Spatial Correlation Analysis of Unemployment Rates in Turkey

Unemployment is a major problem in Turkey as well as in almost all other countries of the World. Unemployment is defined as the situation of being without a job. A decrease in the growth of economies is a major cause of rising unemployment. (Chowdhuryn and Hossain, 2014). According to economic theory, although the unemployment rate is regarded as an important indicator of labor market performance, there are many other indicators affecting unemployment. Some are listed as the value of imports and exports, the dollar cost of imports and exports, the exchange rate of imports and exports, the exchange rate, population growth, gross national product (GNP) growth at current prices, GNP growth at fixed prices, public investments, private investments and GNP deflator (Goktas and Isci, 2010). Studies regarding the unemployment rate for Turkey generally consider determining the relationship between unemployment and other indicators or variables. For example, Bildirici et al., (2012) investigate unemployment generating effects. Kabaklarli et al., (2011) analyze the economic determinants of the unemployment problem in Turkey. Abstract


Introduction
Unemployment is a major problem in Turkey as well as in almost all other countries of the World.Unemployment is defined as the situation of being without a job.A decrease in the growth of economies is a major cause of rising unemployment.(Chowdhuryn and Hossain, 2014).According to economic theory, although the unemployment rate is regarded as an important indicator of labor market performance, there are many other indicators affecting unemployment.Some are listed as the value of imports and exports, the dollar cost of imports and exports, the exchange rate of imports and exports, the exchange rate, population growth, gross national product (GNP) growth at current prices, GNP growth at fixed prices, public investments, private investments and GNP deflator (Goktas and Isci, 2010).Studies regarding the unemployment rate for Turkey generally consider determining the relationship between unemployment and other indicators or variables.For example, Bildirici et al., (2012) Berument et al. (2006Berument et al. ( , 2009) ) and Berument (2008) research the impact of macroeconomic policy on unemployment in Turkey.Dogrul and Soytaş (2010) investigate the relationship between the price of oil, interest rates and unemployment.Dogan (2012) studies macroeconomic variables and unemployment in Turkey.
On the other hand, there is a large amount of empirical literature that tries to understand the difference between geographical areas in terms of unemployment rates.For examples, Niebuhr (2003) studies spatial interaction and regional unemployment in Europe.Lottmann (2012) explains regional unemployment differences in Germany using a spatial panel data analysis.2002) provide the provincial distribution of unemployment rates in Spain.Filiztekin (2009) studies regional unemployment disparities in Turkey from 1980 to 2000, using spatial and nonparametric techniques.Khamis (2012) examines the spatial pattern of each of illiteracy rate and unemployment rate in Egypt.Odeyemi (2013) considers poverty rates in Nigeria with unemployment rates and illiteracy.Kantar and Günay (2015) investigate the spatial pattern of unemployment rates with the socio economic development index (SDI) and literacy rates for 81 provinces of Turkey.
Turkey is divided into seven main geographical regions; Marmara, Aegean, Mediterranean, Black Sea, Central Anatolia, Eastern Anatolia and South Eastern Anatolia.In Turkey, there are 81 provinces which exhibit substantial differences in terms of economic and social variables.The main differences between west (Marmara and Aegean) and east (Eastern Anatolia and South Eastern Anatolia) are clearly seen.These differences are known as the basic characteristic of the geography of Turkey.In this study, we investigate unemployment rates in Turkey in order to explain differences between geographical areas.A map of Turkey's regions and provinces is shown in Fig 1 .

Spatial Autocorrelation Statistics
To evaluate spatial dependence, we first have to determine what is meant by two observations being close together.In other words, we have to determine the distance measure between locations.Depending on the determined distance, a weight matrix, which defines relationships between locations, is formed.W is the weight matrix with zeros (wij, i ≠ j and i, j=1,2,…,n) on the diagonal and with weights (wii, i=j) on the off-diagonal.wij, i ≠ j is the main component of the spatial autocorrelation measure.
The well-known global spatial autocorrelation measure is the Moran's I given by: where n is the total number of spatial observations (i.e.districts), xi is the value for the spatial location i, xj is the value for another spatial location j. x is the mean value of all spatial locations, wij is the spatial weight between locations i and j.Moran's I can be positive or negative.While its positive value arises when similar values occur near one another, a negative value arises when dissimilar values occur near one another.If the Moran's I value is zero, no spatial autocorrelation is present.
Similarly, the local Moran's I statistics is a well-known local spatial autocorrelation measure.Based on local Moran's I statistics, local spatial autocorrelation analysis (LISA) is conducted (Anselin, 1995).The LISA map is drawn to identify potential local clusters and spatial outliers.While the LISA significance map shows locations with significant local Moran's I statistic, the LISA cluster map provides essentially the same information as the significance map, but shows significant locations in color, coded by the type of spatial autocorrelation.High-high (HH) and low-low (LL) regions show clustering of similar values of the considered variable, while high-low (HL) and low-high (LH) regions indicate spatial outliers (Anselin, 2005)

Spatial distribution of unemployment rate of Turkey (2004-2013)
The distribution of the unemployment rate according to province is explained by maps drawn for the periods (2004, 2011, 2012 and 2013).The breakdown of the maps is calculated using the Natural Break Classification method.
The map for unemployment rates for 2004  Moran's I scatter plots for unemployment rates.Standardized unemployment rates are on the standardized average of neighbors' unemployment rates are given on axis.

patial autocorrelation analyses
In order to further understand the spatial distribution of unemployment rate autocorrelation analyses (LISA) are conducted (Celebioglu, 2010).The results of the LISA cluster maps are summarized in Table 1.As can be seen from Fig 5 ., for all years, most clusters are observed in the South East Anatolia Region.Over the number of provinces in LL clusters increased in the South East Anatolia Region.This means that the neighboring effect of unemployment rate at the provincial level in Turkey increased.For these reasons, public and private sector investment attracted to the east part of Turkey in order to decrease unemployment.

Conclusions and Discussion
High unemployment rates are a concern for all countries in the world, Turkey.
In this study, regional unemployment rates at province Turkey from 2004

• Taking into account Global
Moran's I statistics, it can be observed that there is spatial autocorrelation between geographical areas in terms of unemployment rates at a provincial level in Turkey, and also that spatial autocorrelation has increased over time.
• Macro-economic problems, such as unemployment, have negatively affected social issues, such as the crime rate (Comertler, 2007:15).
For this reason, unemployment should not be considered as solely an economic problem, and should be considered as a factor that can lead to social problems.This study shows that spatial clustering at a provincial-level is observable in Turkey and that terrorist actions in the South Eastern Anatolia Region have an influence on unemployment.

Figure 2 :
Figure 2: Unemployment rates in Turkey (http://globalpse.org/turkiyede-issizliksorunu-(2002-2015)Takinginto account the last three years for all provinces in Turkey, it can be observed that 16.1% of the maximum unemployment rate is seen in Adiyaman in 2011.Next to Adiyaman, Izmir, Batman and Gaziantep have high unemployment rate.Although Izmir is one of the largest industrial provinces in Turkey, its unemployment rate is at a very high level.One reason for this is that Izmir has experienced high inmigration over the last 30 years (Isik, 2009).Kütahya, Kahramanmaraş, Manisa, Uşak and Çorum have the lowest rates, at around 4.7%.For 2012, Batman has the highest rate, at 25%.Morever, Siirt and Mardin follow Batman.Kutahya and Usak have the lowest unemployment rates in 2011, 2012 and 2013.The rates for Batman, Mardin and Siirt are, respectively, the highest in Turkey in 2013, Turkey.When the last three years are taken into account, it can be said that the province experiencing the highest unemployment rate is Batman.

Fig 4 .
Fig 4.Moran's I scatter plots for unemployment rates.Standardized unemployment rates are on the standardized average of neighbors' unemployment rates are given on axis.

Figure 5 :
Figure 5: LISA cluster map for LL clusters (three provinces) are identified in the Moran cluster map for unemployment rates and HH clusters are observed in eight provinces in East Region in 2004.Six provinces are outliers.9 and 4 provinces are in the HH and LL groups, respectively.The number the LL and HH clusters increase by 2 11 provinces in the east part of Turkey are identified in the HH clusters Similar to 2011, there are three LL clusters in 2012 (see Fig 5); one in the Central Region, one in the Black Sea one in the Central West Aegean s of Turkey.One HH cluster is seen the South East Region.
. Spatial outliers have different meanings in the context of spatial statistics.Significant spatial outliers indicate that high values are surrounded by low values while low values are surrounded by high values.