Panel Data Analysis of Rural to Urban Migration Mobility in Türkiye from a Sustainable Development Perspective

Abstract

Rural-to-urban migration is most prevalent in developing countries and has been a key driver of urban growth since the onset of industrialization. Initially beneficial, this migration trend has become unsustainable due to its rapid and uncontrolled rise, posing challenges for both rural and urban development. As a result, attention has shifted toward reducing rural–urban migration and encouraging reverse migration to achieve sustainable development. This study investigates the factors influencing rural-to-urban migration in Türkiye, aiming to contribute to rural development policies in similar economies. Using the Arellano–Froot–Rogers estimator, the study analyzes data from 81 Türkiye provinces over a 12-year period, focusing on variables such as population, human development index, agricultural and industrial income, terrorism, agricultural subsidies, and unemployment. The findings reveal that population, agricultural GDP, terrorism, and human development index significantly impact migration trends. These results suggest that rural outmigration is driven not only by economic factors but also by social and political dynamics. Effective rural development strategies, especially those aligned with the Sustainable Development Goals (SDGs), must therefore adopt integrated and collaborative approaches to reduce migration pressure and potentially reverse the trend in the long term.

1. Introduction

Human migration is an ancient phenomenon that has continually reshaped societies, yet its drivers and forms have varied markedly across historical periods [1]. In recent decades, the global share of international migrants rose from roughly 2.8 percent in the 1990s to about 3.6 percent by 2020, with education, employment, and family reunification among the leading motives [2]. Cyclical economic and social dynamics also shape migration [3]. While international migration follows these broad patterns, internal migration is highly context-specific, and governments formulate policies tailored to national institutional and spatial conditions [4]. Lucas modeled internal migration as the search by individuals to maximize their expected lifetime earnings. In this model, cities are attractive because they offer not only higher wages but also the accumulation of human capital and the potential to acquire new skills required by modern production. Furthermore, he stated that this migration process should be viewed as an irreversible part of a society’s transition from a traditional agricultural economy to a continuously growing, industrialized economy [5]. Internal migration is generally more intense in developing economies and is predominantly rural-to-urban. It is comparatively limited in both least developed and highly developed countries. In the former, the dominance of agriculture in the national economy constrains mobility, while in the latter, the urban transition has largely matured [6,7,8,9,10]. Rural-to-urban migration has been the primary driver of urban growth since the onset of the industrial age, and can be defined as a labor transfer from a traditional land-intensive technology to a human capital-intensive technology with infinite growth potential [11]. In many developing settings, including Türkiye, internal population movements are predominantly rural-to-urban. Although industrialization initially catalyzed this flow, contemporary motivations also include education, health, and broader social-welfare considerations alongside income differentials [12,13,14,15].

Furthermore, it shows that the urban–rural income gap is not only caused by the average income difference but is also related to cities providing greater access to higher-paying, complex, and productive jobs [13]. Irrespective of the cause, the unabated continuation of rural-to-urban migration has led to unplanned and often uncontrolled urbanization, while simultaneously resulting in a continuous depletion of the rural population. This rapid and uneven spatial redistribution has generated structural challenges across economic, social, and political domains. The resulting demographic imbalance between urban and rural regions not only transforms space but also erodes social cohesion and strains local economies. Although many developing countries initially encouraged rural-to-urban flows as a development strategy, unmanaged migration has increasingly exacerbated pressures in both cities and the countryside [10,12,16,17].

Therefore, internal migration must be understood as a complex, multi-layered process tightly interwoven with rural development trajectories. Deficits in rural infrastructure, limited access to public services, and constrained economic opportunities push households toward cities in search of better living conditions. These movements, in turn, lead to unplanned urban growth, mounting infrastructure demand, and widening social disparities. In rural areas, these developments contribute to population decline, population ageing, and lower agricultural output, which together can hinder progress toward sustainable development [4,18,19,20].

The Sustainable Development Goals (SDGs) aim to eradicate poverty, reduce inequality, protect the environment, and ensure dignified lives for all. Rural migration is a salient variable in attaining these aims. The relationship is bidirectional: shortfalls in achieving SDG targets can trigger rural out-migration, while the urban pressures and rural dislocation produced by migration can hinder SDG progress [21,22]. Consequently, rural development strategies explicitly aligned with the SDGs are indispensable.

Within this wider literature, rural-to-urban migration constitutes a central component of internal mobility, especially in developing countries, and has attracted interdisciplinary attention from sociology, economics, geography, and development studies. With the prominence of the SDGs on the global agenda, scholarly interest in rural migration has grown further. A bibliometric analysis was conducted using data from the Web of Science (WoS) core collection to visualize the scope and relationships of the literature on rural migration and to identify the positioning and gap of the study’s main focus (rural-to-urban migration, agriculture, and sustainability) within the literature [23]. A search of the WoS core collection identifies roughly 14,470 publications on rural migration, and a VoSviewer-based bibliometric analysis of this corpus links the topic to 164 keywords (Figure 1).

The VoSviewer (version 1.6.19) analysis presented in Figure 1 makes significant contributions to the study’s core argument and design. This bibliometric analysis reveals several key clusters demonstrating the strongest thematic connections and fields within the literature:

Central Cluster (Large Orange/Yellow): The most frequently used descriptors—migration, rural, and agriculture—are central to this cluster. This indicates that the literature is concentrated around the role of rural areas as centers of agricultural production and the importance of this sector for livelihoods.

Rural–Urban Migration and Country Clusters (Red/Light Blue): This represents a more specific cluster focusing on rural–urban migration and countries like China.

Global/Dynamic Clusters (Green/Purple): These clusters encompass dynamic and broader social processes, such as internal migration, globalization, and entrepreneurship.

The VoSviewer analysis contributes to the fundamental argument of this research in three important ways: The analysis proves the existence and scope of the rural migration literature, demonstrating the study’s positioning within this broad academic field. The emergence of migration, rural, and agriculture as the most frequent descriptors supports the decision to focus on these themes. Considering that the existing literature focuses primarily on the general and socio-economic factors of migration, while agriculture-centric analyses remain limited (despite the sociological and economic importance of the sector), this clearly establishes the original contribution (literature gap) of this study, which examines the role of agriculture in the context of Türkiye.

Regarding previous studies in the literature on ‘rural migration and its causes in Türkiye,’ it has been emphasized that rural migration, which began with industrialization in the 1950s, was accepted as a driving force for economic development and growth [24], and that a massive migration flow from rural to urban areas occurred parallel to socio-economic policies implemented in the post-1980 period in the country [25,26]. Furthermore, some of these studies state that the reduction or removal of state subsidies in agriculture, the introduction of production quotas at the scale of production planning, and increases in agricultural input prices disrupted the rural production structure and created livelihood difficulties, increasing the tendency to migrate to cities [27]. The most important causes of migration in the provinces where migration is most intensive in Türkiye Ağrı, Van, Kars, Iğdır, and Ardahan are highlighted as the inadequacy of health, unemployment, and education services, lack of social activities, terrorism, harsh climatic conditions, and seeking better employment opportunities [28,29] emphasized the causes of internal migration as regional development disparities, cultural and ethnic differences, structural changes in agriculture, mechanization, industrialization, and social and economic factors. Previous research on Türkiye generally addressed migration through themes such as development, education, health, income, and employment [30,31,32,33,34,35,36,37,38,39].

Considering population dynamics in Türkiye, the rural share of the total population (85,664,944 people as of 2024) has declined steadily, mirroring patterns in other developing countries [40]. In 1960, the rural population ratio was 68.5 percent, falling to approximately 22.5 percent today. United Nations projections suggest a further decrease to 15.9 percent by 2050 [41]. Prior to 1980, rural development relied primarily on agricultural policies and state intervention. Although the role of civil society expanded thereafter, rural development did not become a core policy priority. In the 2000s, the policy orientation shifted markedly, with direct rural development programs introduced, in part through European Union harmonization, consistent with the Sustainable Development Goals (SDGs). Until 2006, policies were largely agriculture-oriented, but subsequent approaches have been more holistic [42]. Nevertheless, implementation challenges persist, and the rural-to-urban migration pattern shows little structural change.

An examination of Türkiye’s administrative structure reveals a system based on the principle of a unitary state and built upon a duality of central administration and local governments. The country is divided into 81 provinces to ensure the spatial organization of public services. Governance at the provincial level is carried out by the Governorship (representing the decentralized provincial organization of the central administration) and by municipalities and special provincial administrations (in non-metropolitan provinces), which execute local public services. In Türkiye, a two-tiered local governance structure (metropolitan and district municipalities) is applied in the 31 provinces with metropolitan status, while a three-tiered structure (provincial municipality, district municipality, and special provincial administration) is implemented in the remaining provinces. Within this framework, the province assumes a dual administrative function, serving both as the scale for implementing central policies and as the area for coordinating local services [43].

On the other hand, in Türkiye, as in many developing countries, structural constraints in rural areas persist, and rural-to-urban migration continues to increase at similar yet varying rates. Therefore, there is a clear need for national and international policies that both reduce the propensity to migrate and holistically support rural development. In this context, this study aims to multi-dimensionally reveal the determinants of internal migration in Türkiye using panel data analysis at the 81-province level and to propose policy recommendations for rural development based on the SDGs (Sustainable Development Goals). The study specifically considers the limitations of the agricultural sector due to its direct relevance to rural areas and its potential to cause migration. In this sense, the study differentiates itself in terms of the dataset used, scope, and methodology. While existing studies frequently use regional groupings covering subsets of provinces, this research utilizes data for all 81 provinces of Türkiye. Given the substantial geographical, economic, and socio-cultural heterogeneity across provinces, this comprehensive coverage offers valuable contributions to policymaking for migration management.

2. Materials and Methods

2.1. Data

In this study, for each province annual net migration rate, agricultural gross domestic product index, industrial gross domestic product index, population (%), agricultural support amount (%), human development index, unemployment rate, terror situation data for the period 2009–2020 related to 81 provinces in Türkiye were examined. Data were obtained from various secondary sources such as the Türkiye Statistical Institute, the Ministry of Agriculture and Forestry, and Global Terrorism (

Table 1. Data and variable definition.

Variables	Explanation	Reference
M	Net migration rate	[40]
AGDP	Agricultural GDP index	[40]
IGDP	Industrial GDP index	[40]
P	Population ((Province/Country) × 100))	[40]
AS	Agricultural supports ((Support given to the province/total support) × 100))	[44]
HDI	Human development index	[45]
E	Unemployment rate (%)	[40]
T	Terror	[46]

). In this study, all statistical analyses were performed using the Stata 17 software package.

Providing explanations for the variables used in the study will facilitate the selection and understanding of the variables. In this regard, an explanation for each variable has been provided.

Net Migration Rate (M): This variable is considered the dependent variable. It expresses the net number of migrants per thousand persons who are eligible to migrate. For instance, in Tokat, a known rural city in Türkiye that experiences high out-migration, the number of people who migrated in was 24,840 and the number of people who migrated out was 41,746 in year t. Tokat’s net migration in year t, which is the difference between the number of in-migrants and out-migrants in year t, is −16,906. This value indicates that Tokat’s population decreased by 16,906 people due to internal migration in year t. According to the net migration rate, which is calculated by multiplying the ratio of the change in internal migration in a province to the province’s population by 1000, Tokat’s net migration rate in year t is −27.9. This means that in Tokat, with a net migration rate of −27.9 in year t, approximately 28 people per 1000 net migrated out of the city. At the outset of the study, the intention was to determine the village populations as the output of rural-to-urban migration. However, due to the conversion of villages into neighborhoods under the Metropolitan Municipality Law implemented in Türkiye in 2013, healthy data on village/neighborhood populations for 30 metropolitan cities could not be reliably accessed from that year onwards. Although official correspondence was exchanged several times with Turkish Statistical Institute (TurkStat) regarding this matter, it was stated that some villages had undergone name changes or were merged, which would cause significant difficulty in matching the data with the pre-2013 period. For this reason, it was decided that using the publicly available net migration rates at the provincial level would yield more realistic results as the dependent variable. In the literature, there are studies where the net migration rate data has been used as the dependent variable in macro-analyses concerning internal migration movements [31,32,33,35].

In order to reveal the factors that may influence changes in the net migration rate, province-level independent variables were obtained for the study. These are explained below.

Agricultural GDP Index and Industrial GDP Index (AGDP and IGDP): It is expected that an increase in the share of the agricultural and industrial sectors within the province’s GDP will positively affect the net migration rate. Therefore, the data for these variables were compiled on an index basis at the 81-province level [40].

Agricultural Supports (AS): An increase in total agricultural supports, as an output of rural development investments made in the region, is expected to positively affect the net migration rate. The supports included in this study comprise area-based payments, difference premium payments, livestock supports, rural development supports, supports from the Agricultural and Rural Development Support Institution, young farmer supports, insurance support, and other supports [44]. This data was calculated as the ratio of the total agricultural supports in the province to the total national support for all years at the 81-province level.

Population (P): The rate of increase or decrease in the annual population changes in the provinces has a direct link with the net migration rate. This data was calculated as the ratio of the province’s population to the total country population for all years at the 81-province level. Particularly in provinces with a high rate of urbanization, high population density will lead to an increase in opportunities such as education, health, and labor force, and consequently, will positively affect the net migration rate.

Human Development Index (HDI): The HDI is a summary of the progress achieved in the three fundamental dimensions of human development: a long and healthy life, knowledge level, and standard of living. The basic components forming the index are variables representing demography, education, and per capita income level. In this study, the human development index values prepared by Yiğiteli and Şanlı (2020) by calculating sub-indices at the 81-province level were used [45]. It is expected that as the province’s Human Development Index increases, the net migration rate will show a positive and increasing trend.

Unemployment Rate (E): An increase in the unemployment rates in the province is expected to negatively affect the net migration rate. The data covers the percentage of unemployed persons aged 15 and over from TurkStat’s labor force statistics [40].

Terror (T): An increase in terrorist incidents in the province is expected to negatively affect the province’s net migration rate. The relevant data was compiled from the Global Terrorism database [46]. The terror incidents for each province in the relevant year were examined individually. This variable was added to the panel model as a dummy variable. A value of 1 was used for provinces where no terror incidents occurred during the time period considered, 2 for provinces where 1–10 terror incidents occurred, and 3 for provinces where 11 or more terror incidents occurred. As observed in similar studies within the literature, it can be stated that the independent variables HDI, P, AS, AGDP, and IGDP were used for the first time in this study.

Furthermore, we can state that the VoSviewer analysis results presented in Figure 1 directly influenced the study’s design and variable selection. Despite the limited number of agriculture-based analyses in the existing literature, the central role of agriculture in the keyword analysis justified the study’s focus on variables such as Agricultural GDP and Agricultural Supports. The diversity of keywords (migration, rural, urban, agriculture, entrepreneurship, livelihoods, etc.) validates the core idea that migration is influenced by a comprehensive set of economic, social, environmental, and cultural factors. This led to the inclusion of multi-dimensional variables such as HDI (social factor) and Terror (political/security factor) in the model.

According to Equation (1), the net migration rate is obtained by multiplying the ratio of the internal migration change (migration inside–migration outside) in the region by the population of the region by 1000 [40]). Where $M_{.i-i.}$: Net migration rate, $m_{.i}$: In-migration, $m_{i.}$: Out-migration, $m_{.i}-m_{i.}$: Net migration, $P_{i,t+n}$: Population residing in “i” at the time “t + n”, i: The place in which migration and $k$ Constant ($k=1000$).

$$M_{.i-i.}=\left({\displaystyle \frac{m_{.i}-m_{i.}}{ \left({ P}_{i,t+n}-0.5\ast \left(m_{.i}-m_{i.}\right)\right)}}\right)\ast k$$

(1)

To visualize the provinces that are net migration receivers and net migration senders across Türkiye as a whole, and to conduct a regional assessment within the examined time period, a Migration Map was created. In Türkiye’s migration map provided in Figure 2, the Net Migration Rates (M) of the provinces are scaled within various value ranges from −25 to 20. The colors are determined starting from the provinces with the highest negative M (−20.0 to −10.0) (from lightest to darkest) towards the provinces with the highest positive M (5.01 to 20.0).

The map generally illustrates that the eastern provinces are predominantly out-migrating (net-sending), while the western provinces are net-receiving. Accordingly, Tekirdağ (M, 18.8) is understood to be the province with the highest net in-migration rate, located in the western region. It is known that the largest source of in-migration to Tekirdağ is the province of Istanbul. Conversely, Ağrı (M, −23.7), located in the eastern region, is observed to be the province with the highest net out-migration rate. It is also known that Istanbul province ranks first in terms of the destinations for Ağrı’s migration movements. Consequently, it can be stated that Istanbul (M, 1.43) is a province that both receives and sends out a significant volume of migration.

The mean was tested under descriptive statistics in terms of standard deviation, kurtosis, skewness, and data normality (

Table 2. Descriptive statistics.

	M	AGDP	IGDP	P	AS	HDI	E	T
Mean	−2.32	121.38	154.21	1.23	1.24	0.72	10.02	1.26
Std. Dev.	14.59	26.55	52.67	2.20	1.62	0.05	4.77	0.51
Max	138.50	284.80	715.40	18.66	24.15	0.82	30.90	3.00
Min	−119.00	51.70	81.30	0.09	0.02	0.55	3.50	1.00
Skewness	0.95	1.54	2.71	6.09	6.11	−0.71	1.58	1.78
Kurtosis	23.86	7.79	20.81	46.02	65.28	3.39	6.15	5.30
Jarque–Bera	17,762.06	1311.09	14,043.73	80,964.80	163,150.70	88.61	807.18	727.48
Prob.	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Obs.	972	972	972	972	972	972	972	972

). Kurtosis and skewness provide information about the nature of the variables, and Jarque–Bera is used in the calculation of test results. When evaluated in general, it was observed that the skewness values were positive and the kurtosis values were considerably higher than the desired value. Based on the results of the Jarque–Bera statistics, it was concluded that the error terms of the series did not have a normal distribution.

2.2. Empirical Model

In this study, a theoretical framework based on the expected endogenous growth model was used to explain the dependence of population, agricultural GDP index, industrial GDP index, human development index, agricultural subsidies, unemployment rate, and terror on migration (Equations (2) and (3)).

$$M_{it}=f\left(\mathrm{A}\mathrm{G}\mathrm{D}\mathrm{P},\mathrm{E}\mathrm{G}\mathrm{D}\mathrm{P},\mathrm{P},\mathrm{S},\mathrm{H}\mathrm{D}\mathrm{I},\mathrm{E},\mathrm{T}\right)$$

(2)

$$M_{it}=\alpha +{\beta }_1{AGDP}_{it}+{\beta }_2{EGDP}_{it}+{\beta }_3{P}_{it}+{\beta }_4{S}_{it}+{\beta }_5{HDI}_{it}+{\beta }_6{E}_{it}+{\beta }_7{T}_{it}+{\mu }_{it}$$

(3)

where i = 1,…N express cross-section units, t = 1,…T denotes the periods and µ the error term.

2.3. Cross-Section Dependence

Cross-sectional dependency is another factor to be considered. Cross-sectional dependence can occur as a result of unidentified common factors or spatial and distributional patterns. Neglecting cross-sectional dependency can lead to non-disclosure of dependencies of error terms and invalid test statistics, as well as loss of efficiency in estimates. On the other hand, the identification of cross-sectional dependence would help in the selection of econometric techniques to be applied to avoid misleading and inefficient statistics. In this study, the CD test developed by Pesaran [47] was preferred to test the existence of cross-sectional dependency, taking into account the N > T situation.

$$CD=\sqrt{{\displaystyle \frac{2T}{N(N-1)}}} \left({\sum }_{i=1}^{N-1}{\sum }_{j=i+1}^N{\mathrm{ṕ} }_{ij}\right)$$

(4)

Here,

$${\mathrm{ṕ} }_{ij}={\sum }_{t=1}^T{e}_{it}{e}_{jt }/ {\left({\sum }_{t=1}^T{e}_{it}\right)}^{1/2} {\left({\sum }_{t=1}^T{e}_{jt}\right)}^{1/2}$$

(5)

Test results for cross-sectional dependency of all variables are given in

Table 3. Pesaran cd test results.

Variables	Tstatistics	Probability
M	25.44	0.000
P	34.58	0.000
AGDP	103.99	0.000
IGDP	161.10	0.000
HDI	192.14	0.000
AS	6.10	0.000
E	80.10	0.000

. The hypothesis of “there is no cross-sectional dependency for each variables” was rejected.

2.4. Panel Unit Root Test

As the second step of the analysis, unit root test was applied. The decision of which unit root test is to be used depends on whether they take into account the cross-sectional dependency or not. If there is no cross-sectional dependency between the groups that make up the data set, first-generation panel unit root tests should be used, and if there is cross-sectional dependency, second- and third-generation unit root tests should be used. In Table 3, the existence of correlation between units is revealed and the Cross-Sectionally Augmented Dickey–Fuller (CADF) test developed by Pesaran [48] was used in the study (Equation (6)).

$$∆y_{i,t}=a_i+b_i{y}_{i,t-1}+c_i{y}_{t-1}+d_i∆y_t+{\epsilon }_{i,t}$$

(6)

2.5. Panel Data Analysis and Model Prediction

To decide on the appropriate model estimator, the study was conducted in four phases and can be summarized as follows:

1st stage: To determine whether there is a unit and/or time effect in panel data models;

2nd stage: To determine whether unit time effects are correlated with independent variables in the presence of unit and or time effects;

3rd stage: To test the deviations from the basic assumptions of the appropriate model after the selection of the appropriate model according to the tests at the stages 1 and 2 (fixed or random effects);

4th stage: To implement the resistant estimation model if there are deviations in the basic assumptions of the model.

3. Results and Discussion

In the selection of the appropriate panel data model, it should first be tested whether there is a cross-sectional dependence between units. Afterwards, the choice of the unit root test to be applied is determined according to whether there is a cross-sectional dependency or not. Considering the results in Table 3, it was concluded that there was a correlation between the units, in other words, a cross-sectional dependence. Accordingly, in the study, the Cross Section Extended Dickey–Fuller (CADF) unit root test, which takes into account the cross-sectional dependency developed by Pesaran, was applied. The test results are given in

Table 4. Pesaran CADF unit root test results.

Variables	T Statistics	Critical Value 10%	Critical Value 5%	Critical Value 1%	Z (t-bar)	Probability
M	−2.94	−2.00	−2.07	−2.19	−10.07	0.000
ΔP	−3.76	−2.00	−2.07	−2.19	−16.73	0.000
AGDP	−2.32	−2.00	−2.07	−2.19	−5.01	0.000
ΔIGDP	−2.77	−2.00	−2.07	−2.19	−8.66	0.000
HDI	−2.75	−2.00	−2.07	−2.19	−8.55	0.000
AS	−2.42	−2.00	−2.07	−2.19	−5.80	0.000
ΔE	−3.21	−2.00	−2.07	−2.19	−12.30	0.000

. Of the variables, Migration (M), Agricultural Gross Domestic Product (AGDP), Human Development Index (HDI), Agricultural Support (AS) were stationary at I(0) level while the variables of Population (P), Industry Gross Domestic Product (IGDP) and Unemployment (E) I(1) became stable at the 1% significance level in the first difference.

There are commonly three basic models in panel data model prediction. These are the classical model, the fixed effects model, and the random effects model. For the selection of the appropriate estimation model, the validity of the unit and time effects was tested by ANOVA F test (or constrained F) [49]. According to the test results, the H0 hypothesis, which states that there is no unit and/or time effect, is rejected at the 1% significance level, and the H1 hypothesis, which states that at least one of them has an effect, is accepted. Thus, it was decided to apply the fixed or random effects model instead of the classical model (Table 4).

The Hausman test [50] was used for the selection of fixed effects or random effects estimators. The aim of the Hausman test statistic tests whether the difference between the coefficients estimated by the fixed effects model and the coefficients estimated by the random effects model is statistically significant [51]. Hausman test was used in the study and the test results are given in

Table 5. Pre-tests for prediction method preference for panel data model.

ANOVA F test	F (80,803)	1.66
	Prob > F	0.000
Hausman test	chi2	8.83
	Prob > chi2	0.066

. In the model established according to the results of the Hausman test, H is considered 0 because p > 0.05. In this case, it can be stated that the method of random effects would be appropriate in the estimation of the model.

In the presence of heterogeneity, in other words, variable variance and autocorrelation problems, alternative estimation models should be applied. In this context, it is important to test the heterogeneity and autocorrelation for the random effects model. As a matter of fact, the existence of these factors would necessitate the use of a different estimator [49,52]. In the random effects model, variance homogeneity was checked using Levene [53], Brown and Forsythe [54] test, and presence of an autocorrelation was checked using the Lagrangian Multiplier and Adjusted Lagrange Multiplier tests (

Table 6. Varying variance and autocorrelation tests for random effects model.

	Test	Statistic	Probability
Levene, Brown and Forsythe tests	W0	6.097	0.000
	W50	5.058	0.000
	W10	5.680	0.000
Breusch–Pagan Lagrange LM-ALM test	LM	18.510	0.000
	ALM	51.220	0.000

).

Based on the test statistics (W0, W50, W10) (80,810) of Levene, Brown and Forsythe, the H₀ hypothesis, which is established as “there is no variable variance problem” is rejected. Thus, it was concluded that there was a problem of varying variance in the model. Similarly, when the LM and ALM tests were examined on the results of the Lagrange Multiplier and Adjusted Lagrange Multiplier test results, which were established based on the null hypothesis for the autocorrelation test, it was determined that both the random unit effect and the autocorrelation coefficient were equal to 0, that the H₀ hypothesis was rejected and that autocorrelation existed. As a result of the tests performed in the study, it was revealed that there was a problem of varying variance and autocorrelation (Table 6). For this reason, Arellano, Froot and Rogers [55,56,57] standard robust error estimators were used in the estimation of the model, with which we can predict in such a way that varying variance and autocorrelation-resistant (robust) standard errors are obtained. The random effects model predicted using Arellano, Froot and Rogers [55,56,57] robust standard errors estimator to reveal the factors affecting interprovincial migration in Türkiye is given in

Table 7. Results of Arellano–Froot–Rogers Predictor.

Dependent Variable: M
Variables	Coefficient	Robust S.E.	z Statistic	p > |z|
ΔP	171.13	85.06	2.01	0.044
HDI	101.60	9.67	10.50	0.000
AGDP	−0.04	0.02	−2.15	0.032
ΔIGDP	−0.03	0.03	−1.14	0.253
ΔE	−0.12	0.105	−1.19	0.233
AS	0.41	0.286	1.41	0.155
T	−2.62	0.78	−3.36	0.001
c	−66.72	6.28	−10.62	0.000
R2= 0.65	Wald (chi2) = 200.58		Prob > chi2
0.0000

. The results indicated that the probability value of the F statistic was significant (p < 0.01) and that the R2 value, which indicates the goodness of fit, was at a high level of 0.65 for this model. According to the results of the model, population, human development index, agricultural GDP index and terror incidents were effective and significant factors for the net migration rate.

As expected, the population turned out to be a factor that positively affected the net migration rate. The coefficient of this variable in the model was significant (p < 0.05). As the proportion of the provincial population in the total country population increased, the net migration rate also increased positively. In other words, the province found itself in a position where it received migration rather than sends migration. Although the industry variable was not a significant factor affecting the net migration rate in the model, it could be stated that the population growth rate in the province took shape depending on the development in industry and other sectors. Especially in provinces with a high urbanization rate, the high population density causes an increase in opportunities such as education, health, and labor force, and ultimately positively affects the net migration rate. In addition to all these, since the provinces with a high urbanization and industrialization rates in Türkiye are mainly located in the Marmara and Aegean Regions, the population movement towards these regions is constantly active. As a matter of fact, according to the average net migration rate figures between 2008 and 2020 in Türkiye, 52% of the 31 provinces that receive migration were provinces in the Marmara and Aegean Regions. These are all the provinces of the Marmara Region ((Tekirdağ (18.80), Yalova (14.32), Kocaeli (10.69), Çanakkale (8.12), Bursa (6.01), Sakarya (5.04), Bilecik (2.55), Kırklareli (3.48), İstanbul (1.43), Balıkesir (3.39) and Edirne (0.30)) and five of the eight provinces in the Aegean Region ((Muğla (9.64), İzmir (4.51), Aydın (4.91), Denizli (1.35) and Manisa (0.52)). In addition to these, Antalya (10.58), Eskişehir (8.67), Ankara (6.12), Bolu (5.47), Sinop (2.92) and Karabük (2.84) were also among the provinces that take a high rate of migration. These provinces were generally the developed ones in terms of organized industrial facilities, tourism, health, education and security and therefore they provide employment, which makes migration to these cities attractive. A study determining the socio-economic factors influencing internal migration in Türkiye found that as the young population and the population with a high school degree increase, provinces tend to both send and receive more migration; however, as the university graduate population increases, the out-migration from provinces decreases while the in-migration increases [58].

It was determined that the human development index had a positive and statistically significant effect on the net migration rate at the 1% level. Human development can be broadly defined as the advancement of knowledge and living standards throughout a long and healthy life. According to the results of the analysis, a one-unit change in the HDI could lead to a 101.60 unit rise in the net migration rate. These findings are consistent with the results reported by a study in Türkiye. Their study revealed a positive relationship between per capita income levels above the national average and the increase in employment in regions receiving migration. Furthermore, the research emphasized that underdevelopment in the areas of health and education constituted the most critical problem for out-migration regions. In this context, a significant and expected negative relationship was identified between the number of hospital beds and the level of outward migration from a region [35]. On the other hand, a different study determined that both employment and being a high school graduate had a mitigating effect on the net migration rate [39]. A study identifying the socio-economic factors influencing internal migration in Türkiye found a negative relationship between the provinces’ own income levels and their out-migration. A 1% increase in the provincial level national income was found to increase out-migration from the province by 40% [58]. Numerous studies have shown that in migration decisions, access to education and health services, better housing conditions, and familial ties (dependent migration) are determining factors alongside economic factors. Patterns of young people relocating for educational purposes and family members following household heads based on internal ties (chain migration) have been frequently reported. TurkStat’s internal migration statistics data also indicate that reasons such as education and housing/living conditions stand out among the causes of migration [59]. Another variable with a substantial influence on migration is agricultural GDP. The results of the model indicated a negative relationship between agricultural income and the net migration rate. The outcome for Türkiye can be attributed to three main factors; (a) the gradual decline in the share of the agricultural sector in GDP, as seen in many developing countries, (b) the expansion of other sectors leading to labor flows from rural to urban areas, and (c) the migration decisions of rural youth driven by income, security, prestige and emotional motivations. Indeed, numerous studies have documented a global decline in young people’s interest in agricultural occupations, particularly in Asian, African and Latin American contexts [60,61,62]. These studies highlight that non-agricultural careers are perceived as less demanding and more rewarding as they offer higher wages and greater stability [63]. Similar findings were also reported for Türkiye [35]. In another study investigating the relationship between internal migration and income inequality in Türkiye, it was found that income inequality increased the net migration rate [39]. Terrorism is another critical factor affecting migration dynamics, with potential adverse impacts on economic growth, social welfare, and employment. Terrorist incidents first emerged in the Eastern and Southeastern Anatolia regions of Türkiye during the 1980s and intensifies throughout the 1990s. It was reported that 7.5% of the migration from these regions to the safer Western provinces between 1986 and 1990, and 14.2% between 1991 and 1995, occurred due to security concerns [64]. In the present study, as expected, a negative relationship was observed between terrorist incidents and the net migration rate. The statistical results indicate that a one unit increase in terrorism incidents would reduce the net migration rate by 2.62 units, suggesting that terrorism acts as a significant push factor for migration. Similar conclusions have been reached in other studies examining the nexus between migration and terrorism in Türkiye [32,33,34]. Furthermore, a network-based analysis conducted by Gürsoy and Badur (2022) demonstrated that inter-provincial migration both forms regional clusters and indicates that major centers (e.g., Istanbul, Ankara, İzmir) receive migration from various directions [37]. In parallel, there has been reported continuity of annual migration volumes and the emergence of some new counter-flows using dynamic flow models across approximately 25 thousand migration events at the 81-province level. In the study utilizing multinomial regression analysis, significant determinants influencing migration included population, economic welfare, and the spatial and political distance between the origin and destination of migration, while the ratio of ethnic minorities in a province was found to have no positive relationship with internal and external migration [38].

This study offers significant contributions to understanding the dynamics of rural-to-urban migration mobility in Türkiye. Nevertheless, certain limitations must be considered when interpreting the findings. Firstly, the analysis relies on aggregate province-level data. The use of this aggregated data may obscure important local heterogeneities and migration dynamics at the district, neighborhood, or individual household level. Secondly, some variables included in the model, such as the data representing political and social dynamics like Terrorism (T) incidents and comprehensive indices like the Human Development Index (HDI), are difficult to measure with absolute certainty at the provincial level. Finally, this panel data analysis does not distinguish between different types of migrants (youth and older adults, temporary and permanent migrants). However, the migration of young people for career and educational purposes and the migration of older adults for social services or return are driven by different push and pull factors [65]. Future research could use micro-level datasets to overcome these limitations and apply more sophisticated dynamic models by disaggregating migration by type.

4. Conclusions

As in many developing countries, migration mobility in Türkiye continues predominantly from rural to urban areas. Factors such as industrialization, employment opportunities, education, health, and social opportunities make urban centers attractive. In the long term, sustaining this trend will adversely affect both urban and rural welfare and pose a significant barrier to achieving sustainable development. This concern is closely aligned with SDG targets such as SDG 11.1 (safe and affordable housing) and SDG 8.5 (full and productive employment). For this reason, it is essential to make rural life more attractive and to reduce excessive urban population density. The findings show that human development indicators are among the most influential determinants of rural–urban migration mobility.

Table 8. Alignment of Empirical Findings with Sustainable Development Goals (SDGs).

Empirical Finding	Effect Direction	Relevant SDG Targets	Policy Implication (Reducing Migration Pressure)
Human Development Index (HDI)	Positive and Highly Significant	SDG 3: Good Health and Well-being; SDG 4: Quality Education; SDG 10: Reduced Inequalities	Improving access to health and education services in rural areas enhances quality of life, aligning with SDG 3.8 (Universal Health Coverage) and SDG 4.A (Education Facilities) to potentially trigger reverse migration.
Agricultural GDP (AGDP)	Negative and Significant	SDG 2: Zero Hunger; SDG 8: Decent Work and Economic Growth	Declining agricultural income increases rural out-migration. This is directly linked to SDG 2.3 (doubling agricultural productivity and small-scale food producers’ incomes). Policies must stabilize incomes, improve land access, and introduce digital solutions (SKA 2.a) to boost rural labor productivity
Terror (T)	Negative and Highly Significant	SDG 16: Peace, Justice and Strong Institutions	Terrorism acts as a strong push factor for out-migration, reducing the net migration rate. Achieving SDG 16.1 (significantly reducing all forms of violence) by ensuring regional security and promoting social cohesion is fundamental to limiting forced migration.
Population (P)	Positive (Net In-Migration)	SDG 11: Sustainable Cities and Communities	High population density in industrialized provinces creates pressure on infrastructure and resources. Managing this influx requires policy alignment with SDG 11.1 (safe and affordable housing) and SDG 11.3 (inclusive and sustainable urbanization).

has been prepared to ensure a clearer visualization of the alignment between the empirical findings and the Sustainable Development Goals (SDGs).

The positive impact of the HDI demonstrates that social infrastructure, particularly education, healthcare, and income equality, directly affects migration preferences. While economic drivers remain strong, non-economic dimensions such as accessibility to social services and regional security significantly shape population movements. Therefore, enhancing HDI in rural regions could trigger reverse migration by improving quality of life and employment diversity. Such improvements could also reinforce SDG 10.2, which aims to promote the social, economic, and political inclusion of all individuals.

Rural development policy in Türkiye has traditionally been managed by the Ministry of Agriculture and Forestry. However, effective rural transformation now requires a multi institutional structure involving Ministry of Environment, Urbanization and Climate Change, the Ministry of Youth and Sports, and the Ministry of Family and Social Services. Current programs such as the IPARD II (Instrument for Pre-Accession Assistance in Rural Development) and the LEADER approach implemented under EU harmonization provide strong frameworks for community led local development. These programs embody SDG 17 (partnerships for goals) by fostering collaboration among public institutions, local authorities, and non-governmental organizations. Institutional coordination remains a critical determinant of policy effectiveness. Rather than isolated efforts, Türkiye’s National Rural Development Strategy (2024–2028) emphasizes an integrated governance model. This model aims to improve inter-ministerial collaboration, fiscal efficiency, and data driven decision making. Public–private partnership and regional development agencies also play vital roles in implementing SDG, aligned projects across provinces.

The decline in young rural population has significant implications for sustainability of agricultural production. This demographic shift increases dependency on aging farmers and reduces innovation capacity in agriculture. Policies addressing youth retention should focus on income stabilization, land access, and technology adoption. Revised targets under SDG 2.3 (agricultural productivity), SDG 6.6 (water ecosystem protection), and SDG 13.1 (climate resilience) can serve as guiding frameworks. Moreover, integrating climate smart agriculture and renewable energy solutions aligns with SDG 7.3 (energy efficiency) and SDG 15.3 (land degradation neutrality).

To attract and retain young farmers, Türkiye must scale up initiatives such as Young Farmer Grant Program and IPARD III, which offer financial support for innovative and sustainable agricultural enterprises. Digital transformation in agriculture, including precision farming and smart irrigation systems, can further reduce migration pressure by enhancing productivity. Capacity building programs should also strengthen agricultural entrepreneurship and technological literacy in rural schools and vocational institutions.

The integration of national and international cooperation mechanisms is indispensable for sustainable rural development. The Ministry of Agriculture and Forestry, in partnership with international organizations such as FAO, OECD and the European Union, should coordinate projects that combine environmental stewardship, agricultural modernization, and rural livelihood diversification. Such alignment ensures coherence with the UN’s 2020 Agenda and reinforces SDG 17.14 on policy coherence for sustainable development. Overall, this study highlights the necessity of cross sectoral collaboration and global partnership to ensure that migration contributes positively to sustainable development rather than exacerbating regional disparities.