Keywords

1 Introduction

Happy families are all alike; every unhappy family is unhappy in its own way.

Leo Tolstoy-Anna Karenina

Since ancient Greece, the quality of life or the conditions for happiness has been widely discussed to improve welfare of community. While there are more than one definitions of well-being in the literature that make an individual happy, there is no consensus among these definitions. Consequently, measurement of well-being changes according to different opinions. As a result, different terms are used to define subjective well-being. These include quality of life, happiness and satisfaction. There are vast of studies on subjective well-being in order to understand needs for a good life. While the majority of the studies in the literature aim to determine the factors determining subjective well-being, there are a limited number of studies investigating the effect of social assistance on subjective well-being. Social assistance provides social security to individuals who are lack of contribution to welfare system. In this context, individual social assistance is provided in the form of monetary, in-kind or service, while the financing is provided by the state.

Killburn et al. (2018) analyzed the short-term impact of external positive income shock on caregivers' subjective well-being using a 17 month follow-up panel of 3365 households included in the Malawi outreach cash transfer program for extremely poor and labor-limited households. According to the results of the analysis, caregivers living in households receiving social assistance reported higher life satisfaction and tend to look more confidently into the future. However, the impact of social assistance on individuals’ welfare in Turkey was investigated qualitatively by social scientists and the results of the studies have drawn attention to the problems on the welfare system. Güneş (2012) identified problem areas that prevented the continuation of social assistance. These are sustainability of social assistance, delivery of social assistance, increasing demand for social assistance, dependence of social assistance in the poor and social assistance in the form of political patronage. Kutlu (2015) investigated social assistance in Turkey in terms of the nature of social rights. His findings are based on field research study. In the study, it was revealed that beneficiaries of social welfare have lack knowledge about social assistance and could be involved in social assistance processes by their political relations. Kutlu also is a point to other problems which are the way of distribution of social assistance, the quality of the food assistance, uncertainties in the sustainability of social assistance and realization of the demand for social assistance in the form of social injustice. Taşçı (2017) mentioned three main concerns which were danger of stigmatization and humiliation of social assistance beneficiaries, addiction to laziness and exploit of social assistance. Karadoğan (2018) emphasized that social assistance contributes to poverty sustainability and that poverty is internalized through social assistance. According to the results of the study, social assistance which is not perceived as a social right affects the self-esteem level of individuals.

The above mentioned studies revealed the necessity to investigate quantitatively the effect of social assistance on individuals’ quality of life in Turkey. For this purpose, this study, which is an effort to understand how the poor live in poverty in the transforming world, has taken a holistic approach by making use of the quantitative data to reach scientific conclusions and discussing results by utilizing the shared data of the qualitative studies mentioned above. In this way, not only the quantitative data was used to describe the society, but also the consistency of the information obtained through quantitative data was compared with qualitative observations. Thus, instead of preferring one method to another, the advantages of the two were utilized. In this context, the results obtained based on micro data can be used to determine policy at macro level. In order to achieve this, partial proportional odds model was used in order to take into account ordered structure of dependent variable. Using this model is an attempt to make a contribution to use of ordinal utility theory to the possibility and the relevance of interpersonal comparisons for policy recommendations on welfare system.

The study is organized as follows. Section 2 presents the social assistance system in Turkey. Section 3 discusses the methodology. Section 4 summarizes data and presents the empirical results of the model which shows effect of social assistance on subjective well-being. Last section concludes.

2 Social Assistance System in Turkey

The social security system is a security system that provides protection against the physiological risks (illness, disability, old age, maternity, accidents and death) and economic risks (insufficient family income, unemployment) that individuals may face in life. These physiological and economic risks can lead to decrease in income, income cut or increases in expenses. Owing to the social security system, a minimum guarantee is provided for the person to fight against these risks (İzgi 2008). This guarantee is provided based on the paid premiums (social insurance) or non-contributory payments financed by taxes (public social security expenditure) (Akar 2015). The social assistance system is provided by non-contributory payments. The primary aim of social assistance system is to ensure social security to individuals who are lack of premium payment power. In this context, individual social assistance is provided in the form of monetary, in-kind or service, while the financing is provided by the state. Social assistance is distributed by Social Assistance and Solidarity Foundation with Law No. 3294 on the Encouragement of Social Assistance and Solidarity in Turkey since 1986 (Alper 2017).

Social assistance in Turkey can be divided into six main groups which are family assistance, shelter and food assistance, education assistance, health assistance, assistance for disabled person and old people, employment assistance and assistance for special purposes. The primary objective of these social assistance categories is to alleviate poverty by encouraging social justice. However, each of them achieves this goal using different tools. Family assistance provides a wide range of assistance from food needs to accommodation needs. In addition, this aid covers different types of households. These are military families in need, parentless children and widowed women. Although family assistance covers the need for shelter and food for those in need, shelter-food assistance to the benefit of households living in Turkey are also available. Shelter-food assistance provides electricity consumption support to households in need benefiting from social assistance programs. Also under social cohesion assistance which is part of shelter-food assistance, monthly payments are made to foreigners with temporary protection status, international protection status, international protection applicants and humanitarian residence permits. Education assistance contains free textbook distribution to the primary and secondary school students; free lunch for poor students who move to centers where schools are located within the mobile education; food and accommodation help for primary and secondary school students outside the mobile education system; conditional education aid to families who do not have social security and who are in need within the scope of Law No. 3294 for their children’s formal education on the condition that children do not have more than four days of absence in a month; dormitory construction for secondary school students and also transportation aid for students who need special education. Health assistance is provided to help people in need to meet their health needs. Some of the health assistance are premium payments to individuals without social security; providing all kinds of tools and equipment for disabled citizens in need; electricity consumption support for patients who are dependent on the device due to chronic illness; health assistance provided that families in need on the condition that send their children to health control; conditional health assistance for pregnant women in need provided that they have health checks and births in hospital; regular cash benefits for patients who experience psycho-social and financial loss due to tuberculosis and subacute sclerosing panencephalitis disease. In addition to covering the needs of the disabled person and old people within the scope of health assistance, cash assistance is also provided within the scope of assistance for them. As part of this assistance, monthly payments are also made to relatives of disabled people under the age of 18. In addition to these assistance programs, orientation and start-up assistance to a job for individuals who are able to work between the ages of 18–55 living in households benefiting from social assistance programs are provided within the scope of employment assistance. Apart from these aids, there are assistance programs for special purposes such as soup kitchens in poor neighborhoods, disaster and emergency aids, terrorist damage aids. According to the latest activity reports of the Republic of Turkey Ministry of Family, Labor and Social Services, social assistance in figures is shown in Table 1.

Table 1 Overview of social assistance in Turkey
Full size table

Regular social assistance programs excluding general health insurance premium support are conditional education assistance, conditional health assistance, conditional pregnancy assistance, assistance for widowed women, assistance for military families in need, assistance under Law No. 2022 and home care assistance. However, temporary assistance programs which are food assistance, fuel assistance, accommodation assistance, education assistance, health assistance, disability needs assistance, special purpose assistances, clothing and other family assistance, employment assistance are in the form of one-time assistance. As seen in Table 1, compared to 2014, the share of social assistance in the gross domestic product increased by 0.02% in 2016. The number of households benefiting from regular assistance programs in 2016 increased by 68,764 compared to 2014, while the number of staff responsible working in Social Assistance and Solidarity Foundation for the distribution of these aids increased by 447 and the number of Social Assistance and Inspection staff working in Social Assistance and Solidarity Foundation increased by 47. In addition, in 2016, the rate of individuals whose daily expenditure per capita was below 4.3 USD in purchasing power parity decreased by 0.48% compared to 2014. Therefore, in this study, housing aid, in-cash and in-kind social assistance, cash and in-kind child assistance were taken as social assistance and the effects of these on the subjective well-being were investigated using partial proportional odds model.

3 Empirical Methodology

Models for ordinal outcomes differ according to whether the distance between the categories is equal or not. Therefore, firstly, in the study, the parallel lines assumption was tested in order to test the equality of the distance between the categories of the outcome variable. Brant test was used for this purpose. With respect to the results of Brant test, some variables violate the parallel lines assumption. Therefore, the partial proportional odds model was used due to the ordered nature of the outcome variable.

Brant test is a Wald test that includes individual tests that show by which variables the assumption of parallel lines is violated. For this reason, firstly, j-1 binary logit was created for the outcome variable having J-category. The parallel lines hypothesis was tested as a result of the comparison of these binary logits (Long and Freese 2014). In the study, Eq. (1) was used to test parallel lines assumption for the outcome variable with 11 category.

$$\mathrm{Pr}\left({\text{SWB}}\le j|\mathbf{x}\right)=\mathrm{F}\left({\alpha }_{j}-\mathbf{x}{{\varvec{\upbeta}}}_{\mathrm{j}}\right) {\text{for}} j=\mathrm{0,1},\dots ,\mathrm{9,10}$$
(1)

where SWB is the level of subject well-being which is the outcome variable with 11 category. The null hypothesis that all coefficients are jointly zero which indicates parallel lines assumption holds. Alternative hypothesis shows that βs differ across binary logit comparisons. According to the test results, it was revealed that some variables violated the parallel lines assumption. Partial proportional odds model was used to solve this problem. This model allows the effect of the predictors that violate the parallel lines assumption to vary across all categories of the ordered outcome variable. The original partial proportional model proposed by Peterson and Harrell (1990) reconstructs the data and determines the interaction between the explanatory variables that violate the parallel lines assumption and the different categories of the ordered dependent variable. The partial proportional odds model proposed by Williams (2006) alleviates the proportional odds assumption by allowing the effect of each explanatory variable to vary across different cut points of the ordered outcome variable without reconstructing the data. The model used can be written as in Eq. (2) which follows methodology of Williams (2006).

$$\mathrm{Pr}\left({SWB}_{i}>j\right)=\frac{\mathrm{exp}({\alpha }_{j}{+XM}_{i}\beta M+{XK}_{i}{\beta K}_{j})}{1+\left\{\mathrm{exp}(\mathrm{exp}\left({\alpha }_{j}{+XM}_{i}\beta M+{XK}_{i}{\beta K}_{j}\right)\right\}}$$
(2)

where i shows number of individuals changes between 1 and 20,820; \({\alpha }_{j}\) represents cut points of the model; XM indicates explanatory variables which do not violate parallel lines assumption in the model, M = 1, 2, 3, 4; XK shows explanatory variables which violate parallel lines assumption in the model, K = 5, 6, 7, 8, 9; βs are logit coefficients; j denotes the category of outcome variable, j = 0, 1, 2, …, 10. This model uses maximum likelihood estimation.

4 Data and Results

4.1 Data

TURKSTAT 2013 Income and Living Conditions Survey was used in the study. It has information about 12 regions of Turkey according to NUTS-1. Also, it is the latest data set which collects data on individual’s subjective well-being and detailed income resources. The data set consists of 33,755 observations. It contains information about the income of 20,820 individuals. Therefore, descriptive analysis covers the whole data set, while the predicted models cover only those who report their income. In the analyzes, in order to investigate the effect of being a beneficiary of social assistance on subjective well-being, individuals receiving child benefits in-cash and in-kind, individuals receiving housing allowance and individuals receiving social assistance in cash and in-kind were taken into account. An individual who receives any of these benefits was named as social assistance beneficiary and variable which indicates social assistance beneficiary received a value of 1 if individual receives any of these benefits. It is a binary variable, takes the value 0 for individuals who are not beneficiaries of social assistance. The variables used in the study and their definitions are given in Table 2.

Table 2 Variables used in the study
Full size table

The data set is 44.43% male, 73.80% married, 43.42% household head and 20.93% live in the eastern region. The level of education completed is distributed as follows: 12.60% are illiterate, 8.27% are literate but do not complete a school, 36.64% are primary school graduates, 15.06% are secondary school or equivalent, 8.27% are general high school graduates, 7.43% were technical or vocational high school graduates and 11.73% completed higher education. The age ranges from 15 to 110, and the average age is 43. 5.38% of individuals in the data set are poor. Here, 1.004 TL was used as the equivalent of 1 US dollar in terms of purchasing power parity to determine the poor individuals. Accordingly, poor individuals are determined as those whose income is below 4.3 US dollars according to the daily purchasing power parity. 15.61% of the individuals in the data set are social assistance beneficiaries. 5.70% of the individuals in the data set are not satisfied with their lives at all, however, 7.34% are very satisfied with their lives. To the question about general life satisfaction, 26.35% of the data set answered not bad. This category is the most preferred category among the answers. Average satisfaction value differences were investigated according to poverty status that might be effective in analyzing subjective well-being in terms of social policy. Average satisfaction levels according to poverty are shown in Table 3.

Table 3 Average satisfaction levels by poverty status
Full size table

According to Table 3, individuals who are not poor are generally 8% more satisfied with their lives. Based on this result, quantiles were created to investigate how the subjective well-being of individuals is distributed according to income. For this purpose, first the amount of income is ranked from the lowest to the highest, and then the listed income amount is divided into five equal groups. Table 4 shows the distribution of subjective well-being by income quantiles. According to Table 4, 38.32% of the participants did not report their income. While the group with the lowest income level expressed their subjective well-being level as above not bad (5 points+) at a rate of 39.92%, this rate is 64.89% in individuals with the highest income level. This situation shows the importance of the effect of income status on subjective well-being.

Table 4 Distribution of subjective well-being by income quantiles
Full size table

4.2 Results

In the analysis, the partial proportional odds model was used to investigate the effect of social assistance on subjective well-being. In the estimated models, observations of 20,820 people who reported their income were used. As the explanatory variables of the predicted models, relative income, gender, age, marital status, educational level and health status, which have been widely researched in the social assistance and subjective well-being literature, are used. Despite the frequently encountered problem of heteroscedasticity in cross-section data, robust standard errors were calculated in the models. Since the estimation of ordered regression models is based on the assumption of proportional odds ratio, in other words, the assumption of parallel lines, the Brant test was applied in order to decide which ordered regression model should be used (Table 5).

Table 5 Results of Brant test
Full size table

According to the individual Brant test results, the null hypothesis of the parallel lines assumption for the variables of gender, age, age square and marital status could not be rejected. It was concluded that the parameter estimates pass through the same cut-off point. Therefore, it was estimated using a partial proportional odds model, which takes into account that some estimators provide the assumption of parallel lines and some do not. Partial proportional odds model results were given in Tables 6 and 7. Since the log-pseudo likelihood value is statistically significant in the estimated partial proportional odds model, it has been concluded that the model with all independent variables is significant (Wald \({\chi }_{54}^{2}\) = 2089.43, p < 0.001). Since the parallel lines assumption is provided for the variables of gender, age, age square and marital status in the model, the estimators take the same value for each category of the dependent variable. While the head of the household variable was found statistically insignificant in Model 0, Model 1 and Model 2; it was found statistically significant in Model 3 and Model 4 at 1% significance level; similarly significant in Model 5, Model 6, Model 7 and Model 8 at 0.1% significance level and at 5% significance level significant in Model 9. Throughout binary models, the estimated coefficient of the household head variable is in line with expectations and negative. Being the head of the household decreases the odds ratio of being very happy by 14.94% compared to the probability of being in other categories. The variable included in the model in order to investigate the effect of being a social assistance beneficiary on the subjective well-being was found statistically significant at the 0.1% significance level across all binary models. Being a beneficiary of social assistance decreases the odds ratio of having the highest level of subjective well-being by 43.78% compared to the probability of being in other categories. However, a one-unit increase in relative income increases the odds ratio of having better subjective well-being by 10.31%. Education level variable is insignificant in Model 8; significant in Model 0 at 5% significance level; in Model 1 and Model 9 at 1% significance level and in remaining models significant at 0.1% significance level. Education level has a positive effect on subjective well-being. In order to observe the effect of changes in education variable on subjective well-being, the probability of being in categories 0, 3, 5, 7 and 10 of subjective well-being was estimated and shown in Fig. 1.

Table 6 Results of partial proportional odds model
Full size table
Table 7 Results of partial proportional odds model
Full size table
Fig. 1
figure 1

The relationship between subjective well-being categories and educational status with adjusted predictions with %95 confidence intervals

Full size image

According to the results, while the probability of subjective well-being in the 0, 3, 5 and 10 categories of those who have completed college, faculty or higher education decreased, the probability of being in the 7th category increased. The increase in the education level of the individual decreases the odds ratio of having the highest level of subjective well-being by 4.75% compared to the probability of being in other categories. However, an increase in an individual's education level increases the probability of being in the 8th and 9th categories of subjective well-being by 0.24% compared to the probability of being in other categories. The health status variable was found significant at the 0.1% significance level in all models, and the sign of the variable was positive in accordance with the expectations. The increase in the health status of the individual increases the probability of being in the highest subjective well-being level by 69.33% compared to the probability of being in other categories. Similarly, the variable of being married was found statistically significant at the 0.1% significance level, and it was found that having the highest subjective well-being compared to the probability of being married in other categories increased odds ratio by 28.28%.

5 Conclusions

Since the 1980s, the reduction of the role of the state in the social field has been discussed in developed countries where privatizations, free movement of goods, technology and information, deregulation in competition and neoliberal economic policies are effective in the economic area of the state. This debate stems from the fact that the current social welfare state appears to be an economic burden after the economic crises in the globalizing world under the influence of monetarism. In contrast, social policy applications in Turkey are increasing. Although neoliberal policies are followed in the economy, the interventions of the state to reduce poverty have been increasing in the last decade. These interventions are carried out by the Social Assistance and Solidarity Foundations, which work like a non-governmental organization. However, this situation creates some problems. These problems can be divided into two groups which are individual problems and practical problems. The problems that social assistance can create for the individual are stigma and humiliation, addiction and laziness (Taşçı 2017). On the other hand, the problems that social assistance can create in terms of practice are the recreation of poverty as a result of clientelism, injustice in the form of distribution of social assistance, injustice in the delivery of social assistance, unfairness in the quality of social assistance, unfairness in the continuity of social assistance and increase in demand for social assistance (Kutlu 2015). However, an important point to be noted here is that the problems in terms of practice can also create problems for the individual. In this context, the study investigated how the poverty experienced by the poor in Turkey by using partial proportional odds model. For this purpose, TURKSTAT 2013 income and living conditions which is the latest data set which collects data on individual’s subjective well-being and detailed income resources were used. It is also the limitation of the study. Because, the data of 2013 may imperfectly conform with 2021. However, it is a worthy attempt to look at the subject deeply. According to the partial proportional odds model results, the variable included in the model to investigate the effect of being a social assistance beneficiary on subjective well-being was found statistically significant at 0.1% significance level across all binary models. The fact that an individual is a beneficiary of social assistance decreases the odds ratio of having the highest level of subjective well-being by 43.78% compared to the probability of being in other categories. This situation may be evidence of welfare stigma.

Welfare stigma is an important concept in social management research and is defined as a central problem (Pinker 1971). Stigma marks the person who benefits from prosperity, damages his reputation. In addition, stigma can become a barrier that can prevent the person from accessing social services and an experience that can make them feel degraded. Pinker (1971) used the phrase “it is the most common form of violence in democratic societies” about stigmatization. Stigma is associated with two fundamental problems of social welfare. The first of these is the quality of the services provided. In the quality of services, the attitudes of those who deliver the aid can be humiliating (Spicker 2011). The following participants’ expressions in Kutlu (2015) study show the situations that beneficiaries may encounter in service quality:

Participant 6:

What do you not encounter in the aid distribution. (…)I don't know, they read names, chaos, crowd, you get it hard, even your food is stolen.(…) It also happens, screaming and calling. We take it under difficult conditions. People are accumulating early in the morning, they are waiting, the truck will come, they will read the list, you are chasing the truck, you are looking for your name. It gives; but it disgraces.

Participant 7:

The people do not stop, they do not wait, the people do not stop, as if there is fear in the people. Since They are worried that their food will go, They will not be able to replace it, it will not come again, They will not be able to buy it, the lame comes there, the blind come there, they come in a misery.

The second problem, closely related to the first problem, is the effect of services on demand. The statements given by the participants of the field study in Kutlu (2018) study show this situation.

Participant Gülseren made the following statements regarding the aids (Kutlu 2018).

It's not a good thing to get help, you can't go, when you were ill and go there to get aid, you feel awkward there, you feel like you have lowered yourself a little bit from him. You are human too, I am also human. They offend me, their words are heavy, they either say a word to you, or they say something. You can't respond as it should.

Participant Muhsin, who did not want to get in the bread queue and buy bread, used the following statements (Kutlu 2018).

No bro. I find two bread wherever they are, don't get me wrong, Let me tell you as an example, I tell you bro that I'll buy two bread from here, I take it for the sake of God and go to my house. I can't get up every day and get in line there. I'm looking at those in the queue, all women, not one man. How can I see myself like that among them? No way, I can't fit myself. That's why I never thought of it anyway.

The statements based on field research in qualitative studies confirm that social assistance can have a negative effect on the subjective well-being, as the analysis results show. This is why the study is very important in terms of providing quantitative evidence how poverty is felt in Turkey. Another important feature of the study is that the study used nationwide representative data set. In this way, we can use the analysis results to make a policy recommendations.

In terms of other variables used in the model, the estimation results are consistent with the subjective well-being literature. The subjective well-being studies in recent years show that relative income is effective on subjective well-being. Here, relativity is that one's income depends on one's expectations, habits, and social comparisons (Diener et al. 1993). According to analysis results, there is a positive relationship between relative income and subjective well-being. Another positive relationship was found between the individual's being married and his subjective well-being. This result confirms the protection/support hypothesis of Coombs (1991), which shows that single individuals who do not have an ongoing relationship with a spouse providing emotional and financial support have more difficulty. Another important result of the study is that the subjective well-being of the person increases as the health condition improves.

The study aims to improve impact of the social policy by providing a basis for further research. It also suggests an integrated approach to social studies which include quantitative and qualitative methods together. Next steps will comprise collect a panel data set to test main hypothesis of the study in order to take into account individual effects and time effects together.