The relationship between subjective well-being and a
non-adequate level of food consumption
A bachelor thesis
Sophie van den Ende
Supervisor: Prof. M.P. Pradhan
Bachelor Economics and Business, specialization Economics
Field of study: Development economics
Student number: 10025707
01 – 08 – 2014, Amsterdam
Abstract
This paper focuses on the relationship of two subjective well-being measures; the total well-being based on a sixth step ladder and the food consumption adequacy. In the existing literature, a gap can be found in the relationship between different subjective measures, since the focus lies on a the comparison of the objective measure and subjective measures of poverty.
The hypothesis of this study is that stating to have a non-adequate level of food
consumption has an uniform negative effect on the subjective well-being. Using the data of the Indonesian Family Life Survey from 2000, the hypothesis has been tested by regression the food consumption adequacy on the subjective well-being with different control variables. This regression is an oprobit regression and herewith the marginal effects of the variables are measured which were necessary to test the hypothesis. The overall conclusion of the several regressions is that the relationship of stating to have a non-adequate level of food consumption and subjective well-being is negative but not completely linear or uniform. This means, for example, that stating to have a non-adequate level of food consumption does increase your chances to be in the lower steps of
subjective well-being but you are more likely to place yourself in the 2nd step instead of the 1st step.
With these results the conclusion can be drawn that stating to have a non-adequate level of food consumption has the biggest effect on the lower steps of the subjective well-being, the steps where people consider themselves to be relatively poor. But the magnitude of the influence of the non-adequate statement is lower than what would be expected, this could be due to the improving state of economy of Indonesia.
Table of contents Abstract 2 Table of contents 3 List of figures 3 1. Introduction 4 2. Literature overview 5
2.1 Comparison of income and subjective well-being 5
2.2 Subjective well-being in general 5
2.3 Subjective well-being in Indonesia 6
2.4 Food consumption adequacy 7
2.5 Summary of the literature overview 8
3. Descriptive data 9 4. Methodology 12 5. Results 14 6. Conclusion 18 7. Bibliography 19 8. Appendix 21 List of figures:
Table 1: Distribution of subjective well-being and average total consumption 10 in Indonesia, 2000
Table 2: Distribution of food consumption adequacy and average food 10 consumption, 2000
Table 3: Food consumption adequacy and subjective economic well-being in 2000 11 Figure 1: Total household consumption compared with subjective well-being and 12
food consumption adequacy, 2000
Table 4: Marginal effects of food consumption adequacy on the different 13 well-being steps
Table 5: Ordered probit regression of subjective well-being in 2000 15 Table 6: Marginal effects of a non-adequate level of food consumption on 16
the subjective well-being
Table 7: Marginal effects of non-adequacy of food consumption on merged 17 subjective well-being steps
Table 8: Marginal effects of independent variable on the different steps of 22 subjective well-being
Table 9: Ordered probit regression of subjective well-being with merged steps 23
1. Introduction
Although economist and psychologists have been using subjective measures since the 60’s, it is only quite recently that the measures are being implemented in developing and transition economies. The advantage of subjective being as measure is that it can capture different aspects of well-being than the standard objective measures, such as the poverty line (Lokshin, Umapathi, &
Paternostro, 2006). In the last decade, the research in this field has been focused on comparing the objective and subjective measures and looking for similarities and differences in which they
determine the level of poverty in a population. The resemblance is general reasonably high although the overlap depends on the background of the country, whether the country can be considered as being developed, developing or underdeveloped (Howell & Howell, 2008). Another feature, which has been researched quite often in the last decade, is the influence of various individual and household characteristics on the total subjective well-being. The significance of these defining variables also tend to differ through the state of the country, although some characteristics have a similar influence on the subjective well-being in every country (Helliwell, 2003).
Within the subjective qualitative school, a variety of different measures can be found as well which focuses on the different aspects of life such as health, food or education. These specific well-being aspects have influence on the total perception of well-being, but the extent of this influence hasn’t been researched so far. A literature gap can be found in the research concerning the composition of the total subjective well-being out of the different concentrated subjective topics. So far the research hasn’t tested the magnitude of these smaller subjective measures on the total subjective well-being.
The subjective aspect of food is researched by asking respondents their food consumption adequacy. In this question the respondents state whether they consider their food consumption to be adequate for their needs. This question is the subjective version of the objective food
consumption level, which determines the poverty level on the amount of calories needed for a healthy life. The subjective aspect is assumed to be the most important in determining the level of
subjective well-being (Howe, Hargreaves, Ploubidis, De Stavola, & Huttly, 2011).
In this research two subjective measures are studied; the total well-being and the food consumption adequacy. With the food consumption adequacy, the focus lies on the part of the sample that stated to have a non-adequate level of food consumption and the subjective well-being
is based on a 6th step ladder from poorest to richest. This study answers the following question:
‘What is the effect of stating to have a non-adequate level of food consumption on the tot subjective well-being in Indonesia?’.
The answer to this question will be researched in different steps; first a literature overview will be given on the executed research on subjective measures, then some descriptive data of the survey in Indonesia will be elaborated, in the methodology section the regression used to determine the results will be explained, in the results the hypothesis will tested, and in the end a conclusion will be drawn.
2. Literature overview
In this section the literature is described which is necessary to answer the research question. This means that the existing literature about subjective economic welfare and subjective food
consumption is defined. Next to that, at times, an objective measure is compared with the subjective measures to show the significance or advantages of subjective measures. The literature so far has mostly researched the overlap between the different measure with a focus on the relation between subjective well-being and objective measures.
2.1 Comparison of income and subjective well-being
An example of a research between an objective measure and subjective well-being is the following research in which they considered the gross domestic product as an objective measure and the average happiness as a subjective measure. There is a quite strong relationship between these two measures with correlations between 0.6 and 0.84. These correlations tend to differ a lot among developed and developing countries, although in general the relationship is stronger in poorer and developing countries (Howell & Howell, 2008). This is an outcome of the fact that the relationship between the average subjective well-being and the economic status of a country is concave. As income rises, each additional increase in income adds less additional happiness. The absolute income and assets have a strong influence on the well-being until a certain necessary level of food, clothing and shelter is reached. When a population has acquired this necessary level of income, at which physical functioning is achieved, the relative economic status becomes more important. This relative economic status is determined by several comparisons such as with the past, the
aspirations in the future and the economic status of the people that surround you. So the increase in income will increase happiness most strongly if the extra income improves social functioning; which is the development of social connections, growth of self-respect or self-esteem (Howell & Howell, 2008; Kingdon & Knight, 2006). Even though the correlation between income and well-being is relatively high in developing countries; it is still beneficial to test the subjective well-well-being as well since you can incorporate factors, next to income and consumption, which determine the quality of one’s life. Some of these quality of life determining factors are hard, if not impossible, to capture in objective measures. (Kingdon & Knight, 2006).
2.2 Subjective well-being in general
Since the development of subjective well-being in the 1960’s, the research has been mostly
conducted using respondent’s answers in surveys. The questions that the respondents had to answer could differ a lot among the different researches. There are questions such as the Cantril ladder, in which respondents have to rank themselves in terms of subjective well-being or happiness. This ladder can be used for economic well-being when respondents are asked to put themselves on a poverty scale (Carletto & Zezza, 2006). Another question that is used to measure economic well-being is the minimum income question (MIQ), in which respondents are asked what they consider to be the absolute minimum income level. This level is then used to determine where the respondents
place themselves on the income scale (Ravallion, Poor, or just feeling poor? On using subjective data in measuring poverty, 2012). This MIQ method is less suitable for developing countries since the perception of income may differ greatly among respondents (Carletto & Zezza, 2006). An individual, on behalf of the whole family, usually answers both kinds of questions. But Kingdon and Knight (2006) tested individual characteristics in a regression and found out that these
characteristics are generally insignificant in the subjective well-being equations.
There have been a couple of articles that studied the effects of certain characteristics or socioeconomic factors of the household or individual on subjective poverty by using an ordered probit regression. In general there are factors that influence the subjective economic status in every country, such as health, employed status, family status, age, religion and political
institutions. Helliwell (2003) conducted a regression with subjective well-being as the dependent variable and a comprehensive amount of independent variables. One of the most significant independent variables in this research is the subjective state of health; respondents were asked to rank their health status on a scale from 1, very good health, to 5, poor health. A one percent increase in the reported health increases the subjective well-being with a little more than one percent. Another significant independent variable is the employment status; respondents experience a great reduction in subjective well-being when they get unemployed. The marital status has a significant effect on the 10-point well-being scale as well; being married versus being separated increases your well-being with almost ¾ of a point. This is a bigger change than going from being unemployed to being employed (Helliwell, 2003). Next to these individual
characteristics, the subjective well-being is influenced by reference dependence. The well-being depends, just as much as on absolute income, on the relative income position. This relative income position is in correspondence with as well others, as with previous experiences and future
expectations (Castilla, 2010).
2.3 Subjective well-being in Indonesia
Powdthavee (2007) has extensively researched subjective well-being in Indonesia. Just as Helliwell, he has conducted an ordered probit function to test several independent variables. In this research he has found that, ceteris paribus, households with a higher proportion of children and pensioners consider themselves to be poorer. On the other hand, having a higher proportion of adult women in your household has a positive effect on the subjective well-being. Contrary to the research of Helliwell, the marital status has no significant effect on the subjective well-being in Indonesia. But religion does have a significant effect; respondents belonging to the religious minority (non-Islamic) have a lower self-rated economic status. In line with the research of Castilla (2010), the past expenditure, being a reference dependence, has an positive effect on the current well-being; but the coefficient of the past expenditure is only one-third of the coefficient of current expenditure. Another significant reference dependent is the future expectations; respondents that expect to move up on the ladder in the next year, are less likely to report themselves as poor today
(Powdthavee, 2007). In general the data of Indonesia has, according to the research of Powdthavee, a lot of similarities with the literature current research in other countries.
There is a pitfall that one has to keep in mind when using the subjective well-being measure; when using this measure you have to allow for heterogeneity. With latent heterogeneity Ravallion (2012) means that people with similar backgrounds and the same standard of living might report a different answer to the subjective well-being question. These latent personal
characteristics are important since they might be correlated with as well the independent variables as with the subjective well-being level. (Clark, Etilé, et al., 2005). Next to that respondents will answer the questions according to their frame-of-reference, which is depend on their own
knowledge and experience about, among others, their relative position in the hierarchy of poverty and wealth. This skewness is called frame-of-reference bias and might have influenced regressions performed in previous articles (Ravallion, 2012; Kingdon & Knight, 2006).
2.4 Food consumption adequacy
In developing countries it is generally more useful to use the total consumption as objective
measure than income. So in this study, the monthly food and non-food consumption of households is used. The consumption gives a better image of the economic status in the long run since it
incorporates the possibility of households to smooth their state of living by saving and borrowing, in this way you decrease the likelihood of measurement errors (Powdthavee, 2007).
An objective measure of poverty is derived by the food consumption in a country. One of the ways to look at the food consumption is to make a food poverty line, derived with the food energy intake method. This food poverty line is based on the price of consumption needed to take the necessary calories for a healthy life, this minimal calories level is set at 2,100 calories per person per day (Pradhan, Suryahadi, et al., 2000).
With all the food consumption measures there is an inextricably subjectivity and social character to food security and particularly to the selection of basic needs. The distinction between necessities and luxuries is dependent of the social status of an individual relative to the people he is surrounded by. So when working with food consumption, one has to keep in mind that this line distinguishing the necessities and luxuries is socially determined and will keep on changing (Pradhan & Ravallion, 2000).
Next to the objective measures there is also the subjective food adequacy measure. A way to determine the subjective food security is by asking respondents the food consumption adequacy question (fCAQ). In this question respondents are asked to indicate whether their food
consumption is less than adequate, just adequate or more than adequate (Migotto, Davis, Carletto, & Beegle, 2006). In surveys the food consumption adequacy question is usually a part of the
subjective consumption adequacy section in which other parts of the household consumption are tested as well. The other questions are concerning housing, clothing or health care. Overall these questions should give a good overview of the total adequacy of needs. In general it has been assumed that food consumption of total consumption is the part most relevant for subjective
welfare (Howe, Hargreaves, Ploubidis, De Stavola, & Huttly, 2011). In the literature, consumption
adequacy is set to be reached when people have the most essential goods and services to be able to move past the short-term focus on endured existence (Martin & Hill, 2012).
According to a research conducted in Madagascar, in which they implemented an ordered probit regression on consumption adequacy questions, there are some variables that have a significant effect on the food consumption adequacy. Women have, for example, when controlling for household composition and consumption level, more positive perceptions of food consumption adequacy. This could be the case in that women have more efficient ways to use food at home than men so they reach adequacy at a lower food consumption level. Another significant variable is the share of unemployed family members. A higher proportion of unemployed relatives decreases the satisfaction of the food consumption level. Next to that, urban households are, relative to rural households, less satisfied with their level of food consumption. The research of Lokshin, Umapathi and Paternostro (2006) in Madagascar further concludes that food adequacy is strongly influenced by relative consumption. Poorer families that are living in an area with a higher than average income are more likely to assess their adequacy level, on as well food as on clothing and housing expenditures, as being less adequate, when controlled for their own income and other
characteristics. So inequality within the income distribution in an area has a negative effect on the perception of food consumption adequacy (Lokshin, Umapathi, & Paternostro, 2006).
Furthermore, food consumption adequacy is highly correlated with other subjective perceptions concerning the general household situation and the relative economic status. Next to that, perceptions of food adequacy are not only depending on the current situation but also on past and future changes in economic status. People are more likely to consider their food consumption as being adequate if they have the feeling that their life has improved in the last three years (Migotto, Davis, Carletto, & Beegle, 2006).
Consumption adequacy questions seem a good criterion to measure subjective poverty. A disadvantage though is that it is not possible to compare the answers to the consumption adequacy questions between different countries. Respondents within different countries have very different perceptions of the level at which they reach adequacy (Migotto, Davis, Carletto, & Beegle, 2006).
In the research of Pradhan and Ravallion (2000) in Jamaica and Nepal and the research of Lokshin, Umapathi and Paternostro (2006) in Madagascar, objective measures are compared with a subjective poverty line based on consumption adequacy questions. In both the researches the two measures correspond quite well. Even when there is only looked at the food consumption adequacy, and not at the subjective adequacy of housing, health or clothing, the results of the subjective poverty line stay fairly robust. This in contrast with a subjective poverty line based on the minimum income question which is difficult to use in developing countries. As a result the MIQ-based poverty line shows a low similarity with as well the objective measures as with the other subjective poverty measures (Pradhan & Ravallion, 2000; Lokshin, Umapathi, & Paternostro, 2006).
2.5 Summary of the literature overview
Within the years the literature on subjective poverty in developing countries has been growing, and most of these researches focussed on the relation and similarities between objective measures, such as the poverty line, and subjective measures. On average the objective and subjective measures are resonable similar. As discussed, the subjective well-being is influenced by a lot of
factors of which some are included in the objective measures and others are not. One of the factors that has an influence on the subjective well-being is food consumption. Food security can also be measured in as well objective as subjective ways. To determine the subjective food adequacy, respondents are asked whether they consider their food consumption to be adequate. So far there hasn’t been an extensive research with compares the objective and subjective measures of food consumption. Another aspect which hasn’t been researched extensively is the part subjective food adequacy takes in determining the subjective well-being. In the following part, the basic facts of the data will be shown and then there will be determined which part the subjective food adequacy takes in determining the well-being.
3. Descriptive data
In the coming part the basic statistics concerning the subjective well-being, subjective food consumption adequacy and the objective food and non-food consumption have been elaborated.
The data used in this research is part of the Indonesian Family Life Survey (IFLS) of 2000. IFLS is an ongoing household survey, focussing on socioeconomic and health subjects. The survey is representative for 83% of the Indonesia population, who are living in 13 of the 26 provinces in 1993. The IFLS fielded in 2000 is the third wave; the first and second survey have been conducted in 1993 and in 1997. The sample in 2000 consisted out of 39,000 individual interviews, covering 10,400 different households (RAND).
In the IFLS respondents are asked about their total household consumption level per month. This subject has been subdivided in a lot of sub questions like: ‘In the past week what was the total expenditure to purchase …’ and ‘Approximately what was the total value of items consumed by this household that were self-produced or received from another source during the last week?’. All these questions combined give a reasonable overview of the total food and household consumption. As said in the literature overview, in Indonesia, since it is a developing country, the household consumption makes more sense to use as an objective measure than the income since income can be very unreliable. All the consumption data used in this research has been adapted to the
consumption per person, in this way there can be controlled for the differences in household sizes. Furthermore, the consumption has been processed from weekly to monthly expenditure.
IFLS 3 extended the previous IFLS by adding a set of question about subjective well-being in the past, present and future. This article will focus on a few questions of this set, the questions about the current subjective well-being and the food consumption adequacy. The question about the current subjective well-being in the IFLS is along the lines of a Cantril-ladder (described in the literature review). The respondents were asked the following question: ‘Please imagine a six-step ladder where on the bottom (the first step), stand the poorest people, and on the highest step (the sixth step), stand the richest people. On which step are you today?’ Table 1 shows a raw
distribution of the answers to the subjective well-being question, combined with the average level of the different steps of total consumption per person in the households.
Table 1 – Distribution of subjective well-being and average total consumption in Indonesia, 2000
Approximately 99% of the IFLS respondents answered this question; only 0.31% didn’t know where to place themselves on the ladder. As can be seen above, most individuals consider themselves to be on the third step of the ladder. Only a very small percentage, 1.3%, places themselves on the fifth and sixth step, with which they consider themselves to belong to the richest part of the population. As expected, the average consumption level increases with higher steps of the ladder. The biggest difference consumption can be seen in the change from the third to the fourth step,
whereas the differences between the consumption of the 1st and 2nd step and the 4th and 5th step
are relatively small.
Powdthavee (2007) composed a cross-tabulation of the economic well-being and
expenditures using the IFLS 2000 data. In this table he merged to fifth and sixth step of the ladder since they both contain a very small part of the respondents. He divided the household expenditure in ascending quintiles and compared these with the ascending subjective economic status. Roughly 24 percent of the respondents have placed themselves in the same step in the objective and subjective economic status; so for example by saying that they consider themselves to the poorest in the subjective status and also belonging in the lowest expenditure quintile. He has confirmed this relatively weak correlation by conducting the Cramer’s V test; which indicates a correlation of 0.1435.
Table 2 – Distribution of food consumption adequacy and average food consumption in Indonesia, 2000
Table 2 displays the raw distribution of the food consumption adequacy. Respondents were asked the following question: ‘Concerning your food consumption last month, which of the following is true? It is less than adequate for my needs, it is just adequate for my needs, it is more than adequate for my needs.’ Table 2 shows that a majority of 73 percent states that they have just enough food to meet their daily food demand. Whereas 17 percent affirms that they have more than enough food and 9 percent that they can’t meet their daily food demand. The food
consumption column shows the average amount of Rupiahs spend on food consumption per person for the households in the different steps. Just as with the subjective well-being, the consumption
10
On which economic step are you today? Observations Percentage Total consumption
Poorest 1 1202 4.73 172758.1 2 5280 20.78 174138.5 3 14404 56.69 249637.9 4 4193 16.50 367607.1 5 252 0.99 368470.5 Richest 6 76 0.30 456365.6 Total 25407 100.00
Assessment of food consumption: Observations Percentage Food consumption
Not adequate 2374 9.32 97401.3
Just adequate 18602 73.00 128388.6
More than adequate 4505 17.68 152461.3
increases as expected; respondents stating that they have an not adequate level, have a lower level of food consumption that the respondents stating that they have just enough or more than enough. In the following table the two subjective measures, food consumption adequacy and well-being are processed to be able to take a look at their correlation.
Table 3 - Food consumption adequacy and subjective economic well-being in Indonesia, 2000
In the table the numbers represent percentages, which show the relation between the two subjective measures. The numbers work in the following way: of all the people who place
themselves on the first, poorest, step of the economic well-being ladder 35.86 percent states that their food consumption is not adequate, 57.99 percent states that is it just adequate and 6.16 percent states that their level of food consumption is more than adequate. In the table you can see a reasonable shift from the first to the fifth step of the economic status; the higher the step on the ladder, the higher the percentage of people stating to have a ‘just adequate’ or ‘more than
adequate’ level of food consumption. Only in the sixth step of the ladder, the richest, there is a remarkable percentage; the percentage of people stating to have a level of food consumption which is not adequate is higher than this percentage in the third, fourth and fifth step. But looking at the frequency of the numbers, this isn’t a discontinuity; there are only four people in the sixth step stating not to have enough food, whereas this number is much higher in the third (751), fourth (114) and fifth step (7).
Since the data used here is ordinal collapsed data, the correlation can be calculated by using gamma’s correlation measure. The value of gamma in table 3 is 0.4784; so there is a positive relation between food adequacy and subjective economic well-being but it is only moderately strong.
In the following descriptive data the consumption, total food and total household, are taking into account. On the vertical axe the average monthly consumption in rupiahs is given for the year 2000, the horizontal axe shows the subjective well-being, in which the fifth and sixth steps are taking together. The first bar shows the average consumption in thousand rupiahs for respondents that consider themselves to be on the poorest step of the economic well-being ladder and that consider their food consumption not to be adequate.
11
Assessment of food On which economic step are you today?
consumption: 1 2 3 4 5 6 I don't know Total
Not adequate 35.86 19.87 5.21 2.72 2.78 5.26 22.5 9.31
Just adequate 57.99 70.11 78.09 64.99 51.98 59.21 67.5 72.98
More than adequate 6.16 10.02 16.68 32.29 45.24 35.53 6.25 17.67
I don't know 0 0 0 0 0 0 3.75 0.02
Figure 1 - Total household consumption compared with subjective well-being and food consumption adequacy in Indonesia, 2000
In the graph there is a clear line recognizable; the consumption increases in as well the economic steps as in the prospect of adequacy of food consumption. In the food consumption figure the fourth and fifth plus sixth step are shifted; on average respondents on the fourth step spend more on food than respondents on the fifth and sixth step, regardless their prospect on food adequacy. On the first subjective economic well-being the respondents stating that they have just enough food have a higher consumption than the respondents stating that they have a more than adequate level of food consumption. All the other steps are completely in line with what would be expected.
Unfortunately, all this information doesn’t show the extent to which the food consumption adequacy determines the level of subjective well-being. In the methodology and results section there will be looked for an answer to this question.
4. Methodology
In this section the structure of a regression, with which the effect of stating to have a non-adequate level of food consumption on the total subjective well-being can be measured, will be explained. Since the dependent variable, subjective well-being, is an ordered variable, a normal OLS-regression is not suitable in this case. Instead an ordered probit regression will be used in
12
0 50 100 150 200 250 300 350 400 450 1 2 3 4 5 + 6 Monthly household consumption in thousand RupiahOn which economic step are you today?
Not adequate Just adequate More than adequate
0 20 40 60 80 100 120 140 160 180 200 1 2 3 4 5 + 6 Monthly household food consumption in thousand Rupiah
On which economic step are you today?
Not adequate Just adequate More than adequate
Table 4: Marginal effects of food consumption adequacy on the different well-being steps
which the B-coefficients show a positive or negative effect on the dependent variable. The basic
form of this regression looks the following way: SWi = βXi + εi . In this regression the main
independent variable will be the subjective food consumption adequacy but are some other
variables necessary as control variables. All these independent variables are covered by the X-term and have a certain β-coefficient after the regression. The error term ε rectifies for the inaccuracy of the respondents when they are evaluating their subjective well-being. In the model the food consumption adequacy variable is transformed into a dummy variable to have a better overview of the results. A dummy variable has been made for the ‘not adequate’ answer and the ‘just adequate’ answer. The answer ‘more than adequate’ has the role of reference variable.
Based on the literature, the hypothesis is that one’s state of a non-adequate level of food consumption has a uniform negative effect on the subjective well-being. This means that stating to have a non-adequate level will increase your chances to be in the lowest steps of subjective well-being and decrease your chances to be in the higher steps.
When there is a simple regression conducted, with only food consumption adequacy as independent variable, the β-coefficient of both the dummy variables is negative. This means that if your food consumption has a not adequate or just adequate level, you are more likely to be in the lower steps of the well-being ladder. This result is completely conform the expectations. Since there can’t be anything concluded from the magnitude of the B-coefficient, the marginal effects are calculated. In the table the marginal effects are ranged for the different well-being steps, based on the simple regression model with only the food consumption adequacy variable.
Since the independent variables are dummy variables, the marginal effects are based on a change from 0 to 1. So the table works the following way: if you change to a not adequate level of food consumption (from either the just adequate or more adequate, that doesn’t matter), you are 27.4%
more likely to be in the 1st step, 23.7% more likely to be in the 2nd step but 32.4% less likely to be in
the 3rd step etc. The 5th and 6th step show less predictable results: they only have very low negative
percentages, where higher, but negative, percentages are expected. This could be because there is
only a few respondents in the 5th and 6th or because there are some control variables necessary.
Based on the literature review and the regression made in the articles, there are some variables that can be added as control variables to improve the regression. One section of these control variables will be the objective measure of well-being, which is the consumption. This consumption level has been divided into a food and a non-food consumption level, but both of the levels are the household consumption divided by the household size. Next to that, some personal characteristics have been accounted for as well, such as the gender, age, marriage and
employment status. The marriage and employment status have been transformed into a dummy
13
1st step 2nd step 3rd step 4th step 5th step 6th step
Not adequate 27.4% 23.7% - 32.4% -17.3% -1.1% - 0.3%
variable as well, with a division between being married or not, and being employed or unemployed. According to previous research, there are also some household characteristics that have to be taken into account, being the living area (which is either urban or rural), the total household size and the number of children still living at home.
Powdthavee (2007) concluded in his research that the subjective health status also plays a big role in determining the subjective well-being. The subjective health status is determined in the IFLS data with a similar question as the food adequacy; whether they consider their health status to be more, just or not adequate. Just as the food consumption adequacy, the health status has been transformed into a dummy variable with the ‘more than adequate’ answer as reference point.
In the results, the extended regression and the conclusions that can be drawn from this outcome have been studied.
5. Results
In the table 5 the more extended regressions have been elaborated. In regression A, the only independent variables are the food consumption adequacy levels, in regression B the log of food and non-food consumption have been added and in regression C several individual and personal characteristics have been included as well. The outcome shows that only the male and child variable are not significant, and the living area (so rural or urban) is only significant at a 5% significant level. All the other variables are significant at a 1% significance level.
With adding the extra independent variables, the pseudo R2 increases which means that a
higher level of the model is explained with the variables that are included in the regression. The
biggest increase in pseudo R2 is from regression A to B with adding the log of food and non-food
consumption, this causes the pseudo R2 to increase with 71%. The shift from regression B to C
increases the pseudo R2 with 11%.
As explained in the methodology, since the magnitude doesn’t include any information about the independent variable in an ordered probit regression; only the sign of B-coefficient can be used, not his magnitude. In all the regressions the B-coefficient of the different answers on food consumption adequacy lie in line with the expectations; there is a negative relationship between having a non-adequate level of food consumption and the different steps of subjective well-being. The effect of the log of food and non-food consumption is also in accordance with the expected; if your consumption increases you are more likely to be in the higher steps.
The individual characteristics can be interpreted in the same way as the consumption variables. Being married, for example, increases your chances to be in the higher steps of subjective well-being, this relationship has been stated in the literature overview as well. Since the gender-variable is insignificant, the negative B-coefficient will not be interpreted. But the existing literature states that women report on average a higher subjective status than male respondents (Ravallion & Lokshin, 2002).
Ta bl e 5 - O rd er ed p ro bi t re gr es si on o f su bj ec ti ve w el l-be in g in 2 00 0, In do ne si a * = sig nif ic an t at a 5% le ve l ** = sig nif ic an t at a 1% le ve l
15
Re gr essio n A Re gr essio n B Re gr essio n C Co ef fic ie nt St an da rd e rro r Co ef fic ie nt St an da rd e rro r Co ef fic ie nt St an da rd e rro r N on -a de qu at e le ve l o f fo od c on su mp ti on -1. 393 0, 029* * -1. 189 0, 029* * -1. 016 0, 031* * Ju st a de qu at e le ve l o f fo od c on su mp ti on -0. 487 0, 018* * -0. 402 0, 019* * -0. 378 0, 019* * M or e th an a de qu at e le ve l o f fo od c on su mp ti on (R ef er enc e) (R ef er enc e) (R ef er enc e) Lo g of h ou se ho ld f oo d co ns ump ti on p er c ap it a 0. 042 0, 013* * 0. 082 0, 013* * Lo g of h ou se ho ld n on -f oo d co ns ump ti on p er c ap it a 0. 281 0, 009* * 0. 271 0, 009* * G end er ( be ing m al e = 1) -0. 023 0. 015 M ar riag e st at us (b ein g m ar rie d = 1) 0. 590 0, 018* * H ou se ho ld siz e 0. 041 0, 003* * Ch ild re n (h av in g ch ild re n liv in g at h om e = 1) -0. 013 0. 017 Li vi ng a re a (i n ur ba n en vi ro nm en t = 1) 0. 031 0, 015* Emp lo yme nt s ta tu s (b ei ng e mp lo ye d = 1) -0. 055 0, 015* * N on-ad eq ua te le ve l o f he al th -0. 499 0, 043* * Jus t ad eq ua te le ve l o f he al th -0. 119 0, 037* * M or e tha n ad eq ua te le ve l o f he al th (R ef er enc e) N umb er o f ob s 25400 25307 25307 Pse ud o R2 0. 0414 0. 0707 0. 0787Tabel 6 - Marginal effects of a non-adequate level of food consumption on the subjective well-being, Indonesia
1st step 2nd step 3rd step 4th step 5th step 6th step
B C B C B C B C B C B C
Non-adequate level of
food consumption 19,40% 14,51% 24,20% 22,61% -27.33% -22,24% -15,34% -13,98% -0,79% -0,71% -0,22% -0,20%
The children variable, the dummy variable that has a value of 1 if the household has children living at home, is also insignificant and won’t be interpreted. The household size is a significant variable with a positive B-coefficient; having an extra family member will increase the likelihood of being the higher steps. The living area is significant at a higher life than the other variables, at a 5% instead of a 1% significance level. This variable states that living in an urban area rather than in a rural area will make it more likely to be positioned in one of the higher steps. This is in contrast with the effect of the living area on the food consumption adequacy, which is positively affected by living in a rural area (Lokshin, Umapathi, & Paternostro, 2006). The coefficient on employment status is unexpectedly negative; respondents that are employed are more likely to be in the lower subjective well-being steps. In the literature this conflict with expectations has been noted as well but they haven’t been able to find an explanation for the negative coefficient by adding extra employment status dummy variables (Powdthavee, 2007). The subjective health status has a similar effect on the subjective well-being as the subjective food adequacy; having a non-adequate or just adequate level of health has a negative effect whereas having a more than adequate health
situation has a positive effect.
Table 6 contains the marginal effects of the different independent variables on the different subjective well-being steps, based on regression B and C. The marginal effects of regression A are presented in table 4 in the methodology section. The marginal effects show the effect on the subjective well-being steps of adding one extra unit of an independent variable or the change of a dummy variable from 0 to 1. These marginal effects of food consumption adequacy contain the numbers that answer the research question; what is the effect of stating a non-adequate level of food consumption on the subjective well-being? The marginal effects of regression B and C are reasonably similar; but on average, adding the extra variables in regression C decreases the marginal effects. The marginal effects of all the independent variables can be found in the appendix in table 8. But the marginal effects of stating a non-adequate level of food consumption will be discussed in this section.
The results of the marginal effects are not completely in line with the hypothesis. Based on the literature, the expected was that stating that you have a non-adequate level of food
consumption would increase the chance of being in one of the lower steps and decrease the chance
of being in the higher steps of subjective well-being. In which the likelihood of being in the 1st step
with a non-adequate level of food consumption is higher than being in the 2nd step, which is higher
than being in the 3rd step, etc. Surprisingly, the data doesn’t show these results; there is
non-uniformity in the marginal effects of the dummy variable of non-adequacy of food consumption.
It is an unexpected result that stating that your food consumption is non-adequate increases the
likelihood of being in the 2nd step more than of being in the 1st step. When looking at the
descriptive statistics, one can see that the respondents in the 1st and 2nd step spend on average a
very similar amount of Rupiahs on consumption, the average monthly household consumption per person is 172758 Rupiah in the first step, and 174139 Rupiah in the second step. There is a similar
relationship between the 4th and 5th step, with respectively 367607 and 368471 Rupiah spend on
monthly household consumption. This very propinquity of the average consumption in some of the steps might be caused by subjectivity of the questions. It might be socially undesirable to say that you are in the lowest step of the welfare ladder; respondents with a similar consumption level
seem to prefer to say that they consider themselves to be in the 2nd step instead of the 1st. Within
the 4th and 5th step there is the same correspondence with the consumption but social desirability
probably doesn’t seem play a role here. The similarity might be caused by the spread of the ladders,
it is difficult for respondents to differentiate between the 4th and 5th step.
To control for these similarities, which might be the cause of the non-uniformity of food
consumption adequacy on subjective well-being, a regression has been made in which the 1st and
2nd group and the 4th and 5th group have been merged. The complete outcome of the regression can
be found in the appendix, in table 9. The R2of this regression has increased with 16% and the same
independent variables are significant, respectively insignificant. In table 7, the new marginal effects of stating to have a non-adequate level of food consumption can be found.
These new marginal effect correct for the non-uniformity of the in the lower steps of well-being, but this correction is for a great extent due to decreasing the amount of steps. The non-uniformity
in the higher steps can’t be explained by the combination of the 4th and 5th steps.
There can be concluded that the non-adequacy food consumption has a negative effect on the subjective well-being status, this relationship is not as linear as would be expected. Stating to have a non-adequate level of food consumption will increase the chance of being in the lowest two
steps with almost 40%, it decrease the chance of being in the 3rd step with almost 25%, in the 4th
and 5th step the likelihood decreases with 15% and the chance of being in the highest, richest step
decreases with 0.16%.
1st + 2nd step 3rd step 4th + 5th step 6th step Non-adequate level of food consumption 39,64% -24,45% -15,04% -0.16% Tabel 7 - Marginal effects of non-adequacy of food consumption on merged subjective well-being steps, Indonesia
6. Conclusion
This paper has focused on the relationship of two subjective being measures; the total well-being based on a sixth step ladder and the food consumption adequacy. In the existing literature the focus lies on the comparison of the objective measure and subjective measures of poverty, in which in this case the objective measure the household consumption level is. In general the objective measure and the subjective measure have quite similar results although the subjective measure has the possibility of capturing aspects of well-being which cannot be included in the objective measure. The literature on food consumption adequacy was fairly limited since it is to a lesser extent used as a poverty measure, but nonetheless some useful research has been conducted on the subject.
Using the data of the Indonesian Family Life Survey from 2000, the food consumption adequacy has been regressed on the subjective well-being with different control variables. This regression is an oprobit regression and herewith the marginal effects of the variables are measured which were necessary to test the hypothesis. The overall conclusion of the several regressions is that the relationship of stating to have a non-adequate level of food consumption and subjective well-being is negative but not completely linear or uniform. This means that stating to have a non-adequate level of food does increase your chances to be in the lower steps of subjective well-being
but you are more likely to place yourself in the 2nd step instead of the 1st step.
With these results the conclusion can be drawn that stating a non-adequate level of food consumption has the biggest effect on the lower steps of the subjective well-being, the steps where people consider themselves to be relatively poor. But the magnitude of the influence of the non-adequate statement is lower than what would be expected. This could be the case due to the average poverty level of the Indonesian population. At the moment Indonesia has passed the stage of being an underdeveloped country and can be seen as a highly growing and developing country (Dick, 2002)
Of the IFLS sample only a small 10% states to have a non-adequate level of food
consumption, this means that the other 90% can focus on other needs, such as health, security and education. Looking at Maslow’s hierarchy of needs, one might say that the majority of the
Indonesian population moved up the hierarchy by (substantially) fulfilling their food adequacy. The levels just above the basic psychological needs are the safety needs and the need for affection and love (Simons, Irwin, & Drinnien, 1987). It could be that these ‘higher’ needs play a much bigger role in determining the subjective well-being than the psychological needs in present-day Indonesia. It would be interesting to research if this hypothesis makes sense.
Unfortunately it is difficult to research how the respondents define their subjective well-being. Due to the latitude of the subjectivity questions you can capture a lot of aspects, which are not in objective measures, but this makes it also very difficult to determine the reason behind the particular chosen step on the ladder. For further research it might be better if the questions are more comprehensive; if it is better explained what is meant with subjective well-being or with an adequate level of food consumption.
7. Bibliography
Bank, W. (1986). Poverty and Hunger: Issues and Options for Food Security in Developing Countries.
World Bank Publications.
Carletto, G., & Zezza, A. (2006). Being poor, feeling poorer: combining objective and subjective measures of welfare in Albania. Journal of Development Studies, Vol. 42, No. 5, 739 - 760. Castilla, C. (2010). Subjective poverty and reference dependence, income over time, aspirations
and reference groups. UNU-WIDER Working Paper 2010/76.
Clark, A., Etilé, F., Postel-Vinay, F., Senik, C., & Van der Straeten, K. (2005). Heterogeneity in Reported Well-Being; Evidence from Twelve European Countries. The Economic Journal, Vol.
115, No. 502, 118-132.
Dick, W. (2002). The emergence of a national economy: an economic history of Indonesia,
1800-2000. University of Hawaii Press.
FAO. (2003). Measurement and Assessment of Food Deprivation and Undernutrition. Proceedings of the International Scientific Symposiun. Food and Agriculture Organization of the United
Nations.
Helliwell, J. F. (2003). How's life? Combining individual and national variables to explain subjective well-being. Economic Modelling, 20, 331-360.
Howe, L., Hargreaves, J., Ploubidis, G., De Stavola, B., & Huttly, S. (2011). Subjective measures of socio-economic position and the wealth index: a comparative analysis. Health policy and
planning, 26(3), 223-232.
Howell, R., & Howell, C. (2008). The Relation of Economic Status to Subjective Well-being in Developing Countries: a meta-analysis. Psychological Bulletin, 134 (4), 536-560.
Kingdon, G., & Knight, J. (2006). Subjective Well-Being Poverty vs Income Poverty and Capabilities Poverty? Journal of Development Studies, 42 (7), 1199-1224.
Lokshin, M., Umapathi, N., & Paternostro, S. (2006). Robustness of Subjective Welfare Analysis in a Poor Developing Country. The Journal of Development Studies, 42(4), 559-591.
Martin, K., & Hill, R. (2012). Life Satisfaction, Self-Determination, and Consumption Adequacy at the Bottom of the Pyramid. Journal of Consumer Research, 38(6), 1155-1168.
Migotto, M., Davis, B., Carletto, G., & Beegle, K. (2006). Measuring food Security Using
Respondent's Perception of Food Consumption Adequacy. Research Paper No. 2006/88,
UNU-WIDER.
Powdthavee, N. (2007). Feeling richer or poorer than others: a cross-section and panel analysis of subjective economic status in Indonesia. Asian Economic Journal, Vol 21 No.2 , 169-194. Pradhan, M., & Ravallion, M. (2000). Measuring Poverty using Qualitative Perceptions of
Consumption Adequacy. The Review of Economics and Statistics, 82(3), 462-271. Pradhan, M., & Ravallion, M. (2000). Measuring Poverty Using Qualtitative Perceptions of Consumption Adequacy. The Review of Economics and Statistics, 82(3), 462-471.
Pradhan, M., Suryahadi, A., Sumarto, S., & Pritchett, L. (2000). Measurements of Poverty in
Indonesia: 1996, 1999 and beyond. . World Bank, East Asia and Pacific Region, Environment
and Social Development Sector Unit.
RAND. (n.d.). IFLS 3 Public Release. Retrieved May 2014, from RAND: https://www.rand.org/labor/FLS/IFLS/ifls3.html
Ravallion, M. (2012). Poor, or just feeling poor? On using subjective data in measuring poverty.
Policy research working paper, no. 5968, World Bank.
Ravallion, M., & Lokshin, M. (2002). Self-rated economic welfare in Russia. European Economc
Review, 46(8), 1453-1473.
Simons, J., Irwin, D., & Drinnien, B. (1987). Psychology, the search for understanding. New York: West Publishing Company.
8. Appendix
In this appendix the model that has been used to answer the research question will be specified more extensively by describing the formula which has been used to determin the marginal effects of the dependent variables.
The model that has been used can be typed as a probit model since the dependent variable has
ordinal values. This means that the model has the following form: Y = Xβ + e with e ~ N(0, σ2),
which translates into the probability term: Pr(y*i >0 | xi ) = φ (−𝛽𝛽′𝑥𝑥𝜎𝜎 ) or in the similar term:
Pr(yi=0 | xi) = 1 – φ (−𝛽𝛽′𝑥𝑥𝜎𝜎 ). In a probit function, the independent variable doesn’t have a constant
effect on the dependent variable itself since the magnitude of a change of the dependent variable depends on the starting point. But the independent variable does have a constant effect on the derivative of the dependent variable.
In this thesis the focus lies on the relationship between a non-adequate level of food consumption and the level of subjective well-being. So the main independent variable, a non-adequate level of food consumption is a dummy variable. This makes that the marginal probability effect is
determined by the value of φ(xiβ) when x is 0 and x is 1, while the other variables stay at a
constant level. So the marginal effect of the independent variable of food adequacy is determined
by φ(x1β) – φ (x0 φ). The control variables in the model are either in the form of a dummy variable,
in which case the marginal effect is calculated with the formula mentioned above, or they are variables with continuous values. If the variables can be considered as continuous; their marginal effect on the dependent variable, subjective well-being, can be determined with a different
formula. This formula has the following form: Xi = 𝜕𝜕 Pr( 𝑌𝑌=1)𝜕𝜕 𝑋𝑋𝑋𝑋𝑋𝑋
=
𝜕𝜕 𝜑𝜑 (𝑋𝑋𝛽𝛽)𝜕𝜕 𝑋𝑋𝑋𝑋𝑋𝑋,
which can be divided into athe value of the standard normal probability density function and the marginal index effect of x. The marginal index effect is the effect of an independent variable on the total probit function.
Ta ble 8 - M ar gi na l e ff ec ts o f in de pe nd en t va ri ab le o n th e di ff er en t st ep s of s ub je ct iv e w el l-be in g
22
1st st ep 2n d st ep 3r d st ep 4th step 5th step 6th step B C B C B C B C B C B C No n-ad eq ua te le ve l o f f oo d con su mp tion 19. 40% 14. 51% 24. 20% 22. 61% -27. 33% -22. 24% -15. 34% -13. 98% -0. 79% -0. 71% -0. 22% -0. 20% Ju st ad eq ua te le ve l o f f oo d co ns ump tio n 2. 46% 2. 22% 9. 00% 8. 57% -1. 01% -1. 15% -9. 32% -8. 67% -0. 81% -0. 72% -0. 27% -0. 24% Mo re th an a de qu at e le ve l o f f oo d co ns ump tio n Lo g of h ou se ho ld fo od co nsu m pt io n pe r c ap ita -0. 30% -0. 56% -0. 10% -1. 94% 0. 30% 0. 59% 0. 91% 1. 76% 0. 07% 0. 13% 0. 02% 0. 04% Lo g of h ou se ho ld n on -f oo d co nsu m pt io n pe r c ap ita -2. 02% -1. 86% -6. 62% -6. 45% 1. 98% 1. 95% 6. 06% 5. 81% 0. 45% 0. 42% 0. 14% 0. 13% Ge nd er (b ei ng m al e = 1) 0. 15% 0. 54% -0. 16% -4. 82% -0. 03% -0. 01% Mar riag e st at us (b ein g m ar rie d = 1) -0. 41% -1. 42% 0. 45% 1. 26% 0. 09% 0. 03% Ho use ho ld siz e -0. 28% -0. 98% 0. 30% 0. 89% 0. 06% 0. 02% Ch ild re n (h av in g ch ild re n liv in g at h om e = 1) -Liv in g ar ea (in u rb an en vi ro nm en t = 1 ) -0. 21% -0. 73% 0. 22% 0. 65% 0. 05% 0. 01% Em pl oy m en t st at us (b ein g em pl oy ed = 1) 0. 37% 1. 31% -0. 38% -1. 19% -0. 08% -0. 03% No n-ad eq ua te le ve l o f he al th 4. 90% 12. 22% -7. 69% -8. 78% -0. 51% -0. 14% Jus t a de qua te le ve l o f he al th 0. 76% 2. 79% -0. 62% -2. 65% -0. 20% -0. 07% Mo re tha n ad eq ua te le ve l o f he al thTable 9 - Ordered probit regression of subjective well-being with merged steps
Regression D
Coefficient Standard error
Non-adequate level of food consumption -1,083 0,033**
Just adequate level of food consumption -0,380 0,020**
More than adequate level of food consumption (Reference)
Log of household food consumption per capita 0,081 0,014**
Log of household non-food consumption per capita 0,289 0,010**
Gender (being male = 1) -0,023 0,015
Marriage status (being married = 1) 0,550 0,019**
Household size 0,043 0,004**
Children (having children living at home = 1) -0,006 0,018
Living area (in urban environment = 1) 0,055 0,016*
Employment status (being employed = 1) -0,064 0,016**
Non-adequate level of health -0,489 0,045**
Just adequate level of health -0,116 0,038**
More than adequate level of health (Reference)
Number of obs 25296 Pseudo R2 0.0915 * = significant at a 5% level ** = significant at a 1% level
