effect can only be
observed in Models 2 and 3, but age in all models displays the strong curvilinear effect that
could be anticipated based on earlier research. While besides a small effect in Model 1
there is no effect for German citizenship and also no significant education effect, there is a
very strong positive effect of being married, while being unemployed and having serious
health restrictions both have negative effects.
Results for the GSOEP are displayed in Table 5. In contrast to the data discussed up to
now, the GSOEP also includes wealth information, which will be controlled for in the
models. Since in terms of the approaches for measuring the material situation discussed
above, income and wealth both are considered indirect measures, while deprivation is a
direct measure for the material situation, the former two are grouped together in the models
discussed below and the combined effect of both is compared to that of the deprivation
index.
123
491
M2
XY-std.
Coeff.
Deprivation index
Ln HH-income (new OECD)
1.136***
0.271
-0.337***
-0.080
M3
Coeff.
XY-std.
Coeff.
Coeff.
XY-std.
Coeff.
-0.051***
-0.381
-0.045***
-0.335
0.522***
0.119
-0.376***
-0.085
-0.471***
-0.107
0.440*
0.042
0.354
0.033
0.294
0.027
0.149
0.037
0.187*
0.045
0.195*
0.046
-0.079***
-0.628
-0.075***
-0.570
-0.087***
-0.654
0.001***
0.618
0.001***
0.551
0.001***
0.626
0.825***
0.193
0.626***
0.141
0.645***
0.144
0.122
0.030
0.105
0.025
0.048
0.011
-0.046
-0.008
0.081
0.014
-0.082
-0.014
-0.885***
-0.096
Unemployed
-1.096***
-0.148
LL (intercept)
-3,794.250
High
Health (Ref.: no disability)
Disability
-0.895***
-0.093
-0.885***
-0.091
-0.828***
-0.108
-0.714***
-0.092
-3,794.250
LL (model)
-3,577.344
-3,489.086
-3,477.235
433.811***
610.328***
634.030***
AIC
3.379
3.296
3.286
0.057
0.080
0.084
2,130
2,130
2,130
200.218***
23.702***
Results in Table 5 mainly provide two additional pieces of information: first, the two
wealth variables (savings and debts) both have a strong significant effect on satisfaction,
which also prevails after income and deprivation have been controlled for. The second
additional result is that, even though the difference is considerably smaller than the one
observed in Table 3, the deprivation only model (Model 2) has still more explanatory
power than Model 1, including both income and the two wealth variables. The effects for
the control variables remain largely unaltered, the major difference being that tertiary
education makes some difference in Model 2 and that there is a generally stronger gender
effect.
123
492
B. Christoph
M2
XY-std.
Coeff.
Deprivation index
M3
Coeff.
XY-std.
Coeff.
Coeff.
XY-std.
Coeff.
-0.045***
-0.283
-0.034***
-0.214
0.665***
0.168
0.382***
0.095
Ln savings
0.062***
0.135
0.038***
0.081
-0.051***
-0.110
-0.038***
-0.080
-0.395***
-0.081
-0.523***
-0.106
-0.431***
-0.086
-0.057
-0.008
-0.066
-0.009
-0.146
-0.019
Ln debts
Region (Ref.: West Germany)
East Germany
Nationality (Ref.: other)
German
Gender (Ref.: male)
Female
Age
Age (squared)
0.184**
0.047
0.180**
0.045
0.185**
0.046
-0.068***
-0.626
-0.057***
-0.520
-0.063***
-0.567
0.001***
0.534
0.000***
0.425
0.000***
0.453
0.288***
0.073
0.172**
0.043
0.196**
0.049
-0.077
-0.019
0.060
0.012
-0.060
0.203*
-0.015
-0.128
-0.031
0.040
0.033
0.006
-0.838***
-0.156
-0.796***
-0.147
-0.796***
-0.146
-0.675***
-0.101
-0.657***
-0.098
-0.498***
-0.073
-22,195.709
-22,195.709
-22,195.709
LL (Model)
-21,226.749
-21,132.355
-21,036.626
1,937.921***
2,126.708***
2,318.166***
AIC
3.720
3.703
3.687
0.044
0.048
0.052
11,424
11,424
11,424
380.246***
191.458***
These findings are confirmed by the results from the PASS data in Table 6, which
are pretty much the same. In this case the difference between Model 1, including the
indirect indicators and Model 2 containing the deprivation index is more pronounced,
which might be a result of the somewhat more detailed deprivation information
available in PASS. In any case, the general tendency of deprivation to be somewhat
more strongly related to satisfaction than is income alone and even than the indirect
indicators are taken together, remains unaltered. Nevertheless, just as it appeared when
using these instruments for measuring poverty, direct and indirect measures seem to tab
123
493
M2
XY-std.
Coeff.
Deprivation index
M3
Coeff.
XY-std.
Coeff.
Coeff.
XY-std.
Coeff.
-0.078***
-0.429
-0.063***
-0.346
0.570***
0.166
0.339***
0.094
Ln savings
0.126***
0.216
0.051**
0.083
-0.043***
-0.079
-0.026**
-0.046
-0.282***
-0.056
-0.352***
-0.067
-0.270**
-0.051
-0.042
-0.006
-0.051
-0.007
-0.155
-0.022
Ln debts
Region (Ref.: West Germany)
East Germany
Nationality (Ref.: other)
German
Gender (Ref.: male)
Female
Age
0.154*
0.037
0.238**
0.055
0.231**
0.053
-0.119***
-0.718
-0.127***
-0.737
-0.126***
-0.723
0.001***
0.584
0.001***
0.626
0.001***
0.590
0.636***
0.144
0.465***
0.102
0.466***
0.101
Medium
0.054
0.026
0.052
0.024
0.022
0.010
High
0.077*
0.044
0.126***
0.069
0.067
0.036
-0.674***
-0.099
-0.652***
-0.092
-0.636***
-0.089
-0.808***
-0.121
-0.548***
-0.079
-0.435***
-0.062
Age (squared)
Marital status (Ref.: other)
Married
Education (Ref.: low)
-17,953.866
-17,953.866
-17,953.866
LL (model)
-16,707.073
-16,406.003
-16,314.704
2,493.587***
3,095.725***
3,278.324***
AIC
3.557
3.492
3.473
0.069
0.086
0.091
9,408
9,408
9,408
784.738***
182.598***
somewhat different aspects of satisfaction, since both their effects remain significant
when including them in a common model. This result is further supported by the log
rank test displayed at the bottom of the table, which shows that Model 3 constitutes an
improvement over both alternative models, omitting either the deprivation index (M1)
or the indirect indicators (M2). Thus using all indicators for the material situation
discussed jointly instead of considering them alternatives might be the most reasonable
strategy.
123
494
B. Christoph
6 Conclusion
This paper started out with the well-known finding, that the relation between a persons
subjective well-being and his or her material situation is surprisingly weak. This finding is
puzzling, in particular since it contradicts popular prima facie expectations about the
causes of happiness. Since this is so, in recent years this relationship has been increasingly
in the focus of research efforts. Categorizing these efforts, two general approaches have
been identified, those trying to improve empirical models and those trying to find more
adequate measures for the material situation than income (alone).
The current paper tried to make a contribution to the latter line of research. Doing so, it
started out with giving an overview of the different concepts for measuring the material
situation. A particular focus was given to the distinction between direct and indirect
measures for the material situation, which is widely known and applied in the context of
poverty research, but which has hitherto not been given much consideration when
examining the material situation/well-being relationship. Breaking with that practice, this
paper has argued that this distinction might be quite useful when applied in well-being
research and that the relation between well-being and direct measures should be stronger
than the one found for indirect measures.
In order to support this hypothesis, three different German datasets have been used. All
these datasets included an income variable as well as information on the possession of
various items, which is necessary to compute a so-called deprivation index, a very
prominent direct measure for the material situation. Two of the datasets included additional
information on wealth. Using this data, it could be shown that the relation between wellbeing and the direct measure for the material situation is indeed considerably stronger than
the one found when using its indirect counterparts. This will not only hold, when considering the relationship between satisfaction and income but will also remain true, when
supplementing the income variable by additional wealth-information.
This finding fits well with previous research as e.g. the one by Headey and his colleagues, which argues that using more advanced material measures than income alone
might quite well lead to a re-evaluation of the influence material circumstances have on
subjective well-being. It adds to the previous argument, by noting that applying direct
measures like a deprivation index will be a significant improvement over analyzing indirect
measures like income and wealth alone. It also gives rise to some additional questions,
which might be the focus of future research. The first would surely be to use other than
German data in order to confirm the results and safeguard against the somewhat unlikely
but still possible case that the results reported above represent a German particularity rather
than indicating a generally valid relation. The second would be to start a new attempt to use
detailed consumption data. Even though Headey et al. (2005, 2008) have made a first
attempt to studying the influence of consumption on well-being, in which they found a
small effect for some countries and for others none at all, in the authors opinion this
should not settle the case. This is mainly since Headey and his colleagues used very crude
and strongly aggregated consumption data and it seems not unlikely that their findings are
at least in part a consequence of this crudeness rather than of the general quality of
consumption data as predictors of subjective well-being. Unfortunately, datasets covering
consumption in great detail usually do not account for subjective variables like well-being
measures. Therefore it might take some time until the relation between detailed consumption measures and satisfaction might be up for analysis, which, however, does not
make the endeavor a less promising one. Nevertheless, until then or until some more
advanced measure will be applied, combining direct measures like a deprivation index with
123
495
information on income and wealth should be considered the best option to analyze the
relation between subjective well-being and the material situation.
Acknowledgments The author likes to thank Heinz-Herbert Noll, Gerhard Krug, Torsten Lietzmann and
the anonymous reviewer for helpful comments.
Appendix
See Table 7.
GSOEP
WS98
PASS
No
Yes
Yes
No
No
Yes
Yes
No
No
Yes
No
No
Yes
No
Yes
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
Yes
10
No
Yes
Yes
11
Yes
Yes
Yes
12
No
No
Yes
13
Yes
Yes
Yes
14
Yes
Yes
Yes
15
No
Yes
Yes
16
No
No
Yes
17
A newspaper subscription
No
Yes
No
18
A telephone
Yes
Yes
No
19
A car
Yes
Yes
Yes
20
Yes
Yes
Yes
21
No
Yes
Yes
22
Yes
Yes
Yes
23
A hi-fi system
No
Yes
No
24
A washing machine
No
Yes
Yes
25
A dishwasher
No
Yes
No
26
No
No
Yes
123
496
B. Christoph
Table 7 continued
Item
GSOEP
WS98
PASS
27
Yes
No
Yes
28
Yes
No
No
29
Yes
Yes
Yes
30
No
No
Yes
31
No
No
Yes
32
No
Yes
No
33
No
Yes
No
34
No
Yes
No
35
To be able to pay the rent for the apartment and/or the interest
on the house or apartment one lives in always on time
(GSOEP: no information whether due to financial or other
reasons)
Yes
No
Yes
36
No
No
Yes
37
To be able to buy over-the-counter drugssuch as painrelievers or medicine against coldif you need them, even if
your health insurance does not cover the costs.
No
No
Yes
References
Andress, H.-J., Kruger, A., & Sedlacek, B. K. (2004). Armut und Lebensstandard. Zur Entwicklung des
notwendigen Lebensstandards der Bevolkerung. Koln: Gutachten im Rahmen des Armuts- und
Reichtumsberichts der Bundesregierung.
Andress, H.-J., & Lipsmeier, G. (2001). Armut und Lebensstandard. In Bundesministerium fur Arbeits und
Sozialordnung (BMAS) (Ed.), Lebenslagen in Deutschland. Der erste Armuts- und Reichtumsbericht
der Bundesregierung. Bonn: BMAS.
Andress, H.-J., Lipsmeier, G., & Lohmann, H. (2001). Income, expenditure and standard of living as poverty
indicatorsdifferent measures, similar results? Schmollers Jahrbuch, 121, 165198.
Argyle, M. (1999). Causes and correlates of happiness. In D. Kahnemann, E. Diener, & N. Schwartz (Eds.),
Well-being. The foundations of hedonic psychology (pp. 353373). New York: Russel Sage.
Argyle, M. (2001). The psychology of happiness (2nd ed.). London: Routledge.
Boes, S., & Winkelmann, R. (2006). The effect of income on positive and negative subjective well-being.
Sozialokonomisches Institut der Universitat Zurich, Working paper no. 0605. Zurich: Universitat
Zurich.
Boes, S., & Winkelmann, R, (2009). The effect of income on general life satisfaction and dissatisfaction.
Social Indicators Research, Online First, Retrieved June 2, 2009 from www.springerlink.com.
Bohnke, P. (2005). First European quality of life survey: Life satisfaction, happiness and sense of belonging.
Luxembourg: Office for Official Publications of the European Communities.
Bohnke, P., & Delhey, J. (1999a). Lebensstandard und Armut im vereinten Deutschland. WZB working
paper FS III 99-408. Berlin: Social Science Research Center Berlin (WZB).
123
497
Bohnke, P., & Delhey, J. (1999b). Uber die materielle zur inneren Einheit? Wohlstandslagen und subjektives Wohlbefinden in Ost- und Westdeutschland. WZB working paper FS III 99-412. Berlin: Social
Science Research Center Berlin (WZB).
Brauns, H., & Steinmann, S. (1999). Educational reform in France, West-Germany and the United Kingdom:
Updating the casmin classification. ZUMA-Nachrichten, 44, 745.
Brewer, M., Goodman, A., & Leicester, A. (2006). Household spending in Britain. What can it teach us
about poverty? Bristol: Policy Press.
Brewer, M., Muriel, A., Phillips, D., & Sibieta, L. (2008). Poverty and inequality in the UK: 2008. IFS
Commentary No. 105. London: Institute for Fiscal Studies.
Bungeler, K., Gensicke, M., Hartmann, J., Jackle, R., & Tschersich, N. (2009). IAB-Haushaltspanel im
Niedrigeinkommensbereich. Welle 2 (2007/08). Methoden- und Feldbericht. FDZ Metodenreport Nr. 8/
2009. Nurnberg: Bundesagentur fur Arbeit.
Callan, T., Nolan, B., & Whelan, C. T. (1993). Resources, deprivation and the measurement of poverty.
Journal of Social Policy, 22, 141172.
Christoph, B., Muller, G., Gebhardt, D., Wenzig, C., Trappmann, M., Achatz, J., Tisch, A., & Gayer, C.
(2008). Codebook and documentation of the panel study Labour Market and Social Security (PASS).
Wave 1 (2006/2007). Volume I: Introduction and overview. FDZ Datenreport Nr. 5/2008. Nurnberg:
Bundesagentur fur Arbeit.
Clark, A. E., Etile, F., Postel-Vinay, F., Senik, C., & Van der Straeten, K. (2005). Heterogeneity in reported
well-being: Evidence from twelve European Countries. The Economic Journal, 115, C132C188.
Clark, A. E., & Oswald, A. J. (1994). Unhappiness and unemployment. The Economic Journal, 104, 648
659.
Cummins, R. A. (2000). Personal income and subjective well-being: A review. Journal of Happiness
Studies, 1, 133158.
Cummins, R. A. (2002). Subjective well-being from rich and poor. In W. Glatzer (Ed.), Rich and poor.
Disparities, perceptions, concomitants (pp. 137156). Dordrecht: Kluwer.
DAmbrosio, C., Frick, J. R., & Jantti, M. (2009). Satisfaction with life and economic well-being: Evidence
from Germany. Schmollers Jahrbuch, 129, 283295.
Desai, M., & Shah, A. (1988). An economic approach to the measurement of poverty. Oxford Economic
Papers, 40, 505522.
Diener, E., & Biswas-Diener, R. (2002). Will money increase subjective well-being? A literature review and
guide to needed research. Social Indicators Research, 57, 119169.
Diener, E., & Lucas, R. E. (1999). Personality and subjective well-being. In D. Kahnemann, E. Diener, &
N. Schwartz (Eds.), Well-being. The foundations of hedonic psychology (pp. 213229). New York:
Russel Sage.
Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of
progress. Psychological Bulletin, 125, 276302.
Duncan, O. D. (1975). Does money buy satisfaction? Social Indicators Research, 2, 267274.
Easterlin, R. A. (2001). Income and happiness: Towards a unified theory. The Economic Journal, 111, 465
484.
Easterlin, R. A. (2002). The income-happiness relationship. In W. Glatzer (Ed.), Rich and poor. Disparities,
perceptions, concomitants (pp. 151175). Dordrecht: Kluwer.
Ferrer-i-Carbonell, A., & Frijters, P. (2004). How important is methodology for the estimates of the
determinants of happiness? The Economic Journal, 114, 641659.
Frijters, P., Haisken-DeNew, J. P., & Shields, M. A. (2004). Money does matter! Evidence from increasing
real income and life satisfaction in East Germany following reunification. The American Economic
Review, 94, 730740.
Gerlach, K., & Stephan, G. (1996). A paper on unhappiness and unemployment in Germany. Economic
Letters, 52, 325330.
Hagenaars, A. J. M., de Voos, K., & Zaidi, M. A. (1994). Poverty statistics in the late 1980s: Research
based on micro-data. Luxembourg: Office for Official Publication of the European Communities.
Haisken-DeNew, J. P., & Frick, J. R. (2005). DTC. Desktop companion to the German socio-economic panel
(SOEP). Version 8.0. Updated to wave 21 (U). Berlin: German Institute for Economic Research (DIW).
Hallerod, B. (1994). A new approach to the direct consensual measurement of poverty. SPRC discussion
paper no. 50. Sydney: Social Policy Research Centre.
Hallerod, B. (1995). The truly poor: Direct and indirect consensual measurement of poverty in Sweden.
Journal of European Social Policy, 5, 111129.
Headey, B. (2008). Poverty is low consumption and low wealth, not just low income. Social Indicators
Research, 89, 2339.
123
498
B. Christoph
Headey, B., Muffels, R., & Wooden, M. (2005). Money and happiness: The combined effects of wealth,
income and consumption. Schmollers Jahrbuch, 125, 131144.
Headey, B., Muffels, R., & Wooden, M. (2008). Money does not buy happiness: Or does it? A reassessment
based on the combined effects of wealth, income and consumption. Social Indicators Research, 87, 65
82.
Headey, B., & Wearing, A. J. (1989). Personality, life events and subjective well-being: Toward a dynamic
equilibrium model. Journal of Personality and Social Psychology, 57, 731739.
Headey, B., & Wooden, M. (2004). The effects of wealth and income on subjective well-being and ill-being.
IZA discussion paper no. 1032. Bonn: Institute for the Study of Labor (IZA).
Howell, C. J., Howell, R. T., & Schwabe, K. A. (2005). Does wealth enhance life satisfaction for people who
are materially deprived? Exploring the association among the orang asli of peninsular Malaysia. Social
Indicators Research, 76, 499524.
Kroh, M., & Spiess, M. (2008). Documentation of sample sizes and panel attrition in the German socio
economic panel (SOEP) (1984 until 2007). DIW data documentation, no. 39. Berlin: German Institute
for Economic Research (DIW).
Lipsmeier, G. (1999). Die Bestimmung des notwendigen LebensstandardsEinschatzungsunterschiede und
Entscheidungsprobleme. Zeitschrift fur Soziologie, 28, 281300.
Mack, J., & Lansley, S. (1985). Poor Britain. London: Allen and Unwin.
Markus, H. R., & Kitayama, S. (1994). The cultural construction of self and emotion: Implications for social
behavior. In S. Kitayama & H. R. Markus (Eds.), Emotion and culture: Empirical studies of mutual
influence (pp. 89130). Washington, DC: American Psychological Association.
Meyer, B. D., & Sullivan, J. X. (2003). Measuring the well-being of the poor using income and consumption. Journal of Human Resources, 38(Supplement), 11801220.
Michalos, A. C. (1985). Multiple discrepancies theory (MDT). Social Indicators Research, 16, 347413.
Mller, V. (2007). Satisfied and dissatisfied South Africans: Results from the general household survey in
international comparison. Social Indicators Research, 81, 389415.
Mullis, R. J. (1992). Measures of economic well-being as predictors of psychological well-being. Social
Indicators Research, 26, 119135.
Nolan, B., & Whelan, C. T. (1996). Measuring poverty using income and deprivation indicators: Alternative
approaches. Journal of European Social Policy, 6, 225240.
Noll, H.-H., & Weick, S. (2007). Einkommensarmut und Konsumarmutunterschiedliche Perspektiven und
Diagnosen. Analysen zum Vergleich der Ungleichheit von Einkommen und Konsumausgaben.
Informationsdienst Soziale Indikatoren (ISI), 37, 16.
Oishi, S. (2000). Goals and cornerstones of subjective well-being: Linking individuals and cultures. In
E. Diener & E. M. Suh (Eds.), Culture and subjective well-being (pp. 87112). Cambridge, Massachusetts: MIT Press.
Ringen, S. (1988). Direct and indirect measures of poverty. Journal of Social Policy, 17, 351365.
Rudolph, H., & Trappmann, M. (2007). Design und Stichprobe des Panels Arbeitsmarkt und Soziale
Sicherung (PASS). In M. Promberger (Ed.), Neue Daten fur die Sozialstaatsforschung. Zur Konzeption der IAB-Panelerhebung ,,Arbeitsmarkt und Soziale Sicherung. IAB Forschungsbericht No. 12/
2007. Nurnberg: Institute for Employment Research (IAB).
Schafer, A., & Schupp, J. (2006). Zur Erfassung der Vermogensbestande im Sozio-oekonomischen Panel
(SOEP) im Jahr 2002. DIW data documentation, no. 11. Berlin: German Institute for Economic
Research (DIW).
Schyns, P. (2002). Wealth of nations, individual income and life satisfaction in 42 countries: A multilevel
approach. Social Indicators Research, 60, 540.
Schyns, P. (2003). Income and life satisfaction. A cross-national and longitudinal study. Delft: Eburon.
Statistik der Bundesagentur fur Arbeit [STBA]. (2009). Grundsicherung fur Arbeitssuchende in Zahlen. Mai
2009. Nurnberg: Bundesagentur fur Arbeit.
Statistisches Bundesamt [StaBA] (2009). Verbraucherpreisindex fur Deutschland. (Wiesbaden: Statistisches
Bundesamt). Retrieved July 7, 2009, from http://www.destatis.de/jetspeed/portal/cms/Sites/destatis/
Internet/DE/Content/Statistiken/Zeitreihen/WirtschaftAktuell/Basisdaten/Content100/vpi101a.psml.
Stouffer, S. (1949). The American soldier. Princeton: Princeton University Press.
Thum, M., Kuwan, H., & von Rosenbladt, B. (1999). Wohlfahrtssurvey 1998Methodenbericht. Munchen:
Infratest Burke Sozialforschung.
Townsend, P. (1979). Poverty in the United Kingdom. Harmondsworth: Penguin Books.
Trappmann, M., Christoph, B., Achatz, J., Wenzig, C., Muller, G., & Gebhardt, D. (2009). Design and
stratification of PASS. A new panel study for research on long term uemployment. IAB discussion
paper no. 05/2009. Nurnberg: Institute for Employment Research (IAB).
Veenhoven, R. (1991). Is happiness relative? Social Indicators Research, 24, 134.
123
499
Veenhoven, R. (1994). Is happiness a trait? Tests of the theory that a better society does not make people
any happier. Social Indicators Research, 32, 101160.
Veenhoven, R., & Erhard, J. (1995). The cross-national pattern of happiness. Test of predictions implied in
three theories of happiness. Social Indicators Research, 34, 3368.
von Rosenbladt, B., Siegel, N. A., Stimmel, S., & Stutz, F. (2007). SOEP 2006. Erstbefragung der
Erganzungsstichprobe H. Methodenbericht. Munchen: TNS-Infratest Sozialforschung.
Wagner, G. G., Frick, J. R., & Schupp, J. (2007). The German socio-economic panel study (SOEP)Scope,
evolution and enhancements. Schmollers Jahrbuch, 127, 139169.
123