The data were collected between July and September 2011. First, we requested statistical information from TUIK (Turkish Statistical Institute) regarding the development levels of streets and districts to identify different income groups in Turkey’s three largest cities. Based on this information, three groups of streets and districts were identified for each city in terms of three development levels: high, medium, and low.
The number of the questionnaires required for each city was then determined by calculating the proportion of each city’s population as a percentage of the total population of the three cities. The questionnaires were distributed equally to the three income levels (high, middle, and low) in each city. The required minimum sample size was calculated as 1,067 with a margin of error of 3% and a 95% confidence interval. Ultimately, 1,110 questionnaires were distributed.
Based on the information from TUIK, questionnaires were targeted specifically at high-income households in high-developed areas, middle-income households in medium-developed areas, and low-income households in low-developed areas, with the restriction that no more than questionnaires could be distributed in each building.
Apart from the TUIK data, information about potential participants was also gathered from each district’s local official (in Turkish, muhtar) and by observing each building’s appearance. Finally, informal interviews were conducted with some participants.
The questionnaires were distributed equally to high-developed, medium-developed, and low- developed districts in each city. In total, 678, 249, and 183 questionnaires were distributed in Istanbul, Ankara, and Izmir, respectively, making an overall total of 1,110, as shown in Table 1.
Regarding the independent variables, the following socio-economic and subjective factors were included, based on the literature: education, gender, political belief, age, marital status, perception of class, perception of destiny, and subjective poverty. Table 2 summarizes descriptive statistics for the sample:
Regarding perceptions of poverty, 52.4% of low- income respondents consider themselves as poor whereas 6.8% of middle-income people and 0.3% of low-income respondents consider themselves as poor. No low-income respondents consider themselves as rich whereas 32.7% of high-income respondents do. While 37.8% of low-income respondents consider themselves as neither poor nor rich, 90.8% of middle-income respondents consider themselves as neither poor nor rich and 65.7% of high-income respondents consider themselves as neither poor nor rich. Fewer than half of high-income respondents consider themselves as rich.
Thus, perceptions of poverty vary with income status. Specifically, as income status changes from low to middle, more respondents consider themselves as neither poor nor rich. However, as income status changes from middle to high, fewer respondents consider themselves as neither poor nor rich.
Regarding class perceptions, 23% of low-income respondents, 73.8% of middle-income respondents, and 57.3% of high-income respondents consider themselves as middle class. Very few respondents consider themselves as high class: only 3.8% of even high-income respondents consider themselves as high class.
Regarding destiny in life, 15.4% of low-income respondents, 11.6% of middle-income respondents, and 6.8% of high-income respondents believe that destiny determines everything in life. In other words, high-income respondents are less likely than low-income and middle-income respondents to believe that destiny determines everything in life. In contrast, 22% of low-income respondents, 29.5% of middle- income respondents, and 33.9% of high-income respondents believe that respondents create their own destiny. In other words, high-income respondents are more likely than low-income and middle-income respondents to believe that people determine their own destiny.
Regarding education level, as expected, the high-income group includes more college graduates than the other groups since education level is positively correlated with income, although we do not investigate the direction of causation. More specifically, 66.4% of low-income respondents (4.1% illiterate + 4.3% only literate + 39.8% primary school + 18.2% secondary school) have at most a secondary-school education level.
Regarding age, respondents aged at least 66 years comprised 4.1% of the low-income group but only 1.6 and 0.8% of the middle-income and high-income groups, respectively. This shows that the low-income group has the most old-aged respondents.
To examine poverty attributions, two questions were used to assess the respondents’ preference for each approach (individualistic, structural, and fatalistic). The participants could answer as many questions as they wanted. Table 3 presents the descriptive statistics.
As Table 3 shows, the preferred approach to making poverty attributions varies across income groups. For example, 20.3% of low-income respondents agree that “the poor are poor because they are unlucky” compared to 14.1% of middle-income and 17.8% of high-income respondents. Conversely, 23.5% of low-income respondents agree that “the poor are poor because of their fate” compared to 11.4% of middle-income and 8.9% of high-income respondents. Thus, fewer high-income respondents than middle and low-income respondents support the fatalistic approach. Both the individualistic and approaches were supported by fewer low-income than middle- and high-income respondents.
Most participants prefer structural explanations, with 60% agreeing that poverty derives from income inequality and 46% attributing poverty to a lack of basic services. In contrast, fewer participants preferred individualistic explanations, although a significant proportion (35%) agree that laziness can explain poverty while 14% agree that poverty results from “lack of skill, intelligence, and talent”.
The fourth and fifth most frequently selected explanations were fatalistic: “being unlucky” (17%) and “fate” (15%). Overall, the distribution of responses reflects the dominance of structural attributions of poverty among the participants, followed by individualistic and fatalistic attributions.
In order to investigate the effect of political outlook on poverty attributions, we classify political views into three broad categories: right-wing, left-wing, and neither left nor right. The participants’ political position was measured using the following question: “How do you define yourself in terms of political stance?” Respondents could select from nine response categories that are commonly used in Turkish politics, particularly by commentators during election campaigns: conservative, modern conservative, conservative democrat, liberal, social democrat, socialist, nationalist, others, or “none of above”.
For the empirical analysis, a respondent was defined as right-wing if they selected conservative, modern conservative, conservative democrat, liberal, or nationalist and left-wing if they selected social democrats and socialists. Missing responses were coded as “other” and categorized as “neither left nor right” along with “none of above” responses.
As noted above, respondents who selected “liberal” were included in the political right category. The term “liberal” has multiple meanings in Turkey, although it generally refers to the political right because it is associated with less government intervention and support for the market economy. In contrast, it refers to the political left, especially in the USA, mainly regarding culture.
Of the 1,057 respondents, 54 identified as liberal (*** %), alongside 305 left-wing (28.8%), 523 right-wing (49.48%), and 229 (21.66%) respondents who were neither left nor right. This distribution is consistent with the vote shares of right- and left-wing parties in many of Turkey’s national parliamentary elections, which indicates that the sample is representative of the whole population in terms of political outlook.
The following section presents the regression analyses of the determinants of different poverty attributions.
We conducted two sets of regressions, as presented in Tables 4 and 5. The dependent variable in the first set was a binary yes/no choice, coded as 1 for yes and 0 for no responses to questions about poverty attribution. For each approach to explaining poverty (structuralist, individualistic, and fatalistic), there were two statements, as presented in Table 3 in the previous section. If the respondent answered yes to either one or both questions in each pair, the dependent variable took the value of 1 for that group whereas if the respondent answered no to both questions, the variable took the value of 0. For these types of dependent variable, binary choice regression analysis is appropriate. We therefore used a probit method to estimate the coefficients of independent variables.
For the second set of regressions, the dependent variable was an ordered choice variable. Respondents answered the questions using a Likert-type scale from 1 to 5, with 1 labelled as “not important at all” and 5 as “very important”. The dependent variables in the second set of regressions also included policy variables about reducing poverty. The respondents stated their preference for each policy by assigning any number from 1 to 5. Since the dependent variable is an ordered response, ordered probit regression analysis was used to estimate the coefficients of the independent variables.
As Table 4 shows, people who define themselves neither left nor right or right-wing political do not believe that poverty results from income inequality or lack of basic services. That is, they do not support the structuralist view. Therefore, Hypothesis One was confirmed.
Regarding the socio-economic variables, the coefficients for gender, age, and marital status were not significant, indicating that support for the structuralist view is not affected by these variables. Literate people and elementary, middle, and high school graduates were more likely than respondents with university and graduate degrees to reject structuralist explanations of poverty. Annual per capita income had a positive and significant coefficient. That is, respondents with higher incomes (not perceived income) were more likely to agree with structuralist explanations, specifically income inequalities and inadequate access to basic services. Conversely, respondents with high perceived income defined themselves a high class and rejected structuralist explanations. Thus, Hypothesis Two was also confirmed. Respondents who said that destiny was very important in their life also did not agree with the structuralist view.
Regarding the individualistic approach, our regression results show that it is not preferred by women, respondents with high actual annual income (not perceived income), illiterate respondents, and respondents who consider destiny to be very important. That is, they do not think that the poor are poor because they are lazy or lack skill, intelligence, and talent. In contrast, the individualistic approach is preferred by respondents who define themselves as neither left nor right, or identify as right wing, respondents who define themselves as upper-middle or high class, and describe themselves as rich. That is, they tend to think that poor are poor because they are lazy, or lack skill, intelligence, and talent. These findings confirm Hypotheses Three and Four.
Regarding the fatalistic approach, the regression results indicate support for this approach from lower-class respondents, respondents who self-report as poor, respondents who consider that destiny is very important in their life, and married respondents. That is, they think that poor are poor because they are unlucky, and that poverty is their fate. In contrast, older respondents, and high-income respondents (not perceived income) disagree with the fatalistic approach.
It is important to consider endogeneity in our regression analysis as there could be reverse causality from the dependent variable (poverty attribution) to political views. However, we believe that this reverse causality is weak since almost 60% of respondents agree with structuralist poverty attributions although only about 28% of respondents defined themselves as left wing. In other words, while most left-wing people believe that poverty has structuralist causes, not all the people who think like this are left wing.
We further investigated endogeneity by designing an online experimental survey conducted with a more educationally homogeneous group of undergraduate and master’s economics students. Out of 800 students approached, 81 responded to the survey, of whom 72 (88.88%) think that poverty is caused by income inequality. Of these 72 students, 39 students defined themselves as centrist or right wing while 33 students defined themselves as left wing. Given that 36 students out of the full sample of 81 defined themselves as left wing, almost all left-wing students (33 out of 36) think that poverty is caused by income inequality, although not all students who think that poverty is caused by structuralist factors are left wing. These results thus indicate weak endogeneity in our regression results.
Table 5 presents the results of the second set of regressions.
As Table 5 shows, people with structuralist views think that policies guaranteeing minimum income should be implemented along with better income distribution, vocational courses, free education and health services, and new job creation. Conversely, these respondents do not believe that poverty can be reduced by association and foundation aid, coal aid, and only community aid.
In contrast, respondents with a fatalistic view agree that poverty can be reduced by increasing association and foundation aid, coal aid and community aid can reduce the poverty. As usually low-income respondents, believe instead that poverty can be reduced with the help of the people around them (community aid, association and foundation aid) and coal aid from the government, especially before national or local elections.
Respondents with an individualistic view think that policies for new job creation, vocational courses, and policies for better income distribution can help poverty reduction whereas association and foundation aid cannot.
Table 5 indicates that Hypothesis Five is confirmed since respondents with individualistic or structuralist views support some of the same policies, such as new job creation, vocational courses, and better income distribution. Comparing the fatalistic and structuralist views highlights clear policy preference differences. More specifically, only the former agrees that association and foundation aid, coal aid, and community aid can reduce poverty.