Ethnic Diversity and Public Goods Provision in Turkey
The relationship between ethnic diversity and publicly provided goods, on the sub-national level in Turkey. Comparison of socio-economic indicators of the development of the Kurdish provinces and other regions. Efficiency of investment in education.
Рубрика | Экономика и экономическая теория |
Вид | дипломная работа |
Язык | английский |
Дата добавления | 24.08.2017 |
Размер файла | 291,7 K |
Отправить свою хорошую работу в базу знаний просто. Используйте форму, расположенную ниже
Студенты, аспиранты, молодые ученые, использующие базу знаний в своей учебе и работе, будут вам очень благодарны.
Graph 3 Average number of Hospitals per capita Data source: TURKSTAT (2016)
In general, the number of hospitals per 100,000 inhabitants was increasing. However, there was a decline in hospitals per capita from 2010 to 2011. The sharpest increase happened after 2011, where the average number of hospitals rose to 0.248 per million people. Hospital beds have been steadily increasing too. Therefore, we can say that the boost in economic growth was followed by the growth in the number of public hospitals.
I calculated changes in per capita number of hospitals, beds, physicians, and nurses in the year 2000. Data was collected from the Turkish National Statistical Institute (TURKSTAT, 2016). Therefore, from the observed period, I subtracted a number of hospitals, beds, physicians, and nurses in the year 2000. These variables will be used to analyze the relationship between changes in the public goods provision in health, and the ethnic and economic characteristics of provinces.
Graph 4 Average Changes in the Number of Nurses Data source: TURKSTAT (2016)
Chart 4 plots the changes in the number of nurses per capita, per thousand people. Change in the number of nurses was steadily increasing through the whole observed period. Because of the argument mentioned above that nurses per capita should provide a better picture in how changes in the provision of goods correlate with independent variables, we will test the changes in Nurses per capita in the following analysis.
Finally, the most used public health provision variable in literature are the infant mortality rates (IMR). Infant Mortality Rate compares the number of deaths of infants under one year old, in a given year, per 1,000 live births in the same year. I used the data on the number of infant deaths, and the number of daily births from Turkish National Statistical Institute for the observed period. Additionally, I used data on live births, and infant deaths for the year 2002 to estimate the changes of each observed year with regards to the 2002 infant mortality rates. Similar to the changes in Illiteracy, using this method we ignore previous disparities between provinces and focus only on the changes during the AKP mandate.
Graph 5 shows the average infant mortality rates in each year.
Graph 5 Average Infant Mortality Rates Data source: TURKSTAT (2016)
Mean infant mortality rates have experienced a very steep decline from 2009 to 2013. They reached very low level of 10.5 deaths per thousand live births in the 2013, but in 2014 the rate increased to 11 deaths per thousand people. The reason for this increase might be the inflow of refugees from Syria, who are settled in the refugee camps in the south-eastern areas of the country.
However, the biggest surprise was that changes in infant mortality rates with respect to 2002 rates were negative. Infant mortality rates in the period 2009-2014 were higher than in 2002. However, the changes were declining in the observed period, meaning that the infant mortality rates have been decreasing in the observed period.
Graph 6 Average changes in IMR (t - t2000) Data source: TURKSTAT (2016)
2.4.5 Control Variables
Income levels
To control for the income levels in each of the provinces, I used the consumption expenditure on food and non-alcoholic beverages, as a share of total consumption expenditures. The reasoning behind this approach is that the higher shares of food expenditures, in total spending, usually imply that an individual (in this case a province), is poor. Provinces were clustered in 26 sub-regions. For example, TRA1 sub-region consists of three provinces: Bayburt, Erzurum, and Erzincan. Thus, in the final dataset, they will have the same score for the food consumption expenditure. Istanbul, Ankara, and Izmir have the lowest share of food expenditures in total spending, while Kurdish provinces have the highest score, because of lower development in this area. Hence, including this variable means accounting for the disparities in regional development. The decline in the share of consumption expenditure on food in total expenditure means that overall regions performed better, and the income increased. There has been a significant decrease of 4% between 2009 and 2014, which is shown on the graph bellow.
Graph 7Average share of consumption expenditure on food in total consumption
Data source: TURKSTAT (2016)
Population
Data on the entire population per province was collected from the Address based Population Registration System, downloaded from TURKSTAT (2016). Since population varies a lot across provinces, it is necessary to control for that variation, especially for gross regional product components, where highly populated areas have significantly higher population.
Share of urban population
The proportion of urban population in total population per province is also collected from the Address based Population Registration System from the TURKSTAT website. The dataset consists of the absolute values of urban population, rural population, and its related shares. Several provinces: Adana, Ankara, Antalya, Aydin, Bursa, Balikesir, Denizli, Diyarbakir, Erzurum, Eskisehir, Gaziantep, Hatay, Istanbul, Kahramanmaras, Kayseri, Kocaeli, Konya, Malatya, Manisa, Mardin, Mersin, Mugla, Ordu, Sakarya, Samsun, Sanliurfa, Tekirdag, Trabzon, and Van, were missing data on the rural population in the period 2013-2014. Consequently, the share of the urban population was 100%, which means that the absolute number of urban population was equal to the total population in the province. For some provinces, such as Istanbul and Izmir, the share of the urban population was as high as 98 %, but for province Hatay was around 50%. I decided to use the latest proportion of urban population which was for the year 2012 and to calculate missing values for the rural population. If I kept the proportion of the urban population at a 100%, it would seem like there was a big migration flow from rural to urban areas, which wasn't the case. Therefore, using the latest available shares provides a more realistic picture of how the population in each province is structured and reduces the error term.
2.5 Model
When analyzing panel data, there are three widely used models: Pooled OLS, Fixed effects, and Random effects model. Pooled OLS model would be a waste since it does not account for time series. Fixed effects model produces estimates within provinces, and it is used when controlling for all time-varying variables. Kohler and Kreuter (2009) highlight that it is impossible to test time-invariant variables, such as ethnicity, religion, and race since they are perfectly collinear with the entity's dummies. Fixed effects model, thus, controls for all fixed effects, such as ethnicity shares, but it does not produce the coefficient for those time-invariant characteristics. Since the treatment variable, share of the Kurdish population is time-invariant, fixed effects model would not give the desired estimate, and it would be impossible to see the correlation between the dependent and treatment variable.
On the other hand, Random effects model would give the estimates for a time-invariant variable, but it would produce biased results. If all time-invariant variables were included in the model, heterogeneity bias wouldn't be an issue. However, it was difficult to control for all province specific characteristics due to the lack of available data. Therefore, errors would be underestimated, and heterogeneity bias would occur. Additionally, it is very hard to interpret the results since within-between entity coefficients are not separated. A solution for this can be found in the Mundlak's work from 1978, where he proposed the usage of the so-called within-between model or a hybrid model. Bell and Jones (2015), argue for using hybrid models, which use the random effects model, and control for between effects by including individual means (the higher-level entity mean), of each time-varying covariate, to separate the within and between effects. This hybrid model was explained in the next few lines, and it follows Bell and Jones's adaptation of Mundlak's (1978a) formulation. They argue that with this approach, heterogeneity bias is not only corrected but explicitly modeled.
Micro models are formulated by the following equations:
,
where represents all time-invariant components of the model., are all controlled time varying variables. is composed of , which is a time invariant variable (in this case share of Kurdish population in a province), and , which represents the higher-level entity j's mean of all time varying variables. Since the mean is estimated for each entity j, now represents the time invariant component of controlled variables (Snijders & Bosker, 2012) (Bell & Jones, 2015).
The macro model equation, therefore, is:
.
In the macro model, , now, represents the within effect, because controls for between effect. Jones & Bell state that is the contextual effect, since it models the difference between the within and between effects. Berlin et al. (1999), rearranged , by deriving it as:
,
which rearranges to:
.
In the final model, is the within model, and is the between effect of the variables . (Bartels, 2008) (Leyland , 2010) (Bell & Jones, 2015)
The residuals are - the between entity error, and - the within entity error, and they are assumed to be normally distributed:
)
).
Snijders and Bosker (2012), highlight the advantage of the within-between model, when it comes to interpretation of the coefficients since the within and between effects are clearly separated from each other. Jones and Bell also advise the usage of this model, because, in the first formulation, there was a correlation between and , but using the group mean centering this correlation is lost (Raudenbush, 1989), which leads to better and more stable estimates. Lastly, the third advantage is that even if there is a multicollinearity between 's and the treatment, time-invariant variable , those 's can be excluded from the regression without the fear of heterogeneity bias.
The following chapter on results will be presented as follows: Firstly, we will run regressions on several dependent variables, which represent the public goods provision. Secondly, I will conduct the Exploratory Factor Analysis to create the index of public goods provision using all previously mentioned variables, and the two separate indices which run on the indirect, and the direct provision of public goods. Then those indices will be used as dependent variables in regressions. Finally, I will use changes in illiteracy rates, changes in the infant mortality rates, and changes in the number of nursing staff as dependent variables, to explore the relationship between public goods provision, and the ethnic characteristics and economic performances.
3. Findings
This chapter is divided into 4 main sections. Firstly, I will interpret the results based on the several proxies of public goods provision. After that follows the chapter on the Exploratory factor analysis in which the process of constructing indices will be explained. These indices will represent the public goods provision, and will be used as dependent variables in the regressions. Lastly, I will test the public goods provision using changes in illiteracy rates, IMR, and the number of the nursing staff.
3.1 Individual regression results
First within-between model I ran on several variables from the dependent variables dataset. Illiteracy rates, the number of nursing staff per capita, infant mortality rates, and the number of general high schools per capita are variables that represent both indirect and direct provision, in both education and health dimensions. I used natural logarithm of the values of the gross regional product sectors - agriculture, industry, and services. Standard errors are clustered at the sub-region level, which is one level above province grouping. There are 26 sub-regions in which provinces are territorially grouped. Since the within-between model explicitly models heterogeneity bias by including individual means, regression assumes robust standard errors (Bell & Jones, 2015). Results on the above-mentioned Y variables are shown in Table 1.
Table 1 Regression results
Regression Table 1 |
||||||
(1) |
(2) |
(3) |
(4) |
(5) |
||
Illiteracy rate |
Bachelor diploma |
IMR |
Nurses per capita |
High schools per capita |
||
Kurdishness |
0.0534*** |
-0.0312*** |
4.930** |
-0.259 |
0.00106 |
|
(3.82) |
(-4.63) |
(2.83) |
(-1.42) |
(0.09) |
||
W Agriculture |
-0.0472*** |
-0.00399** |
0.133 |
0.123 |
0.00886** |
|
(-6.68) |
(-2.62) |
(0.14) |
(1.94) |
(3.11) |
||
W Industry |
-0.0282*** |
0.0390*** |
-3.128*** |
0.506*** |
-0.0239** |
|
(-4.54) |
(9.77) |
(-3.54) |
(6.21) |
(-2.88) |
||
W Services |
-0.0242* |
0.0696*** |
-4.472*** |
0.816*** |
-0.0337** |
|
(-2.19) |
(19.97) |
(-3.66) |
(7.68) |
(-2.89) |
||
W UR |
0.000579* |
0.0000676 |
0.0880* |
-0.00126 |
-0.000179 |
|
(2.06) |
(0.42) |
(2.13) |
(-0.31) |
(-1.10) |
||
M Agriculture |
-0.00153 |
-0.00872** |
1.589*** |
-0.231* |
-0.000695 |
|
(-0.38) |
(-2.87) |
(4.51) |
(-2.41) |
(-0.22) |
||
M Industry |
-0.00783** |
0.00150 |
-0.753 |
0.0190 |
-0.0128 |
|
(-3.19) |
(0.76) |
(-1.76) |
(0.25) |
(-1.90) |
||
M Services |
-0.00425 |
0.0144*** |
-1.414* |
0.318** |
-0.0152* |
|
(-0.97) |
(4.03) |
(-2.29) |
(2.67) |
(-2.44) |
||
M UR |
0.00125 |
-0.000252 |
0.143 |
-0.0598*** |
-0.00185 |
|
(1.06) |
(-0.38) |
(1.13) |
(-4.14) |
(-1.80) |
||
Income levels |
0.435*** |
0.0110 |
-8.019 |
0.622 |
-0.119 |
|
(3.69) |
(0.34) |
(-0.72) |
(0.57) |
(-1.24) |
||
Population |
1.81e-09 |
-2.00e-09 |
0.000000190 |
-0.000000143*** |
2.11e-09 |
|
(1.00) |
(-1.71) |
(1.03) |
(-3.39) |
(1.44) |
||
Urban pop. |
-0.000318 |
0.000262 |
0.00609 |
0.00761** |
0.000532 |
|
(-1.79) |
(1.85) |
(0.19) |
(2.74) |
(1.81) |
||
constant |
0.0493 |
-0.0207 |
10.21* |
0.224 |
0.314** |
|
(0.82) |
(-0.70) |
(2.26) |
(0.30) |
(3.08) |
||
Note: t statistics in parentheses; * p<0.05, **p<0.01, ***p<0.001 |
Due to the nature of the within-between model, it is easier to interpret the coefficients, since the within effect is clearly separated from the between effect. Therefore, if agricultural value, increases by one percent within a province, everything else constant, we can expect to see a decrease in the share of illiterate people by 4 percent. Similarly, a percent change in Industry and Services, ceteris paribus, is associated with a decline in the share of Illiterate people by 2.8 and 2.4 percent, respectfully. Also, if the unemployment rate increases by 1 percent in a province, ceteris paribus, we can expect to see an increase in the share of illiterate people by 0.05 percent. Therefore, within a province, changes in the economic indicators have a positive effect on the decline in illiteracy levels. The model explains 78.3 % of the variance within a province.
Individual means of time-varying variables capture the between entity effects. Lower education attainment indicators, illiterate, literate without a diploma, and primary education, are positively correlated with the share of the Kurdish population. Hence, if a share of Kurdish population increases by one percent in a province, holding other variables constant, we can expect to see an increase in illiteracy rate by 5.3 percent. This result implies that there is a negative effect of Kurdishness on the illiteracy rates. One explanation for this might be because Kurds are not allowed to use Kurdish language in public, and are not being taught in their mother language. Benson (2005) compared the education in a language that students don't understand to “holding learners under water without teaching them how to swim.” In the report on the discrimination in Turkey's education system, Kaya (2015) emphasized the important role of providing the education in a mother language, to the quality of education, where teaching in a language students do not understand leads to higher drop-out rates and repeated years of schooling. Another explanation for a positive correlation between low educational attainment and the Kurdishness of a province, is the there is a practice of the lower provision of funds for education from central government to Kurdish provinces.
Among the economic factors in the between model, only industrial value has a significant relationship with illiteracy rates. If the industrial value increases by one percent, holding other variables constant, we can expect to see a decline in the share of illiterate people by 0.78 percent.
In the second regression, I tested the effects of ethnic characteristics and economic performance on the share of the population with an undergraduate diploma. Within a province, a one percent increase in agriculture, ceteris paribus, is associated with a decline of 0.39 percent in the share of people with a university degree. On the other hand, industry and service positively correlate with the proportion of the population with a university degree, where a percent increase in these two sectors, ceteris paribus, is associated with an increase in Bachelor's diploma holders by 3.9 and 6.9 percent. This very high increase implies that the economic outcomes largely affect higher education. Between effects are following: A percent increase in agriculture, ceteris paribus, is associated with a decrease in the share of the population with an undergraduate diploma by 0.8 percent. However, higher negative coefficient has a share of Kurdish population variable, where a percent increase in Kurdishness, leads to a decrease in the proportion of the population with the undergraduate diploma by 3.1 percent. Moreover, services are positively associated with the undergraduate diploma shares, with a coefficient of 1.4. The overall R^2 of the regression with Bachelor diploma is 74.5.
Results of the regression with Y variable infant mortality rate are as follows: Within a province, a percent increase in industry and services, ceteris paribus, is associated with the decline in the infant mortality rates by 3.12 and 4.47 per thousand live births. A percent increase in the unemployment rate, ceteris paribus, is associated with an increase in the IMR by 0.8 per thousand live births. Across provinces, a one percent increase in Kurdishness, ceteris paribus, is associated with an increase in the infant mortality rates by 4.9. Additionally, a percent change in the average agricultural product value and the mean services value, ceteris paribus, is associated with an increase by 1.6, and a decrease by 1.4 infant deaths per thousand live births. Again, Kurdishness had the highest coefficient in the regression meaning that Kurdish provinces have the highest effect on the IMR, which is a proxy for low public health provision scores. Additionally, we can see that the higher agricultural activity in a province might lead to more infant deaths per thousand live births.
Within province results for the fourth dependent variable, nursing staff per thousand people, are as follows: A percent change in industry and service value, ceteris paribus, might be associated with an increase in the number of nurses per thousand population by 0.5 and 0.8, respectfully. Thus, positive changes in health provision from year to year, are due to higher economic performances in secondary and tertiary sectors. Across provinces, a percent change in mean agriculture and mean unemployment rate, ceteris paribus, is associated with a decline of 0.2 and 0.05 nurses per thousand population. Additionally, a percent increase in mean Services, ceteris paribus, is associated with an increase of 0.3 nurses per thousand population. Therefore, it seems that provinces with the large tertiary sector, on average, have more nursing staff per capita. Share of the Kurdish population had no significant relationship with the number of nurses in each province. Overall R^2 is 48.9.
Lastly, regression results on the direct provision of public schools, number of general high schools per capita, show a positive relationship between the agricultural value and the number of secondary schools on both levels. Additionally, services and industry are negatively correlated with the dependent variable. Share of Kurdish population has no significant relationship with the number of secondary schools. But, R^2 of both within and between effects (0.18, and 0.3 respectfully), is low meaning that these results are not adequate for further discussion.
Since regressions results with different proxies for public goods provision vary, I decided to create a Public goods provision index using the EFA to capture the effects of the common variable which affects all dependent variables. In this case, that variable is the public provision in education and health.
3.2 Exploratory Factor Analysis
There has been a growing trend in the empirical research of the effects of race or ethnicity or cultural disparities between groups in various countries, and on the cross-country level. To estimate the relationship between ethnic and economic characteristics, and the delegation of those goods, I decided to create a public goods provision index. If the ethnic structure really does affect lower public goods provision in Turkish provinces, then it will help in redesigning the policies to stimulate the take-up in the future, or to design specific projects made only for underdeveloped areas Kurdish areas. Again, all variables used for the creation of index measure only health and education dimension of public goods provision. All variables were standardized before conducting the analysis.
To construct the index, I used the Exploratory Factor Analysis, whose goal is to identify the underlying relationships between measured variables. Specifically, factor analysis groups sub-indicators that are collinear to form a composite indicator that captures as much of common information among variables as possible. (Nardo, et al., 2015) There are two popular techniques used in EFA, maximum likelihood and principal axis factoring. Maximum likelihood method is widely used on the data that is normally distributed, which is not the case in the data used for this research. Therefore, I used the principal axis factoring, which is similar to the Principal Components Analysis, because they both produce results where the first factor/component explains the highest possible common variance, the second factor next highest common variance, and so on. The first advantage of the Principal axis factoring is that it works when the normality assumption is violated (Fabrigar, Wegener, MacCallum, & Strahan, 1999). Secondly, it is less likely to produce incorrect information, than the maximum likelihood method (Finch & West, 1997). However, the major limitations are that it does not allow for the computation of confidence intervals and significance tests.
Since several treatment variables that are trying to explain the public goods provision in Turkey are highly correlated with each other, I used the Cronbach's alpha method, which is a measure of internal consistency, or reliability of the data, or how closely related a set of items are as a group. Cronbach's alpha ranges from 0 to 1, and it shows the ratio of two variances. It also shows the overall alpha scores if each variable was excluded from the dataset. If all the scale items included in the dataset are highly correlated or have high covariances, then alpha is going to be high, which means that the items measure the same underlying effect.
Kaiser-Meyer-Olkin's test (KMO), indicates whether the variables can be grouped into a smaller set of factors. KMO ranges from 0 to 1, where the rule of thumb is to have the score higher than 0.6. Keiser and Rice (1974), classify the scores from unacceptable (> .5), to marvelous (<= .9).
3.2.1 Indices
Firstly, I decided to create one index capturing overall provision of public goods. Overall means it runs on variables that both, directly and indirectly, measure public goods in health and education. Using the EFA method, we are trying to measure the latent variable, Public goods provision. This latent variable cannot be directly measured with a single variable. Instead, it is seen through the relationships it causes in a set of dependent variables. Therefore, first EFA has been conducted on 16 variables:
· shares of illiterate, literate without a diploma, primary education, high school, Bachelor, Masters, and Doctorate diploma population,
· number of Preprimary, primary, general secondary, and vocational high schools,
· number of hospitals, beds, physicians, and nurses per thousand population, and
· the infant mortality rates.
Exploratory factor analysis starts with producing eigenvalues which show the proportion of the total variance explained by each factor. Each factor is a specific combination of variables whose factor loadings represent the weight of each variable in the given factor. Any factor with an eigenvalue ?1 explains more variance than a single observed variable. However, there is no specific rule when choosing the right number of factors. The Kaiser's stopping rule (Factor Analysis: A Short Introduction, Part 1 , n.d.) says that we should include the components whose eigenvalues are above 1. The rule of thumb is to include all components which have a steep slope in the scree plot. Finally, it is the choice of the researcher to decide on how many factors he/she should continue. Since the factor one explains 59% of the total variance, and loads on all variables, I decided to use only one factor. Cronbach's alpha score was 0.90 Cronbach's alpha scores table is in the Appendix 2, which implies that the dataset is reliable and that it could be used for the applied scenarios research. Lance et.al. (2006), and Nunally (1978).
Factor loadings for the Overall Provision Index are shown in the first column of the Table 1. It loads negatively on low performances and low infant mortality rates, and positively on higher education and the health indicators. The KMO score of .7681 is classified as Meritorious, and the variables are considered adequate for the exploratory factor analysis (EFL). Ankara province has the highest index score, followed by the Eskisehir province. Both provinces are considered developed, urban areas. The lowest Overall Provision Index scores are for the provinces Mus, Agri, Sanliurfa, followed by Mardin, Bitlis, and Sirnak, which are the least developed provinces, and provinces populated by Kurdish ethnic group.
Table 2 Factor loadings for the Overall Provision, Direct Provision, and Indirect Provision Index
Variables |
Overall provision loadings |
Direct provision loadings |
Indirect provision loadings |
|
% Illiterate |
-0.77 |
-0.80 |
||
% Literate without diploma |
-0.70 |
-0.72 |
||
% Primary education diploma |
-0.30 |
|||
% High school diploma |
0.81 |
0.81 |
||
% Bachelor diploma |
0.89 |
0.92 |
||
% Masters diploma |
0.82 |
0.87 |
||
% Doctorate diploma |
0.80 |
0.77 |
||
Infant Mortality Rate |
-0.56 |
-0.58 |
||
No. Pre-prim schools |
-0.60 |
|||
No. Primary schools |
-0.73 |
|||
No. General secondary schools |
0.11 |
0.51 |
||
No. Vocational secondary schools |
0.21 |
0.56 |
||
No. Pre-prim teachers |
||||
No. General secondary teachers |
0.62 |
|||
No. Vocational secondary teachers |
0.67 |
|||
No. Hospitals |
0.21 |
0.58 |
||
No. Hospital beds |
0.64 |
0.73 |
||
No. Physicians |
0.77 |
0.57 |
||
No. Nurses |
0.77 |
0.88 |
Secondly, I grouped dependent variables in 2 types: Variables capturing direct effects of the funding from the central government such as public institutions (schools and hospitals per capita), hospital beds per thousand people, and staff in both sectors (teachers and medical staff per capita). The second, indirect type, are the effects of government projects, and funding, such as share of illiterate people, share of the population with Master's degree, and the infant mortality rates. All variables loaded strongly on the first factor in both EFA's.
Hospital beds, Nurses, and the high school level teachers have the highest loadings, and therefore will have the highest weight in the Direct Provision Index variable. In the Indirect Provision Index, all variables have very high loadings. Variables that have lower loadings are still important since they still have a strong relationship with the underlying factor. The Cronbach's alpha for Direct Provision Index was 0.79, and for the Indirect Provision 0.88, with KMO's of 0.65, and 0.77, respectfully. Direct Index explains 57% of the variation in variables loading in the Index and has an eigenvalue of 3.3. The explained variation of variables loading in the Indirect Index is higher, 82%, with an eigenvalue of 4.4. These two indices were used as dependent variables in the regressions with the same treatment variables as in the previous chapter.
3.3 Regression results on the Public goods provision indices
To test whether the public goods provision was correlated with the ethnic characteristics and economic performances of provinces, I ran two models, one with GRP values (in TL), and another with the sectoral shares of each GRP component. Additionally, I generated individual mean, or the higher-level entity mean, for each independent variable to exclude the between province effect from the within effect. In the model explained previously, that is the , or the between province effects. Also, to get the within province effects , or time varying variables, I subtracted the individual means from actual values of each variable. Therefore, the regression results on three indices are presented in Table 3.
Table 3 Regression results with Indices
Regression Table 2 |
|||||||
(1) |
(2) |
(3) |
(4) |
(5) |
(6) |
||
Public Goods Provision Index |
Direct Provision Index |
Indirect Provision Index |
|||||
Kurdishness |
-1.529*** |
-1.513*** |
-1.080** |
-0.886* |
-1.593*** |
-1.655*** |
|
(-5.51) |
(-5.74) |
(-3.06) |
(-2.35) |
(-5.81) |
(-5.95) |
||
W Agriculture |
0.135*** |
0.275*** |
0.126** |
||||
(3.32) |
(3.88) |
(2.80) |
|||||
W Services |
1.424*** |
1.091*** |
1.505*** |
||||
(13.65) |
(6.17) |
(14.89) |
|||||
W Industry |
0.829*** |
0.624*** |
0.841*** |
||||
(7.33) |
(4.90) |
(7.81) |
|||||
W % Agri. product |
-1.958* |
-0.835 |
-1.888* |
||||
(-2.13) |
(-1.09) |
(-2.10) |
|||||
W % Serv. Product |
-0.236 |
-0.202 |
-0.476 |
||||
(-0.24) |
(-0.22) |
(-0.49) |
|||||
W % Ind. Product |
-0.566 |
0.744 |
-1.133 |
||||
(-0.52) |
(0.68) |
(-1.14) |
|||||
W GRP |
1.265*** |
1.033*** |
1.358*** |
||||
(11.71) |
(5.67) |
(16.49) |
|||||
W UR |
0.00265 |
-0.000990 |
0.00176 |
0.0000855 |
0.00373 |
-0.00121 |
|
(0.53) |
(-0.17) |
(0.34) |
(0.02) |
(1.05) |
(-0.27) |
||
M Agriculture |
-0.417*** |
-0.422* |
-0.305** |
||||
(-3.45) |
(-2.04) |
(-2.76) |
|||||
M Service |
0.640*** |
0.311 |
0.412*** |
||||
(4.18) |
(1.37) |
(3.66) |
|||||
M Industry |
0.167 |
-0.160 |
0.115 |
||||
(1.64) |
(-0.97) |
(1.53) |
|||||
M % Agri. product |
-2.760*** |
-0.834 |
-2.891** |
||||
(-3.49) |
(-0.74) |
(-3.23) |
|||||
M % Serv. product |
2.252* |
5.121*** |
0.511 |
||||
(2.56) |
(3.93) |
(0.51) |
|||||
M % Ind. product |
-2.242* |
-5.128*** |
-0.486 |
||||
(-2.52) |
(-3.95) |
(-0.47) |
|||||
M GRP |
0.160* |
-0.176 |
0.0467 |
||||
(2.11) |
(-1.56) |
(0.59) |
|||||
M UR |
-0.0491 |
-0.0459 |
-0.120*** |
-0.122*** |
-0.0293 |
-0.0231 |
|
(-1.77) |
(-1.74) |
(-4.15) |
(-3.60) |
(-0.98) |
(-0.79) |
||
Income levels |
0.347 |
1.021 |
-1.946 |
-0.829 |
0.894 |
1.417 |
|
(0.35) |
(0.95) |
(-1.04) |
(-0.43) |
(1.55) |
(1.89) |
||
Population |
-5.15e-08 |
-0.000000152** |
-0.000000164*** |
-0.000000237** |
0.000000103** |
1.62e-08 |
|
(-1.95) |
(-3.14) |
(-3.88) |
(-2.91) |
(2.62) |
(0.38) |
||
Urban pop. |
0.0211*** |
0.0193*** |
0.0266*** |
0.0178** |
0.0139* |
0.0151** |
|
(3.44) |
(3.70) |
(4.37) |
(3.19) |
(2.48) |
(2.91) |
||
constant |
-3.815** |
-4.515*** |
0.714 |
1.804 |
-1.152 |
-2.802** |
|
(-3.05) |
(-4.02) |
(0.42) |
(1.03) |
(-0.84) |
(-2.93) |
||
Note: t statistics in parentheses; * p<0.05, **p<0.01, ***p<0.001 |
First two regressions in the Table 1 I ran on the dependent variable Public goods provision index. Overall R^2 for both models was high, explaining 73, and 76 percent of variation between variables. Within a province, a percent increase in the share of agricultural product in total GRP, ceteris paribus, is associated with a decline in the Overall Provision Index by 0.019. However, a percent increase in the total GRP, ceteris paribus, is associated with an increase in the Overall Provision Index by 1.26. The coefficient for the overall GRP is in line with the coefficients from the second model, where within a province, a percent increase in the agriculture, industry, and services values, ceteris paribus, are associated with the increase in the Index score by 0.13, 0.8, and 1.4 respectively.
Between provinces, a percent increase in the share of Kurdish population, ceteris paribus, is associated with a decline in the Overall Provision Index score by 1.53. Additionally, higher shares of agricultural and industrial product seem to correlate negatively with the public provision index, with the coefficients of -2.76, and -2.42 respectively. Total GRP value and the share of services product in total GRP, have positive relationship with the index scores. However, when we look at the results with sectoral values, agriculture has a significant negative relationship with the index, where a percent increase in the agricultural value, ceteris paribus, is associated with the decline in the index score of 0.42. Results from the regression table with the Overall provision index imply that the provision of public goods in education and health is negatively associated with the share of Kurdish population, and the higher shares of the primary and secondary sector in total GRP. Therefore, if a province has a high share of Kurdish population and high share of agricultural production in total provincial production, it will most likely have lower provision of public goods. However, if in general, economic activity in a province is increasing, public goods provision will be higher. Regression produced insignificant coefficient for unemployment rates using both, values, and sectoral shares.
Above-mentioned results lead to a conclusion that across provinces, the biggest determinant of a public goods provision in health and education is the ethnic structure of a province. However, it is important to highlight the limitations of the process in construction of the Index and the usage of indirect variables. Limitations of the Index are that it explains 59% of the total variance of 16 variables included in the Exploratory factor analysis. Additionally, out of 59% of the variation, index loads highly on the variables such as illiterate, literate without a diploma, and other educational levels, which have higher correlation with the shares of the Kurdish population. On the other side, it also loads highly on hospital beds, physicians, and nurses, which show the direct and the fastest change in the public goods provision. Therefore, I ran two regressions both for Direct and Indirect Provision Index to see if the clustering variables would affect coefficients in the model.
Direct Provision Index, as previously mentioned, runs on the number of public institutions per capita, and on the teachers and medical staff in hospitals, as well as hospital beds per thousand people. Within province effects imply that a one percent change in the agriculture, industry, and service value, ceteris paribus, is associated with the increase in the Direct Provision Index of 0.27, 1.09, and 0.62, respectfully. Additionally, none of the three sectoral shares in total GRP were significant on the within level. On the between province level, share of Kurdish population negatively correlates with the Direct Provision Index. Therefore, a percent change in the share of the Kurdish population, ceteris paribus, leads to the decline in the Direct Index score by 1.08, and 0.9, depending on the economic dimension variables used. Additionally, agricultural value negatively correlates with the Direct Provision Index, meaning that a province experiencing higher agricultural activity might be receiving proportionally less than other provinces. Share of services sector in total GRP positively correlates with the index, where a percent increase in the share of tertiary sector, ceteris paribus, is associated with the increase in the Direct Provision Index by 5.12. on the contrary, share of industrial product has negative relationship with the Index with the coefficient of -5.12. That said, a province focusing the economic activity on the industry instead of on the services sector, might on average have lower scores in the Direct Provision Index. Similarly, unemployment rates are negatively correlated with the index scores, where a percent increase in the unemployment rate, ceteris paribus, is associated with the decline in the Index scores by 0.12. Therefore, we can argue that provinces with the higher agricultural and industrial product in total GRP have less public goods provision, measured by the number of public institutions and staff employed in the health and education sectors. However, Kurdishness is significantly correlated with the Direct Index which is contrary to the findings from the individual regression models on the size of the nursing staff per capita, and the number of general secondary schools. These results, also imply that there has been a practice of lower provision in health and education to Kurdish areas, which is in line with the findings of the Overall Provision Index.
Lastly, the third index runs on the direct effects of the public funds delegated to provinces, specifically, how did government decision of funding affect the outcomes in human capital. The model with the Indirect Provision Index confirmed the results from the previous two sets of regressions. On the within level, higher economic activity in all three sectors positively affects the Index scores. However, share of the agricultural product in the total GRP, again, has a negative relationship with the public goods provision. Therefore, a percent increase in the share of the agricultural product, ceteris paribus, is associated with a decline in the index score of 2.89. According to these findings, the effects of the public goods provision are most likely going to be lower in the provinces with higher shares of agricultural product in total GRP. Between level results show that again higher share of agricultural sector, and agricultural value, negatively associate with the Indirect Provision Index. Finally, share of Kurdish population has the most significant coefficient, where a percent increase in the share of Kurdish population, ceteris paribus, leads to the decline in the Indirect Index score by 1.6. R^2 for both indices was high, 73 for Indirect, and 63 for the Direct Provision Index.
Therefore, the overall conclusion is that the Kurdish provinces profited significantly less in terms of a public goods provision. More to it, provinces with the higher agricultural and industrial activity have most likely lower scores than the provinces with the higher shares of services sector in total GRP. Since Kurdishness doesn't correlate with the share of agricultural product in total GRP, we can conclude that these are two separate factors shaping public goods provision in Turkey. However, this research paper focuses on the AKP-run government, and the mechanisms of public provision in the period of higher economic growth. Using variables in the present without excluding the historical effects doesn't necessarily show the provision of goods now. Since Kurdish population has been marginalized for a whole century, illiteracy levels and other proxies for public goods provision are worse than in other areas of Turkey. Therefore, in the next chapter I will present the results on the changes in the public goods provision by ignoring the historical practice which affected the current amount of goods in provinces.
3.4 Changes in Illiteracy Rates
Changes in illiteracy rates have been calculated by subtracting the 2000 Illiteracy shares from the corresponding province in the period 2009-2014. If in Adana province, proportion of illiterate people in 2009 was 9.8, and in 2010 7.9, we subtract the 2000 share 13.11. New values for 2009 and 2010 are going to be -3.2, and -5.1, respectively. Using this method, we only account for the actual changes from each year. Therefore, we can analyze the trend in changes, specifically whether Kurdish provinces had a significantly different trend in these changes than other provinces, during the AKP mandate. Similarly, using this method, we can analyze the changes in the infant mortality rates, and nursing staff. Regression results on the changes in illiteracy rates are in the following table.
Table 4 Regression results with Delta illiteracy rates
Regression Table 3 |
|||
(1) |
(2) |
||
Changes in Illiteracy rates |
|||
Kurdishness |
0.0939*** |
0.0907*** |
|
(4.49) |
(4.57) |
||
W Agriculture |
0.0498*** |
||
(6.31) |
|||
W Services |
0.0473*** |
||
(4.50) |
|||
W Industry |
0.0378*** |
||
(6.06) |
|||
w % Agriculture product |
0.18** |
||
(2.73) |
|||
w % Services product |
-0.227* |
||
(-2.34) |
|||
w % Industry product |
0.209* |
||
(2.01) |
|||
w GRP |
0.0602*** |
||
(4.70) |
|||
w UR |
-0.000161 |
-0.000400 |
|
(-0.32) |
(-0.98) |
||
M Agriculture |
0.00817 |
||
(1.51) |
|||
M Services |
-0.00827 |
||
(-1.27) |
|||
M Industry |
-0.00766 |
||
(-1.71) |
|||
M % Agriculture product |
0.0996 |
||
(1.75) |
|||
M % Services product |
0.0706 |
||
(1.31) |
|||
M % Industry product |
-0.0697 |
||
(-1.31) |
|||
M GRP |
0.00108 |
||
(0.24) |
|||
M UR |
0.00208 |
0.00198 |
|
(1.37) |
(1.30) |
||
Income levels |
-0.241 |
-0.150 |
|
(-1.75) |
(-1.23) |
||
Population |
-4.29e-09* |
-3.37e-10 |
|
(-2.26) |
(-0.14) |
||
Urban pop. |
-0.000354 |
-0.000497 |
|
(-1.03)... |
Подобные документы
The levers of management of a national economy, regions and enterprises. The prices for the goods. Taxes to the proceeds from realization of commodity production. Proceeds from realization of services to the population, establishments and organizations.
реферат [18,7 K], добавлен 12.04.2012The use of computers in education. Improvements in health, education and trade in poor countries. Financial education as a mandatory component of the curriculum. Negative aspects of globalization. The role of globalization in the economic development.
контрольная работа [57,9 K], добавлен 13.05.2014Negative consequences proceeding in real sector of economy. Social stratification in a society. Estimation of efficiency of economic safety. The parity of the manufacturers of commodity production. Main problems of the size of pension of common people.
статья [15,4 K], добавлен 12.04.2012Prospects for reformation of economic and legal mechanisms of subsoil use in Ukraine. Application of cyclically oriented forecasting: modern approaches to business management. Preconditions and perspectives of Ukrainian energy market development.
статья [770,0 K], добавлен 26.05.2015Investments as an economic category, and their role in the development of macro- and microeconomics. Classification of investments and their structure. Investment activity and policy in Kazakhstan: trends and priorities. Foreign investment by industry.
курсовая работа [38,8 K], добавлен 05.05.2014The essence of economic efficiency and its features determination in grain farming. Methodology basis of analysis and efficiency of grain. Production resources management and use. Dynamics of grain production. The financial condition of the enterprise.
курсовая работа [70,0 K], добавлен 02.07.2011General characteristic of the LLC DTEK Zuevskaya TPP and its main function. The history of appearance and development of the company. Characteristics of the organizational management structure. Analysis of financial and economic performance indicators.
отчет по практике [4,2 M], добавлен 22.05.2015The influence of the movement of refugees to the economic development of host countries. A description of the differences between forced and voluntary migration from the point of view of economic, political consequences. Supply in the labor markets.
статья [26,6 K], добавлен 19.09.2017Defining the role of developed countries in the world economy and their impact in the political, economic, technical, scientific and cultural spheres.The level and quality of life. Industrialised countries: the distinctive features and way of development.
курсовая работа [455,2 K], добавлен 27.05.2015Chinese economy: history and problems. Problems of Economic Growth. The history of Chinese agriculture. The ratio of exports and imports of goods and service to gross domestic product at current prices. Inefficiencies in the agricultural market.
курсовая работа [162,1 K], добавлен 17.05.2014Establishing a favorable environment for investments, removing administrative barriers. Establishing high-technology parks. Formation of financial mechanisms to attract and support investments, tax stimulation measures. Brand promotion of Russian regions.
реферат [15,9 K], добавлен 04.06.2013Theoretical aspects of investment climate in Ukraine. The essence of investment climate. Factors that forming investment climate. Dynamics of foreign direct investment (FDI) in Ukraine. Ways of improving the mechanism of attracting foreign investment.
курсовая работа [155,2 K], добавлен 19.05.2016The major structural elements of economic safety of a national economy branches. The structural analysis of economic activity. Share of wages in ВВП, of productivity of Russia and western countries. The essence of the economic taxes and their purpose.
статья [166,3 K], добавлен 12.04.2012Government’s export promotion policy. Georgian export promotion agency. Foreign investment promotion. Government’s foreign investment promotion policy. Foreign investment advisory council. Taxation system and tax rates in Georgia.
курсовая работа [644,0 K], добавлен 24.08.2005Socio-economic and geographical description of the United states of America. Analysis of volumes of export and import of the USA. Development and state of agroindustrial complex, industry and sphere of services as basic sectors of economy of the USA.
курсовая работа [264,5 K], добавлен 06.06.2014Analysis of the causes of the disintegration of Ukraine and Russia and the Association of Ukraine with the European Union. Reducing trade barriers, reform and the involvement of Ukraine in the international network by attracting foreign investment.
статья [35,7 K], добавлен 19.09.2017The essence of agrarian relations: economic structure and specificity. The land rent, land price as a capitalized rent. History of the formation of agricultural sector of Ukraine, its reforms. Assessment of the investment attractiveness of AIC of Ukraine.
курсовая работа [1,1 M], добавлен 04.01.2016Assessment of the rate of unemployment in capitalist (the USA, Germany, England, France, Japan) and backward countries (Russia, Turkey, Pakistan, Afghanistan). Influence of corruption, merges of business and bureaucracy on progress of market economy.
реферат [15,5 K], добавлен 12.04.2012The stock market and economic growth: theoretical and analytical questions. Analysis of the mechanism of the financial market on the efficient allocation of resources in the economy and to define the specific role of stock market prices in the process.
дипломная работа [5,3 M], добавлен 07.07.2013Concept and program of transitive economy, foreign experience of transition. Strategic reference points of long-term economic development. Direction of the transition to an innovative community-oriented type of development. Features of transitive economy.
курсовая работа [29,4 K], добавлен 09.06.2012