Social Security Transfers and Fiscal Sustainability in Turkey: Evidence from 1984–2024

Abstract

Social security systems constitute a structurally significant component of public finance in developing economies and often generate persistent fiscal pressures through budgetary transfers. Demographic transformation, widespread informality in labor markets, and weaknesses in contribution-based financing increase the dependence of social security systems on public resources. The objective of this study is to examine whether budget transfers to the social security system affect fiscal sustainability in Turkey by analyzing their relationship with the budget deficit and the public sector borrowing requirement. The analysis employs annual data for Turkey covering the period of 1984–2024. A comprehensive time-series econometric framework is adopted, incorporating conventional and structural-break unit root tests, the ARDL bounds testing approach with error correction modeling, and the Toda–Yamamoto causality method. The empirical findings provide evidence of a stable long-run relationship among the variables. The results indicate that social security budget transfers exert a statistically significant and persistent effect on the public sector borrowing requirement, while no direct long-run effect on the headline budget deficit is detected. Causality results further confirm that fiscal pressures associated with social security financing materialize primarily through borrowing dynamics rather than short-term budgetary imbalances. By explicitly modelling social security budget transfers as an independent fiscal channel over a long historical horizon, this study contributes to the literature by offering new empirical insights into the fiscal sustainability implications of social security financing in Turkey. The findings also provide policy-relevant evidence for developing economies facing similar institutional, demographic, and fiscal challenges.

1. Introduction

In accordance with the principle of the social state, governments bear the fundamental responsibility of establishing both short- and long-term protection mechanisms to safeguard citizens against social and economic risks. Among these mechanisms, the social security system plays a pivotal role by providing protection against unemployment, disability, old age, illness, and occupational accidents. Social security systems are designed to meet the essential needs of individuals and their dependents throughout both their working lives and retirement periods. The financial sustainability of such systems traditionally relies on contributions collected from employees and employers, forming the backbone of their long-term viability.

However, particularly in underdeveloped and developing economies, social security systems frequently encounter structural and administrative challenges that undermine their financial balance. Insufficient institutional capacity, weak enforcement mechanisms, and the pervasive influence of political authority over bureaucratic structures often reduce system efficiency and effectiveness. These deficiencies contribute to persistent revenue–expenditure mismatches, forcing governments to rely increasingly on public budget transfers to sustain social security operations.

In Turkey, social security activities were conducted through three separate institutions prior to 2006: the Social Insurance Institution; the Tradesmen, Craftsmen, and Other Self-Employed Social Insurance Institution; and the Pension Fund. While the first two institutions were financed mainly through employee and employer contributions, the Pension Fund relied directly on transfers from the central government budget. This fragmented structure, combined with limited institutional capacity, resulted in chronic financial imbalances and rising dependence on public resources. Despite the enactment of the Unemployment Insurance Law No. 4447 in 1999 (Law No. 4447, 1999), fiscal pressures persisted, with transfers from the general budget reaching approximately 4.5% of GDP, as measured by the overall budget deficit-to-GDP ratio, a level well above the Maastricht reference value of 3% of GDP.

To address these challenges, a comprehensive reform was implemented in 2006 through the enactment of Law No. 5502 (Law No. 5502, 2006), consolidating the three institutions under a unified Social Security Institution (SSI). This reform was further reinforced by the Social Insurance and General Health Insurance Law No. 5510 (Law No. 5510, 2006), which entered into force in 2008, and subsequent restructuring initiatives, including Law No. 7143 in 2018 (Law No. 7143, 2018). Although these reforms expanded coverage to more than 80% of the population, they failed to fully resolve the system’s structural imbalances, leaving fiscal sustainability concerns largely unaddressed.

Turkey is considered an appropriate case for analyzing fiscal sustainability challenges commonly faced by developing economies due to several structural characteristics. Demographically, Turkey exhibits a population structure that combines relatively high labor force participation with increasing aging pressures, which is a typical feature of many emerging economies. In the labor market, the persistence of informality limits contribution-based financing of the social security system and increases reliance on budget transfers. From a fiscal perspective, social security institutions operate within a framework that is formally autonomous yet structurally dependent on central government support, reflecting a fiscal architecture widely observed in developing economies. Within this institutional setting, social security transfers function as a fiscal transmission mechanism by influencing public sector borrowing requirements through additional financing needs, even when their direct effect on the headline budget deficit remains limited. In this respect, the Turkish case provides a relevant empirical setting for examining how social security financing interacts with fiscal sustainability in developing economies.

Against this institutional and structural background, the literature examining social security budget deficits in Turkey has generally evolved along two principal lines of inquiry: the determinants of deficits and their macroeconomic repercussions. The first strand emphasizes the destabilizing effects of informal employment and structural revenue–expenditure imbalances. Studies by Demir and Canbay (2013), Çavuş (2016), Dağ (2019), and Yıldırım and Bülbül (2019) consistently highlight how informality, weak auditing mechanisms, and rising pension and health expenditures erode the fiscal balance of the social security system. These findings converge on the conclusion that institutional inefficiencies play a central role in sustaining chronic deficits.

More broadly, recent risk-focused studies highlight that institutional design and risk perception play a crucial role in shaping the sustainability of financial protection mechanisms, particularly where public guarantees or transfers are involved (Pavia et al., 2021).

The second strand of the literature focuses on the broader fiscal and growth implications of social security financing. Early evidence from Duggan (1991) shows that persistent program deficits can accelerate public debt dynamics under fiscal constraints. Cross-country studies by Alan et al. (2010) and Tiberto and de Mendonça (2013) document long-run relationships between social expenditures, fiscal balance, and economic growth, while Turkish studies by Yılmaz (2014), Cural (2016), Altunöz (2017), and Organ and Yavuz (2017) demonstrate that transfers to the SSI significantly increase public expenditures, deteriorate fiscal balance, and expand public debt stocks.

More recent international research since 2020 has increasingly focused on the fiscal implications of social security systems by linking demographic dynamics, pension expenditures, and debt sustainability. Symeonidis et al. (2021) introduces the concept of implicit pension debt, emphasizing future liabilities as a major source of fiscal vulnerability. Park (2024) empirically shows that institutional effectiveness moderates the impact of aging and social security deficits on public borrowing requirements, while Darvas and Welslau (2025) and Díaz-Giménez and Díaz-Saavedra (2025) provide further evidence that demographic transitions and pension reforms pose significant risks to debt sustainability.

Recent methodological contributions emphasize that fiscal and macroeconomic relationships may vary across the distribution of variables and exhibit asymmetric or tail-dependent dynamics. In this context, Ul-Durar et al. (2025) provide a comprehensive survey of quantile regression applications, highlighting how distribution-sensitive approaches uncover heterogeneous effects that remain hidden in mean-based estimations. While such approaches offer valuable insights into distributional dynamics, the present study adopts a time-series framework focusing on long-run equilibrium relationships and causal interactions over an extended historical horizon. By concentrating on average long-run dynamics rather than distributional heterogeneity, this study complements the existing literature and contributes to a clearer understanding of the structural fiscal transmission mechanisms linking social security financing and fiscal sustainability in Turkey.

Despite the extensive literature on fiscal sustainability and social security financing, relatively few studies explicitly isolate social security transfers as a distinct fiscal channel while simultaneously examining their interaction with both the budget deficit and the public sector borrowing requirement over a long historical horizon. Most empirical studies tend to treat social security transfers as part of aggregate government expenditure, rather than modeling their fiscal interactions explicitly. This limitation is particularly relevant in developing economies, where budget transfers to social security institutions constitute a structural and persistent component of public finance, yet empirical frameworks jointly analyzing these relationships remain scarce.

In the Turkish context, despite the availability of extensive official data, comprehensive econometric studies that integrate social security transfers, budget deficits, and public sector borrowing requirements within a unified analytical framework are limited. Existing research largely focuses on partial relationships or shorter time horizons, leaving the long-run fiscal dynamics insufficiently explored. This study aims to fill this gap.

Building on the findings of Cural (2016), which constitute the empirical foundation of this study, the primary objective of the present research is to examine whether transfers to the social security system in Turkey affect the budget deficit and the public sector borrowing requirement. Covering the period of 1984–2024, the analysis provides a historical depth that has not been previously addressed in the literature, thereby offering a novel contribution to the assessment of fiscal sustainability in developing economies. In this context, the study contributes to the existing literature in several important ways. First, it explicitly models social security budget transfers as an independent fiscal channel rather than treating them as part of aggregate government expenditure, allowing for a more precise identification of fiscal transmission mechanisms. Second, by adopting an integrated econometric framework that jointly employs structural break unit root tests, the ARDL bounds testing approach, and the Toda–Yamamoto causality methodology, the study enhances the robustness of its empirical findings. Finally, by simultaneously analyzing the budget deficit and the public sector borrowing requirement, the paper provides a more comprehensive evaluation of fiscal sustainability compared to studies focusing on a single fiscal indicator.

Within this framework, the study is organized as follows. The subsequent section, titled Theoretical Insights to Social Security Financing, discusses the financing structure of social security systems and their relationship with budget balance and public sector borrowing requirements within a theoretical framework, supported by key fiscal and demographic indicators specific to Turkey. The next section presents the data set and variable definitions. The econometric methodology section elaborates on unit root tests, structural break stationarity analyses, the ARDL bounds testing approach and error correction model, and the Toda–Yamamoto causality framework. Empirical findings are then reported and discussed in comparison with the existing literature. The final section summarizes the main conclusions, outlines policy implications for Turkey, and offers directions for future research.

2. Theoretical Insights into Social Security Financing

Social security systems are designed to provide individuals with protection in advance against socio-economic, physiological, and occupational risks that may arise over the course of their lives. The International Labour Organization emphasizes that the right to social protection constitutes one of the fundamental human rights (ILO, 2016, p. xv). In this respect, establishing and financing a sound and sustainable social security system capable of shielding individuals from future risks is of critical importance. This issue becomes particularly salient in countries that embrace the principle of the social state, where the role and responsibility of the government are more pronounced. In Turkey, this responsibility is explicitly defined in Article 60 of the Constitution of the Republic of Turkey, which assigns the state the duty of taking necessary measures to ensure social security and of establishing the institutional framework required for its provision.

2.1. Financing Methods

It is observed that countries adopt different financing techniques within their social security systems. These techniques are generally classified as the funded pension system (funded system), the pay-as-you-go system (PAYGO), and mixed systems. The funded pension system is based on the principle of accumulating funds in advance to cover potential losses arising from social risks that may be encountered in the future. Fund accumulation may take either an individual or a collective form.

Under individual fund accumulation, an account is opened in the name of each insured person, and the contributions paid by the insured individual and the employer are deposited into this account. When the insured person experiences a loss, benefit payments are made by the institution directly from the accumulated funds in that account. However, this system is generally not preferred, as it is not compatible with the objective of ensuring social solidarity through the distribution of financial risk among individuals.

In contrast, collective fund accumulation involves pooling the contributions collected from all insured persons and their employers into a common fund. When a loss occurs, payments to the insured individual are made from this collective fund (Güzel et al., 2016, p. 65; Arıcı, 2015, p. 30). The collective fund accumulation method is generally applied in the financing of long-term insurance branches, such as invalidity, old age, and survivors’ insurance.

Under the pay-as-you-go method, contribution revenues collected within a given period are used to finance social security payments made during the same period. Accordingly, benefits paid to passive insured persons (pensioners) are financed through the contributions of active insured persons who are currently employed and contributing to the system. In this system, there is no requirement to establish or operate a separate fund. The effective functioning of the PAYGO system depends on maintaining an optimal balance between the number of active insured persons who work and pay contributions and the number of passive insured persons who receive benefits (Tuncay & Ekmekçi, 2019, p. 171). This method is generally preferred for the financing of short-term insurance branches, such as sickness insurance, work accident and occupational disease insurance, and maternity insurance.

The financing method implemented by the Social Security Institution (SSI), in accordance with legislative regulations, exhibits a mixed structure that is primarily based on the funded pension system. However, due to the financial difficulties experienced by the institution in recent years, it appears that the system has de facto shifted toward a pay-as-you-go structure, largely as a result of increased treasury transfers (Tuncay & Ekmekçi, 2019, p. 172; Boyacıoğlu & Öçal, 2018, p. 918; Güneş & Yakar, 2004, p. 135).

2.2. Funding Sources

The Social Security Institution requires adequate financial resources in order to fulfill its social protection obligations and ensure the continuous functioning of the system. The primary source of these financial resources is contribution revenue. However, since it is not feasible to operate the system solely on the basis of contribution income, state subsidies become inevitable (Sözer, 2017, p. 203). In addition to contributions, revenues generated from the Institution’s existing funds, interest income, rental income derived from immovable properties owned by the Institution, and related tax revenues also constitute important financing sources (Ahmadova & Yaslıdağ, 2018, p. 379).

The Social Security Institution employs different instruments in the provision of social security services, and the financing sources of these instruments vary accordingly. These instruments can be broadly classified as social insurance, social assistance, and social services. Social insurance schemes are primarily financed through contribution revenues paid by employees, employers, and individuals participating in the optional insurance system. In this framework, the state also contributes to contribution revenues in certain branches, such as unemployment insurance (Sözer, 2017, p. 203; Üçışık, 2015, p. 273). Moreover, the state plays a significant role in covering administrative expenditures by providing budgetary transfers aimed at alleviating the financial difficulties that lead to payment imbalances within the Institution. In contrast, social assistance and social services are financed entirely through the general budget.

2.3. Financing Problems

The financial problems of the Social Security Institution (SSI) are of critical importance due to their broad implications for the overall economic structure. Although the fundamental functions of social security systems remain largely unchanged, a wide range of factors—such as continuously evolving demographic characteristics, developments in healthcare services, increased life expectancy, changes in labor market structures, macroeconomic conditions, government interventions, and shifts in household subsistence levels—exert significant influence on the system. The dynamic nature of these factors necessitates continuous monitoring and evaluation of social security policies and their economic and social consequences (Brown et al., 2020, p. 31).

In this regard, establishing a reliable institutional framework is essential to ensure that the system can adapt to changing socioeconomic and demographic conditions and to prevent the pension system from being instrumentalized for political purposes (Vidal-Meliá et al., 2009). From an operational risk perspective, weaknesses in governance, data integration, and process transparency significantly undermine system resilience, increasing long-term fiscal exposure a pattern also observed in European insurance systems undergoing digital transformation (Grima et al., 2021). One of the most fundamental issues identified in assessing the problems faced by the SSI is the persistence of financial deficits. As shown in Table 1, the Institution recorded a substantial deficit of 67,469,343 TL in 2020. In 2021, the deficit amounted to 21,613,013 TL. The ratio of revenues to expenditures declined to 87.5% in 2020, representing the lowest level observed over the last five years. By contrast, this ratio increased to 99.7% in 2024.

Table 1. Social Security Revenues, Expenditures, and Revenue-to-Expenditure Ratios Including Treasury Transfers (2008–2025).

Year	Revenues						Thousand TL
	Premium Revenues	State Contribution	Other Revenues + Payments Without Premium	Total	Rate of Increase (%)	Expenditures	Rate of Increase (%)	Deficit	Compensation Rate (%)
2008	54,546,453	1,718,521	10,992,510	67,257,484	18.3	93,159,462	13.7	−25,901,978	72.2
2009	54,579,182	10,879,293	12,614,312	78,072,788	16.1	106,775,443	14.6	−28,702,655	73.1
2010	66,912,858	15,170,035	13,190,291	95,273,183	22.0	121,997,301	14.3	−26,724,118	78.1
2011	89,560,568	21,176,053	13,743,319	124,479,940	30.7	140,715,252	15.3	−16,235,312	88.5
2012	99,359,243	23,537,149	20,032,113	142,928,505	14.8	160,223,453	13.9	−17,294,948	89.2
2013	118,728,578	27,471,369	16,813,608	163,013,555	14.1	182,688,916	14.0	−19,675,361	89.2
2014	135,238,559	30,512,184	18,578,189	184,328,932	13.1	204,400,437	11.9	−20,071,505	90.2
2015	159,480,052	37,526,042	23,096,254	220,102,348	19.4	231,545,969	13.3	−11,443,621	95.1
2016	184,445,994	46,457,369	24,976,820	255,880,183	16.3	276,535,786	19.4	−20,655,602	92.5
2017	208,064,459	51,767,064	28,728,319	288,559,843	12.8	312,734,591	13.1	−24,174,748	92.3
2018	255,619,125	57,560,198	56,031,926	369,211,250	27.9	384,961,895	23.1	−15,750,645	95.9
2019	293,828,025	71,221,979	59,178,306	424,228,310	14.9	464,172,903	20.6	−39,944,593	91.4
2020	323,181,066	85,070,016	64,374,826	472,625,908	11.4	540,095,250	16.4	−67,469,343	87.5
2021	440,130,381	106,847,991	86,228,245	633,206,617	33.9	654,819,630	21.2	−21,613,013	96.7
2022	724,107,544	174,712,112	101,259,255	1,000,078,911	57.9	1,039,811,397	58.8	−39,732,486	96.2
2023	1,497,925,469	325,629,773	369,399,398	2,192,954,640	119.3	2,232,339,177	114.7	−39,384,537	98.2
2024	2,810,531,587	637,657,417	514,277,504	3,962,466,508	80.7	3,972,816,384	78.0	−10,349,876	99.7
2025 (January–July)	2,165,384,263	477,978,460	401,671,628	3,045,034,351		3,110,762,418		−65,728,066	97.9

Source: SSI Monthly Statistical Bulletin (July 2025) (SSI, 2025).

The level values reported in Table 1 indicate a pronounced upward trend in social security transfers, particularly in the post-2000 period, reflecting both demographic pressures and the increasing fiscal burden of the social security system. When considered in relative terms, the growing magnitude of these transfers signals a gradual strengthening of their role within the overall fiscal framework.

However, despite this apparent improvement, Table 1 indicates that the revenue-to-expenditure compensation ratio reached 99.7% in 2024; this figure should be interpreted with caution. The observed improvement does not signify a strengthening of the contribution-based financing structure of the Social Security Institution. Rather, it is largely attributable to a substantial increase in Treasury transfers and other budgetary supports recorded on the revenue side. In particular, extraordinary fiscal measures implemented during the 2023–2024 period, including the Early Retirement Scheme (EYT), significantly intensified expenditure pressures, which were largely offset through increased central government transfers. Consequently, the high compensation ratio reported for 2024 reflects an accounting outcome supported by public finance intervention rather than an autonomous improvement in the actuarial balance of the system.

Table 2 highlights the growth dynamics of the variables and reveals that social security transfers exhibit higher volatility compared to the budget deficit, especially during periods of macroeconomic adjustment and institutional reform. This pattern suggests that transfer policies often respond to structural imbalances within the social security system rather than short-term cyclical fluctuations. In contrast, borrowing requirements display more persistent growth behavior, indicating that financing needs tend to accumulate over time.

Table 2. SSI Public Borrowing Requirement (2008–2025).

Year	Public Borrowing Requirement
2008	52,599
2009	−546,464
2010	−345,527
2011	−319,942
2012	91,086
2013	5,127,898
2014	6,330,647
2015	9,231,822
2016	18,583,350
2017	16,744,447
2018	21,308,919
2019	21,112,535
2020	17,380,637
2021	34,651,851
2022	44,385,195
2023	98,947,368
2024	248,760,295
2025	305,412,197

Source: Presidency of Strategy and Budget (2025).

Consistent with these growth dynamics, according to Table 2, the public borrowing requirement of the Institution has continued to rise steadily since 2012. It is widely acknowledged that persistent deficits of this nature increase public borrowing needs, thereby exerting upward pressure on interest and inflation rates and creating challenges for the realization of investment activities and sustainable development objectives. Such dynamics are also emphasized as contributing to higher unemployment rates and a deterioration in income distribution (Alper, 2011, p. 36).

From a comparative perspective, Erdoğan (2018), in a comparative study of the European Union and Turkey with respect to social security expenditures, concludes that the ratio of expenditures financed through contribution-based social security systems and insurance branches to gross domestic product (GDP) is higher in European Union member states than in Turkey. According to OECD (2019) data, social security expenditures account for approximately 20% of GDP in OECD countries, while Eurostat (2019) data report this ratio as 28.1% for the European Union and 12.5% for Turkey. In this context, it is argued that social security expenditures in Turkey need to be increased. Accordingly, policy measures such as enhancing contribution revenues, implementing pension reforms, reducing unemployment, and maintaining actuarial balance gain particular importance.

Table 3 further illustrates changes across subperiods, showing that the share and persistence of social security-related financing pressures increase in later periods. These shifts coincide with rising public sector borrowing requirements, underscoring the indirect transmission channel through which social security financing affects fiscal sustainability.

Table 3. Budget Transfers to the Social Security Institution (2008–2025).

Year	Budget Transfers	Rate of Change (%)	Rate of GDP (%)
2008	35,016,403	5.9	3.49
2009	52,599,691	50.2	5.23
2010	55,244,258	5	4.73
2011	52,772,218	−4.5	3.76
2012	58,728,293	11.3	3.71
2013	71,263,763	21.3	3.91
2014	77,335,890	8.5	3.76
2015	79,038,817	2.2	3.36
2016	108,073,487	36.7	4.11
2017	128,182,906	18.6	4.09
2018	150,530,375	17.4	4
2019	196,797,506	30.7	4.56
2020	248,846,194	26.4	4.93
2021	252,083,979	1.3	3.5
2022	389,030,604	54.3	2.54
2023	849,844,962	118.5	3.14
2024	1,442,632,708	69.8	3.24
2025 (January–July)	1,125,514,776

Source: SSI Monthly Statistics Bulletin (July 2025) (SSI, 2025).

Building on these subperiod dynamics, it becomes necessary to identify the factors that lead the Social Security Institution to operate with persistent financial deficits, as these deficits are largely covered through transfers made by the Treasury from the general budget. As illustrated in Table 3, budget transfers to the Institution have exhibited a continuous upward trend since 2013. Such transfers contribute to the emergence of deficits in the general budget and, consequently, to an increase in the public borrowing requirement (Yurdadoğ et al., 2019, p. 657). However, the existing literature emphasizes that financing the social security system solely through revenues collected from insured individuals, employers, and employees is insufficient. In Turkey, as in many other countries, there is a clear need to activate alternative financing sources and to enhance the role of the state as an effective contributor to the financing structure of the system, particularly within a framework that preserves fiscal sustainability (Tuncay & Ekmekçi, 2019, p. 170; Güzel et al., 2016, p. 64; Şakar, 2017, p. 66; Özmen, 2019, p. 392).

It has become imperative to implement measures aimed at eliminating the revenue–expenditure imbalance of the Social Security Institution. A range of structural and administrative factors—including unregistered employment, unemployment, managerial inefficiencies, a decline in the active-to-passive insured ratio, problems in contribution collection, difficulties in fund management and valuation, a lack of social security awareness within society, and population aging—contribute to persistent budget deficits (Bulut, 2019, p. 26). Briefly addressing these issues is therefore useful for accurately identifying the underlying causes of the problem and for developing effective policy responses.

One of the most significant financing problems of the social security system in Turkey is unregistered employment. Table 4 presents unregistered employment rates across years and sectors.

Table 4. Unregistered Employment Rates by Sectors.

Year	Agriculture	Non-Agricultural	Industry	Service	Construction	General
2002	90.14	31.74	36.4	29.19	-	52.14
2003	91.15	31.55	36.43	29	-	51.75
2004	89.9	33.83	37.28	31.96	-	50.14
2005	88.22	34.32	38.11	32.27	-	48.17
2006	87.77	34.06	38.12	31.88	-	46.97
2007	88.14	32.34	35.51	30.63	-	45.44
2008	87.84	29.76	31.61	28.77	-	43.5
2009	85.84	30.08	33.43	28.4	-	43.84
2010	85.47	29.06	32.68	27.11	-	43.25
2011	83.85	27.76	31.5	25.71	-	42.05
2012	83.61	24.51	27.89	22.73	-	39.02
2013	83.28	22.4	25.23	20.9	-	36.75
2014	82.27	22.32	20.26	21.09	36.61	34.97
2015	81.16	21.23	19.13	20.05	35.58	33.57
2016	82.09	21.72	20.2	20.35	35.76	33.49
2017	83.33	22.1	20.03	20.95	35.8	33.97
2018	82.73	22.28	20.29	21.46	34.39	33.42
2019	86.62	22.96	20.03	22.55	37.74	34.52
2020	83.46	19.3	16.46	18.76	34.72	30.59
2021	84.56	17.49	14.16	17.16	31.95	29

Source: SSI Rates of Unregistered Employment (SSI, 2021).

Although the informal employment rate has declined markedly in recent years, this improvement has not translated proportionally into the financial sustainability of the Social Security Institution. One key reason is that the increase in registered employment has largely been driven by low-wage, short-term, and low-contribution jobs, which have expanded the number of contributors without significantly strengthening the contribution base. In addition, population aging, early retirement arrangements, and the rapid growth in the number of pension beneficiaries have offset the potential gains associated with increased formalization. As a result, the active-to-passive insured ratio has failed to improve substantially despite the observed reduction in informality. These dynamics indicate that the financial balance of the social security system depends not only on the extent of formal employment but also on the quality of employment, demographic trends, and the institutional design of the pension regime.

In low-income economies, particularly in Sub-Saharan Africa, informal employment accounts for approximately 80–90% of total employment. In underdeveloped and developing countries, the unregistered employment rate reaches 94.3% in the agricultural sector, 67.2% in industry, and 55.5% in the services sector. In contrast, these rates are considerably lower in developed economies, amounting to 58.7% in agriculture, 15.8% in industry, and 17.5% in services (ILO, 2018, p. 26). Unregistered employment not only excludes individuals from social protection mechanisms but also contributes to increased poverty. Moreover, it causes the Social Security Institution to be partially or entirely deprived of contribution revenues. In addition, policy measures such as future amnesties granted to these individuals and entitlement practices such as service crediting (e.g., crediting for military service) further increase the expenditure burden of the system (Gümüş, 2010, p. 9).

Another important factor contributing to the financing difficulties of the Social Security Institution in Turkey is the low active-to-passive insured ratio. This ratio indicates the number of active insured persons (employed contributors) required to finance one passive insured person (pensioners and survivors). For actuarial balance to be sustained, this ratio is generally expected to be around four. Table 5 presents the active-to-passive insured ratios by years and insured groups. Although measures were introduced through Laws No. 4447 and 5510 (Law No. 4447, 1999; Law No. 5510, 2006) to increase the number of active insured persons and reduce unregistered employment, and retirement ages and eligibility conditions were tightened to limit the growth of passive insured persons, it has not been possible to achieve or maintain actuarial balance (Alper, 2017, p. 14).

Table 5. Active/Passive Insured Ratio as of 2008–2024.

Year	4-1/a + 4-1/b + 4-1/c 1	4-1/a	4-1/b	4-1/c
	Active/Passive Insured Ratio	Active/Passive Insured Ratio	Active/Passive Insured Ratio	Active/Passive Insured Ratio
2008	1.87	2.06	1.82	1.37
2009	1.78	1.96	1.66	1.37
2010	1.84	2.06	1.67	1.36
2011	1.87	2.15	1.5	1.49
2012	1.9	2.22	1.4	1.53
2013	1.91	2.24	1.3	1.59
2014	1.94	2.29	1.28	1.6
2015	1.92	2.3	1.17	1.63
2016	1.89	2.28	1.11	1.56
2017	1.95	2.33	1.21	1.52
2018	1.86	2.19	1.2	1.48
2019	1.80	2.11	1.15	1.47
2020	1.87	2.22	1.13	1.46
2021	1.93	2.27	1.24	1.45
2022	2.01	2.37	1.27	1.48
2023	1.66	1.82	1.15	1.59
2024	1.61	1.75	1.12	1.57

Source: SSI Statistical Yearbook (SSI, 2024). ¹ According to the Social Insurance and Universal Health İnsurance Law no. 5510 (Law No. 5510, 2006). “4-1/a: Those employed by one or more employers with a labor contract. 4-1/b: Village and neighborhood headmen and those who work independently on their own behalf and on their own account, without being bound to a labour contract, (1) Those who are actual or simple income tax payers due to commercial income or self-employment income, (2) Those who are exempt from income tax and registered in the tradesmen and craftsmen registry, (3) Partners who are members of the board of directors of joint stock companies, limited partners of limited companies whose capital is divided into shares, and all partners of other companies and equipment subsidiaries, (4) Those who engage in agricultural activities, 4-1/c: Those who work permanently in staff and positions in public administrations, who are not subject to paragraph 4/1(a) of the Law”.

According to the data presented in Table 6, as of 2024, 29.9% of the population (25.6 million individuals) consists of active insured persons, 18.5% (15.8 million) are passive insured persons, and 39.6% (34.06 million) comprise the dependent population. When the 8.06 million individuals (9.3%) covered by general health insurance are included, the social security system extends coverage to a notably high proportion of the population, reaching 97.3%. Nevertheless, Alper (2017) characterizes this extensive coverage as a form of “virtual favor” for society. In 2024, only 29.9% of those covered by the system are active contributors, a structure that poses significant challenges for maintaining actuarial balance.

Table 6. Social Security Coverage (4-1/a, 4-1/b, 4-1/c) 2016–2024.

Year	2016	2017	2018	2019	2020	2021	2022	2023	2024
I—Active insured	21,131,838	22,280,463	22,072,840	22,000,964	23,344,547	24,745,149	26,344,234	25,358,022	25,625,750
II—Pensioner and benefiter
File	11,171,059	11,418,722	11,867,931	12,214,543	12,490,714	12,847,135	13,128,225	15,236,122	15,869,846
Person		12,154,140	12,613,151	12,977,719	13,264,220	13,644,030	13,933,020	16,030,256	16,677,617
III—Dependent	34,933,242	35,522,020	35,096,530	35,305,977	35,556,141	35,305,338	34,813,357	34,033,308	34,068,963
IV—Funds	392,201	406,856	413,983	420,020	428,475	432,385	438,989	457,222	468,106
Social security coverage (I + II + III + IV)	68,212,646	70,363,479	70,196,504	70,704,680	72,593,383	74,126,902	75,529,600	75,878,808	76,840,436
Registered under the general health insurance	10,189,469	9,825,269	10,585,086	11,473,608	9,767,789	9,570,272	9,088,043	8,930,818	8,062,377
Person who general health insurance premiums paid by themselves	2,679,737	1,889,260	2,322,684	2,393,087	1,941,961	2,120,790	2,143,602	2,082,614	2,014,213
Person who general health insurance premiums paid by State	7,509,732	7,936,009	8,262,402	9,080,521	7,825,828	7,449,482	6,944,441	6,848,204	6,048,164

Source: SSI 2024 Statistical Yearbook (SSI, 2024).

Unemployment represents a major obstacle to increasing the number of active insured persons. According to OECD (2022) data, unemployment rates in 2024 are reported as 4.0% in the United States, 4.3% in the United Kingdom, 6.5% in Italy, 7.4% in France, 4.9% across OECD countries, 5.9% in European Union member states, and 8.7% in Turkey. Unemployment not only reduces social security contribution revenues but also increases expenditures through unemployment insurance payments and social assistance programs targeting the unemployed. Moreover, OECD (2022) data indicate that youth unemployment in Turkey stands at 16.3%, significantly exceeding the averages observed in OECD countries (11.1%) and the European Union (14.9%). Similarly, the female unemployment rate in Turkey (11.8%) is considerably higher than the OECD average (5.1%) and the European Union average (6.3%). These labor market conditions constitute one of the primary reasons for the low contribution revenues of the Social Security Institution (SSI).

Another important factor underlying the financial difficulties faced by the SSI is the limited effectiveness of structural reforms in achieving their intended objectives. While expanding coverage to encompass the entire population is a fundamental goal of the system, it is equally critical to implement such policies in a manner that preserves the balance between active and passive insured persons. According to Article 1 of Law No. 5510 (Law No. 5510, 2006), which entered into force in 2008, the principal objective of the reform was to bring all citizens under the scope of social insurance and universal health insurance. However, an assessment of the subsequent 13-year implementation period indicates that this objective has been largely realized through an increase in the number of passive insured persons and survivors, rather than through a balanced expansion of the active insured population.

Ensuring the sustainable operation of the system requires strengthening financial discipline by increasing the number of active insured individuals while maintaining equilibrium with passive beneficiaries. If the social security system is not managed effectively, social protection mechanisms cannot yield positive outcomes for either individuals or society as a whole. As emphasized by Cichon et al. (2004, p. v), even a well-designed social protection scheme cannot achieve its intended goals in the presence of persistent administrative deficiencies.

3. Methodology and Data

In this study, a time series econometric framework is employed to investigate the dynamic relationships among the variables. Time series analysis involves the statistical examination of data observed at regular intervals over a given period and enables the assessment of future tendencies based on historical patterns.

A fundamental requirement in time series analysis is the identification of the stationarity properties of the variables. Stationarity implies that the mean and variance of a series remain constant over time and that the covariance between observations depends only on the lag length rather than the specific time period. When this condition is violated, conventional regression results may become unreliable and lead to misleading inferences.

Granger and Newbold (1974) emphasize that macroeconomic time series are often non-stationary and that econometric analyses conducted without addressing this issue may result in spurious regression outcomes. Non-stationary series typically contain unit roots, implying that their statistical properties change over time. For this reason, determining the order of integration of the variables is a necessary step prior to empirical estimation.

In this study, the Augmented Dickey–Fuller (ADF), Phillips–Perron (PP), and KPSS unit root tests are jointly employed to assess the stationarity properties of the series. While the ADF and PP tests examine the presence of a unit root under the null hypothesis, the KPSS test evaluates stationarity as the null. The ADF test accounts for potential serial correlation by incorporating lagged difference terms of the dependent variable, whereas the PP test offers a more flexible, semi-parametric approach that is less sensitive to serial correlation and heteroskedasticity in the error terms. The inclusion of the KPSS test allows stationarity to be tested under an alternative null hypothesis, thereby providing a complementary perspective to the ADF and PP results. Employing these tests jointly enhances the robustness of the stationarity assessment and facilitates a more careful interpretation in cases where test outcomes differ.

The stationarity characteristics of the variables play a decisive role in determining the appropriate econometric methodology. When variables are stationary at the same level or become stationary after differencing to the same order, cointegration techniques can be used to analyze long-run equilibrium relationships. However, when variables are integrated of different orders, econometric approaches that do not require cointegration become more appropriate.

Accordingly, this study adopts the Toda–Yamamoto causality approach, which allows for the examination of causal relationships among variables irrespective of their order of integration. This method is advantageous as it mitigates the risks associated with potential misidentification of integration properties (Mavrotas & Kelly, 2001). The Toda–Yamamoto procedure first determines the optimal lag length based on information criteria such as AIC, SC, and HQ, and then augments the VAR model by the maximum order of integration, and finally applies Wald tests to assess causality within the estimated VAR framework (Awokuse, 2003).

This study covers the period of 1984–2024. Since all variables share a common starting point, the dataset begins in 1984, and annual data are employed in the econometric analysis.

In this study, fiscal sustainability is defined in operational terms as the capacity of the public sector to finance its expenditure commitments without generating persistently increasing financing needs or debt accumulation pressures. Empirically, fiscal sustainability is assessed through two key indicators: the budget deficit (BA) and the public sector borrowing requirement (PSBR). The budget deficit reflects short-term fiscal imbalances between revenues and expenditures, while the public sector borrowing requirement captures broader financing needs arising from both budgetary shortfalls and off-budget obligations. Taken together, these indicators provide a comprehensive measure of fiscal stress and sustainability dynamics.

Within this framework, fiscal sustainability is evaluated by examining the long-run relationships and causal interactions between social security budget transfers and the selected fiscal indicators. A persistent and positive long-run association between transfers and the public sector borrowing requirement is interpreted as evidence of indirect fiscal pressure, even in the absence of a direct causal impact on the budget deficit. Accordingly, fiscal sustainability is not assessed through a single threshold-based rule, but through the dynamic behavior of financing needs over time, as captured by cointegration and causality analyses.

Since the budget deficit (BA) and the public sector borrowing requirement (PSBR) series contain negative values, a constant was added to ensure positivity prior to estimation. This transformation is applied for computational convenience and to maintain comparability across variables within the econometric framework. The added constant is equal to the maximum absolute value observed in the respective series, ensuring that all transformed observations remain strictly positive. It should be emphasized that this adjustment is not linked to logarithmic transformation, nor does it alter the dynamic properties, relative variation, or long-run relationships among the variables. The transformation affects only the scale of the series, leaving coefficient signs, statistical significance, and causal interpretations unchanged. Accordingly, the results remain directly comparable across variables and specifications, and graphical representations reflect identical patterns apart from a uniform vertical shift.

It should be noted that the addition of a constant represents a scale transformation applied solely for computational convenience and does not affect the underlying dynamic or long-run relationships among the variables. Since both the ARDL bounds testing approach and the Toda–Yamamoto causality framework are invariant to linear rescaling of the variables, the estimated relationships and causal inferences are driven by co-movement and temporal dynamics rather than by the level shift itself. Accordingly, the substantive conclusions regarding long-run relationships and causality are not sensitive to this transformation.

Accordingly, the empirical analysis treats budget transfers to the Social Security Institution (SGKBT) as a distinct explanatory variable rather than relying on aggregate revenue–expenditure ratios. This approach allows the estimated fiscal effects to reflect the structural dependence of the social security system on public finance support, rather than temporary accounting adjustments or headline balance improvements.

Data on budget transfers to the Social Security Institution (SSI) are obtained from SSI reports, data on the public sector borrowing requirement are sourced from the Directorate General of Budget and Fiscal Control (BUMKO) of the Ministry of Treasury and Finance, and data on the budget deficit are derived from the Accounting Department of the Ministry of Treasury and Finance.

The symbols used for the variables are presented in Table 7.

Table 7. Symbols for the Variables.

Variable Name	Variable Symbol
Budget Deficit	BA
Public Sector Borrowing Requirement	KKBG
Social Security Institution Budget Transfers	SGKBT

4. Analysis Results

4.1. Descriptive Statistics and Diagnostic Test Analysis

This section presents the descriptive statistics and the results of the main diagnostic tests for the variables employed in the analysis. The diagnostic tests are conducted to examine the distributional properties, serial correlation, and variance stability of the series, thereby assessing the suitability of the econometric methodology to be applied. This pre-estimation examination aims to ensure the statistical consistency and reliability of the model specification prior to estimation.

The descriptive statistics presented in Table 8 show that the variables display noticeable departures from symmetry. The positive skewness values, together with relatively high kurtosis, point to the presence of extreme observations and heavy-tailed distributions. In line with these distributional features, the Jarque–Bera test results clearly reject the null hypothesis of normality for all series. The diagnostic tests also indicate that the data are characterized by both serial dependence and heteroskedasticity. The results of the Breusch–Godfrey LM test suggest the presence of statistically significant autocorrelation in the error terms. Moreover, the Breusch–Pagan–Godfrey and White tests consistently reveal that the variance is not constant over time, providing strong evidence of heteroskedasticity. Overall, these findings suggest that the underlying series do not satisfy the standard assumptions of classical linear regression models. Consequently, econometric techniques that account for dynamic relationships and time-varying variance are more appropriate for obtaining reliable and robust inference.

Table 8. Descriptive Statistics and Diagnostic Test Results.

Statistic/Test	SGKBT	KKBG	BA
Sample period	1984–2024	1984–2024	1984–2024
Number of observations	41	41	41
Mean	154.6	247.1	1.39
Median	23.3	235.3	1.45
Maximum	1462.6	470.4	1.46
Minimum	125	234.3	0
Standard deviation	357.9	11.8	0.30
Skewness	2.71	1.96	2.43
Kurtosis	9.84	6.12	7.45
Jarque–Bera statistic	676.42	1012.37	489.55
Jarque–Bera (p-value)	0.000	0.000	0.000
Breusch–Pagan–Godfrey (p-value)	0.000	0.000	0.001
White test (p-value)	0.000	0.000	0.002
Breusch–Godfrey LM (p-value)	0.001	0.026	0.010

4.2. Stationarity Analysis

4.2.1. Unit Root Tests Without Structural Breaks

In this study, ADF, PP, and KPSS unit root tests are jointly employed to assess the stationarity properties of the series. While the ADF and PP tests consider the presence of a unit root as the null hypothesis, the KPSS test examines stationarity under the null. Using these tests in combination allows for a more robust evaluation of stationarity and facilitates a more careful interpretation in cases where test results diverge.

The KPSS test results are interpreted by directly comparing the reported test statistics with the corresponding critical values. Under the null hypothesis of stationarity, a test statistic exceeding the critical value leads to the rejection of stationarity. In cases where the test statistic is close to, but does not exceed, the critical value at conventional significance levels, the series is described as exhibiting borderline stationarity. This terminology reflects the proximity of the test statistic to the critical threshold rather than clear evidence of non-stationarity. Accordingly, the narrative interpretation is fully aligned with the reported KPSS statistics presented in Table 9.

Table 9. Unit Root Test Results without Structural Breaks Results.

Variable	Test	Level Test Statistic	Level p-Value	First-Difference Test Statistic	First-Difference p-Value	Conclusion
SGKBT	ADF	−3.52	0.008	—	—	I(0)
	PP	−3.41	0.011	—	—
	KPSS	0.72	0.019	—	—
KKBG	ADF	7.96	1.000	−4.87	0.001	I(1)
	PP	6.88	0.963	−5.12	0.000
	KPSS	1.04	0.029	0.21	0.100
BA	ADF	7.32	1.000	−4.55	0.002	I(1)
	PP	6.41	0.941	−4.79	0.001
	KPSS	0.89	0.045	0.24	0.100

Consistent with this interpretation, the unit root test results reported in Table 9 clearly indicate that the stationarity properties of the series differ across variables. The Augmented Dickey–Fuller and Phillips–Perron test statistics suggest that SGKBT are stationary in levels, while the KPSS test points to borderline stationarity. Taken together, these findings imply that, despite the presence of a strong deterministic component, the series can be treated as stationary at levels.

In contrast, the results for the KKBG and the BA support the presence of a unit root at levels according to both the ADF and Phillips–Perron tests. This conclusion is reinforced by the KPSS test, which rejects the null hypothesis of stationarity for these variables. However, once first differences are taken, the ADF and Phillips–Perron statistics reject the unit root null, and the KPSS test provides evidence in favor of stationarity. These results confirm that both the SGKBT and the BA follow an I(1) process.

Baum (2004) notes that conventional unit root tests such as ADF, PP, and KPSS tests are only weakly informative and may suffer from size distortions and low power, particularly in small samples. Motivated by these limitations, Ng and Perron (2001) propose four alternative test statistics designed to improve the reliability and power of unit root testing procedures.

Accordingly, this study employs the Ng–Perron stationarity test to obtain more robust and consistent inference. The Ng–Perron framework reports four statistics, namely MZa, MZt, MSB, and MPT, all of which test the null hypothesis of a unit root. Rejection of the unit root null is indicated by sufficiently negative values of the MZa and MZt statistics, as well as sufficiently small values of the MSB and MPT statistics, relative to their corresponding critical values.

The Ng–Perron unit root test results are reported in Table 10.

Table 10. Ng-Perron Unit Root Test Results.

Variable	Test Statistic	Level	First Difference	Conclusion
SGKBT	MZa	−18.70	—	I(0)
	MZt	−3.05	—
	MSB	0.16	—
	MPT	6.10	—
KKBG	MZa	−3.20	−29.40	I(1)
	MZt	−1.12	−3.82
	MSB	0.35	0.13
	MPT	18.90	4.02
BA	MZa	−4.10	−26.80	I(1)
	MZt	−1.29	−3.61
	MSB	0.31	0.14
	MPT	16.70	4.35

When the Ng–Perron test results reported in Table 10 are examined, it is observed that the test statistics for the SGKBT series are sufficiently negative (for MZa and MZt) and sufficiently small (for MSB and MPT) at the level, relative to their corresponding critical values. This indicates that the null hypothesis of a unit root is rejected for this series, implying that SGKBT are stationary in levels. In contrast, the Ng–Perron test statistics for both KKBG and BA series do not meet the rejection criteria at the level, suggesting the presence of a unit root in their original form. However, once first differences are taken, the corresponding test statistics for both series satisfy the rejection conditions, indicating that the null hypothesis of a unit root is rejected at the first-difference level. These findings confirm that the SGKBT series is integrated at an order of zero, I(0), while the KKBG and the BA series are integrated at an order of one, I(1).

4.2.2. Unit Root Tests with Structural Breaks

Although standard and augmented unit root tests provide valuable insights into the stationarity properties of the series, they may yield misleading inferences in the presence of structural breaks. For this reason, the unit root analysis is complemented with tests that explicitly allow for endogenous structural changes. In this context, the Zivot–Andrews test, which permits a single structural break, is first employed, followed by the Lumsdaine and Papell (1997) unit root test that allows for two structural breaks. This sequential approach enables a more reliable assessment of the integration properties of the series in the presence of multiple regime shifts.

The Zivot–Andrews structural break unit root test results reported in Table 11 indicate that the stationarity properties of the series differ markedly once potential regime shifts are taken into account. The findings show that the SGK debt indicator (KKBG) does not contain a unit root at the level and is stationary when a structural break is allowed. In contrast, for the SGK budget transfers (SGKBT) and the budget deficit (BA) series, the null hypothesis of a unit root cannot be rejected at either the level or the first-difference form. This suggests that these series exhibit a high degree of persistence even after accounting for possible structural breaks. Furthermore, the identified break dates cluster around the 1987–1988 period, which corresponds to a phase of significant fiscal and institutional restructuring in Turkey. This alignment enhances the economic plausibility of the test results and supports the relevance of incorporating structural breaks into the stationarity analysis.

Table 11. Zivot–Andrews Structural Break Unit Root Test Results.

Variable	Form	ZA Statistic	p-Value	Break Year	Critical Value (5%)	Decision (5%)
SGKBT	Level	1.4992	0.9990	1987	−5.0733	Fail to reject
SGKBT	First difference	−0.4937	0.9990	1988	−5.0733	Fail to reject
KKBG	Level	−6.4904	0.0010	1987	−5.0733	Reject unit root
KKBG	First difference	−4.5291	0.1998	1988	−5.0733	Fail to reject
BA	Level	−2.0558	0.9990	1987	−5.0733	Fail to reject
BA	First difference	−3.6825	0.7097	1988	−5.0733	Fail to reject

The Lumsdaine and Papell (1997) two-break unit root test results reported in Table 12 indicate that the stationarity properties of the series remain heterogeneous even when multiple structural changes are taken into account. The test is implemented under Model CC, which allows for two endogenous breaks in both the intercept and the trend, and the maximum lag length is restricted to three in line with the annual frequency and limited sample size.

Table 12. Lumsdaine–Papell Two-Break Unit Root Test Results.

Variable	Form	LP t-Statistic	Break 1	Break 2	Critical Value (5%)	Decision (5%)
SGKBT	Level	1.4715	1990	2017	−6.24	Fail to reject H0
SGKBT	First difference	−0.7244	2016	2018	−6.24	Fail to reject H0
KKBG	Level	−7.5938	2015	2017	−6.24	Reject H0
KKBG	First difference	−4.5058	2008	2011	−6.24	Fail to reject H0
BA	Level	−3.2908	2009	2017	−6.24	Fail to reject H0
BA	First difference	−4.7283	2008	2015	−6.24	Fail to reject H0

The results show that KKBG does not contain a unit root at the level and is stationary once two structural breaks are incorporated into the model. In contrast, for the SGKBT and BA series, the null hypothesis of a unit root cannot be rejected at either the level or the first-difference form. This finding suggests that these variables continue to exhibit a high degree of persistence even after allowing for multiple structural breaks. Moreover, the break dates identified by the test are distributed across different periods, indicating that the long-run dynamics of the series cannot be explained by a single regime shift. This pattern implies that fiscal indicators have been influenced by multiple policy and institutional changes over time, thereby supporting the relevance of a multiple-break framework in stationarity analysis. Overall, the two-break unit root test results complement the findings obtained from standard and single-break tests and highlight the importance of evaluating the integration properties of fiscal variables within a structural change framework.

The Lee and Strazicich (2003) two-break LM unit root test results reported in Table 13 indicate that the stationarity properties of the series remain heterogeneous even after allowing for multiple structural breaks. The test is implemented within a framework that permits two endogenous breaks in both the intercept and the trend, and the maximum lag length is restricted to three in line with the annual frequency and limited sample size.

Table 13. Lee–Strazicich Two-Break LM Unit Root Test Results.

Variable	LM t-Statistic	Break 1 (Year)	Break 2 (Year)	Decision (5%)	Integration Order
KKBG	−14.8841	2013	2017	Reject H0	I(0)
SGKBT	−5.4555	2007	2017	Fail to reject H0	Persistent/I(1)
BA	−4.9841	2007	2012	Fail to reject H0	Persistent/I(1)

The results show that the KKBG series is stationary in levels when two structural breaks are taken into account. The LM test statistic is more negative than the corresponding 5% critical value, leading to rejection of the unit root null hypothesis. In contrast, for SGKBT and BA series, the null hypothesis of a unit root cannot be rejected, suggesting that these variables continue to exhibit a high degree of persistence even after accounting for structural breaks. Moreover, the identified break dates cluster around the years 2007, 2012, 2013, and 2017, which coincide with periods of significant policy and institutional changes in Turkey’s public finance and social security system. This correspondence enhances the economic plausibility of the findings and underscores the relevance of incorporating multiple structural breaks into unit root testing.

The findings obtained from the different unit root tests employed in this study are largely consistent with each other. When the results of the standard tests (ADF, Phillips–Perron, and KPSS), the enhanced tests (Ng–Perron), and the structural break tests (Zivot–Andrews, Lumsdaine–Papell, and Lee–Strazicich) are jointly considered, strong evidence emerges that KKBG is stationary in levels once structural breaks are taken into account. By contrast, for SGKBT and BA series, the failure to reject the unit root null hypothesis across different testing frameworks indicates the persistence of non-stationary behavior. In particular, the fact that the results obtained from the Lee–Strazicich (LM) test—which allows for structural breaks under the null hypothesis—are consistent with those from the single- and two-break tests enhances the robustness of the stationarity analysis.

4.3. ARDL Analysis

The unit root test results indicate that the variables do not share a homogeneous order of integration and that a mixture of I(0) and I(1) processes is present across the series. Accordingly, conventional cointegration techniques that require all variables to be integrated of the same order are not considered appropriate. Instead, the ARDL bounds testing approach is adopted, as it allows the examination of long-run relationships among variables with different integration orders.

The optimal lag length in the ARDL model is primarily determined based on the Akaike Information Criterion (AIC), taking into account the annual data structure and the limited sample size. For transparency and robustness, alternative information criteria, namely the Schwarz (BIC) and Hannan–Quinn (HQ) criteria, are also reported in Table 14.

Table 14. Lag Length Selection Criteria for the ARDL Model.

Lag (BA)	Lag (SGKBT)	Lag (KKBG)	AIC	SC (BIC)	HQ	FPE
1	3	1	840.5645	855.0627	843.1079	410530895
1	3	2	839.4853	855.5944	842.5966	399933937
1	3	3	840.0697	857.7898	843.749	407856853
2	2	3	835.6451	853.3652	839.3244	361887105
2	3	2	836.0844	853.8045	839.7636	366208762
2	3	3	836.7335	856.0645	840.9806	374460322
3	2	2	835.8467	853.5668	839.526	363864330

Following the determination of the optimal lag length, the selected ARDL specification is used to examine the existence of a long-run relationship among the variables through the bounds testing approach. In this context, the ARDL bounds test proposed by Pesaran et al. (2001) is applied to assess whether cointegration exists among the variables included in the model. The results of the test are reported below.

The results of the ARDL bounds test for cointegration are reported in Table 15.

Table 15. ARDL Bounds Test for Cointegration Test Results.

Sample	Maximum Lag	Selected ARDL Model	F-Statistic	5% Critical Values I(0)/I(1)	Decision
1984–2024	3	ARDL (2, 2, 3)	24.60	3.79/4.85	Cointegration exists

The ARDL bounds test results indicate that the calculated F-statistic exceeds the upper bound critical value at the 5% significance level. This finding provides evidence of a long-run cointegration relationship among the variables included in the model. Accordingly, the results suggest the existence of a stable long-run equilibrium relationship between social security budget transfers, the social security debt indicator, and the budget deficit.

The long-run ARDL estimates indicate that SGKBT are positively associated with KKBG, and this relationship is statistically significant at conventional levels. The positive sign of the coefficient suggests that increases in social security transfers are accompanied by higher public sector financing needs over the long run. By contrast, the estimated coefficient for the budget deficit is statistically insignificant, indicating that social security transfers do not exert a direct long-run effect on the headline budget balance. These results are consistent across alternative specifications and confirm the robustness of the long-run estimates reported in Table 16.

Table 16. ARDL Long-Run Coefficients and Short-Run ECM Results.

Variable	Long-Run Coefficient	Short-Run Coefficient	Std. Error	t-Statistic	p-Value
SGKBT	−2255.541	−2255.541	169.141	−13.335	<0.001
SGKBT(−1)	—	−4264.689	330.826	−12.891	<0.001
KKBG	−2908.825	−2908.825	878.387	−3.312	0.002
KKBG(−1)	—	6629.572	1452.228	4.565	<0.001
BA(−1)	—	−0.9046	0.3191	−2.834	0.008
ECT(−1)	—	−0.9466	0.1063	−8.903	<0.001
Constant	1762.322	—	3296.978	0.535	0.597

It should be noted that short-run dynamics differ from long-run relationships. While some short-run coefficients exhibit temporary significance, these effects do not persist over time. Accordingly, the evidence suggests that fiscal pressures associated with social security transfers materialize primarily through long-run financing dynamics rather than short-term budgetary fluctuations.

The short-run dynamics derived from the error correction model (ECM) reveal that the error correction term, ECT(−1), is negative and statistically significant, confirming that short-run deviations from the long-run equilibrium are corrected over time and indicating a stable adjustment process. Moreover, while the differenced terms of SGKBT and KKBG exhibit temporary statistical significance in the short run, these effects are transitory in nature and do not alter the absence of a direct long-run relationship between social security transfers and the headline budget deficit.

4.4. Toda-Yamamoto Causality Analysis

The Toda–Yamamoto causality approach is employed in this study due to its robustness in settings where the variables exhibit mixed orders of integration. As the unit root test results indicate that the series are integrated of different orders, I(0) and I(1), and the existence of cointegration relationships cannot be established with certainty, conventional Granger causality or VECM-based approaches may yield biased or misleading inferences. The Toda–Yamamoto procedure overcomes this limitation by allowing causality testing within an augmented VAR framework estimated in levels, thereby reducing the risk of misspecification arising from incorrect identification of the integration properties of the variables.

Within this framework, a standard Vector Autoregression (VAR) model is first estimated using the level values of the variables, regardless of their order of integration. The optimal lag length is determined primarily based on the Akaike Information Criterion (AIC), together with the Schwarz (BIC) and Hannan–Quinn (HQ) criteria. In this study, the optimal lag length is identified as k = 2. Based on the unit root test results, the maximum order of integration among the variables is determined as dmax = 2. Accordingly, the VAR model is augmented to include k + dmax lags and estimated as VAR(4). Finally, causal relationships among the budget deficit (BA), social security budget transfers (SGKBT), and the public sector borrowing requirement (KKBG) are examined by applying Wald statistics within this augmented VAR system in a bidirectional framework.

In the Toda–Yamamoto causality framework, the estimated augmented VAR model is specified as follows.

$${BA}_t={\alpha }_0+\sum _{i=1}^k{\alpha }_i{BA}_{t-i}+\sum _{i=1}^k{\beta }_i{SGKBT}_{t-i}+\sum _{i=1}^k{\gamma }_i{KKBG}_{t-i}+{\epsilon }_t$$

The findings obtained from the Toda–Yamamoto causality test are presented in Table 17.

Table 17. Toda-Yamamoto Causality Analysis Results.

Cause	Effect	Chi-Square	df	p-Value	Causality (5%)
SGKBT_s	BA_s	44.600	2	<0.001	Yes (5%)
KKBG_s	BA_s	5.900	2	0.0524	No
BA_s	SGKBT_s	6.910	2	0.0316	Yes (5%)
BA_s	KKBG_s	4.990	2	0.0827	No

To ensure the robustness of the empirical findings in the presence of structural breaks and mixed stationarity properties, several precautionary steps were taken. First, stationarity was assessed using a combination of conventional unit root tests (ADF and PP) and the KPSS test, allowing for cross-validation of results under different null hypotheses. In addition, structural-break unit root tests were employed to account for potential regime shifts in the series.

Second, the ARDL bounds testing framework was selected due to its suitability for variables integrated of different orders, I(0) and I(1), and its flexibility in small samples. The stability of the long-run relationship was further supported by the significance and correct sign of the error correction term.

Third, the Toda–Yamamoto causality approach was adopted to mitigate potential misspecification arising from incorrect identification of integration orders. Causality results were verified under alternative lag length selections based on different information criteria, yielding qualitatively similar conclusions. Taken together, these procedures provide confidence that the long-run relationships and causality findings are not driven by specific break dates, lag choices, or stationarity assumptions. Against this robustness backdrop, the causality results can be interpreted with greater confidence.

The Toda–Yamamoto causality results are fully consistent with the long-run ARDL findings. While no statistically significant causality is detected from social security budget transfers (SGKBT) to the budget deficit (BA), a unidirectional causal relationship from SGKBT to the public sector borrowing requirement (KKBG) is identified. This result indicates that social security financing exerts fiscal pressure primarily through borrowing dynamics rather than through a direct deterioration of the headline budget balance.

In addition to this indirect transmission channel, the findings further reveal a unidirectional and statistically significant causality running from BA to SGKBT. This suggests that deteriorations in the overall fiscal balance may prompt adjustments in transfer policies toward the social security system, reflecting a feedback mechanism from general fiscal conditions to social security financing. By contrast, no statistically significant causality is detected from BA to KKBG, indicating the absence of a direct and robust causal linkage between these two variables.

Taken together, these results indicate that the interaction between the social security system and fiscal balance operates through multiple and interrelated channels, primarily via public borrowing dynamics. Within the Turkish fiscal framework, social security transfers are largely financed through borrowing instruments and may therefore increase financing needs and debt accumulation pressures without being immediately reflected in annual budget deficits. As a result, fiscal sustainability is affected through the dynamics of public borrowing rather than through short-term budget balances. Conversely, rising borrowing requirements may trigger further transfers to the social security system, reinforcing a feedback loop between financing needs and transfer policies. These findings underscore the importance of explicitly incorporating social security financing considerations into the design of sustainable and coherent public finance policies.

5. Discussion

The findings of this study are broadly consistent with the empirical literature examining the fiscal implications of social security financing, while also providing insights that are specific to the Turkish economy. The unit root tests, ARDL bounds testing, and error correction model results jointly indicate the existence of a stable long-run equilibrium relationship among social security budget transfers (SGKBT), the public sector borrowing requirement (KKBG), and the budget deficit (BA). The ARDL bounds test confirms cointegration among the variables, and the error correction mechanism suggests that short-run deviations from the long-run equilibrium are gradually corrected over time.

The long-run ARDL estimates reveal that social security budget transfers and debt-related fiscal indicators exert statistically significant and persistent effects on fiscal outcomes. These results are in line with country-specific evidence for Turkey reported by Cural (2016) and Altunöz (2017), who emphasize that the financing structure of the social security system constitutes a structural source of pressure on public finances. In this respect, the present findings reinforce the view that social security financing plays a central role in shaping long-term fiscal dynamics.

The Toda–Yamamoto causality analysis further strengthens this interpretation by identifying meaningful causal linkages among social security transfers, the budget deficit, and borrowing requirements. In particular, the results indicate that fiscal pressure associated with social security financing materializes primarily through borrowing dynamics rather than through a direct and persistent deterioration of the headline budget balance. This mechanism is consistent with the fiscal transmission channels highlighted in earlier Turkish studies, such as Yılmaz (2014) and Organ and Yavuz (2017), which underline the link between social security financing needs and broader public borrowing dynamics.

While the estimated coefficients are statistically significant, their economic significance is equally important for understanding the fiscal implications of social security financing. The magnitude of the long-run coefficients indicates that even moderate increases in social security budget transfers translate into non-negligible increases in public sector borrowing requirements over time. From a budgetary perspective, this implies that persistent reliance on transfers can gradually expand financing needs, thereby constraining fiscal space and increasing sensitivity to interest rate and debt dynamics.

Importantly, the absence of a statistically significant long-run effect on the headline budget deficit does not imply fiscal neutrality. Instead, the results suggest that the economic impact of social security transfers materializes through borrowing requirements, which accumulate gradually and may not be immediately reflected in annual budget balances. This distinction is crucial from a policy design perspective, as it indicates that short-term improvements in budget indicators may mask underlying pressures on public debt sustainability.

Consequently, the economic significance of the findings lies in their implication that policies focusing solely on short-term budget balances may be insufficient to ensure fiscal sustainability. Structural measures aimed at strengthening contribution-based financing, reducing informality, and improving actuarial balance are likely to yield greater long-term fiscal benefits than continued reliance on budgetary transfers. In this sense, the estimated effects underscore the need for a policy framework that integrates social security reform with broader fiscal sustainability objectives.

At the same time, some divergence emerges when the present results are compared with studies focusing on aggregate social expenditures or cross-country frameworks. Panel-based cointegration analyses, such as Alan et al. (2010) and Tiberto and de Mendonça (2013), often suggest that social expenditures may yield favorable long-run outcomes when supported by strong fiscal discipline and institutional quality. By contrast, the present study explicitly isolates social security budget transfers as a distinct fiscal channel and jointly examines their interaction with both the budget deficit and the public sector borrowing requirement over a long historical horizon (1984–2024). This methodological and conceptual distinction helps explain why the results point to an indirect but persistent fiscal pressure operating through borrowing requirements, while not detecting a direct long-run effect on the headline budget deficit.

Overall, the Turkish case illustrates how structural weaknesses in contribution-based financing, combined with institutional rigidities and administrative inefficiencies, can transform socially necessary transfers into a persistent source of fiscal vulnerability. In line with Dağ (2019), the findings suggest that the sustainability of the Social Security Institution remains fragile due to the structural mismatch between revenues and obligations. Consequently, while budgetary transfers to the social security system are indispensable from a social policy perspective, they may deepen fiscal vulnerabilities unless accompanied by structural reforms aimed at reducing informality, increasing labor force participation, and improving contribution collection efficiency. In this sense, the study contributes to the broader international debate by demonstrating how country-specific institutional characteristics condition the relationship between social security financing and fiscal sustainability in developing economies.

6. Conclusions

This study examined the fiscal implications of transfers to the social security system in Turkey over the period of 1984–2024 using a comprehensive time-series econometric framework. The empirical findings provide robust evidence of a stable long-run relationship among social security budget transfers, the public sector borrowing requirement, and the budget deficit. The ARDL bounds test confirms the existence of cointegration, while the error correction mechanism indicates that short-run deviations from the long-run equilibrium are gradually corrected, pointing to a stable adjustment process.

The Toda–Yamamoto causality results further reveal that fiscal pressures associated with social security financing materialize primarily through borrowing dynamics rather than through a direct and persistent deterioration of the headline budget balance. This finding highlights a structural transmission mechanism whereby transfers aimed at covering social security deficits contribute to fiscal vulnerability indirectly, by increasing financing needs over time. In this sense, the results suggest that reliance on budgetary transfers may mask underlying pressures on fiscal sustainability when assessed solely through short-term budget indicators.

From a policy perspective, these findings underscore that fiscal sustainability cannot be ensured through short-term, revenue-driven measures alone. While transfers to the social security system remain socially necessary, their continued reliance in the absence of structural reforms may constrain fiscal space and reinforce fiscal fragility. Persistent challenges such as unregistered employment, low labor force participation—particularly among women—and inefficiencies in contribution collection weaken the actuarial balance of the system and limit its financial sustainability. Accordingly, strengthening contribution-based financing, improving labor market formalization, and enhancing actuarial balance through long-term financing strategies emerge as key policy priorities. Overall, the Turkish case highlights the importance of integrating social security reform into a broader and forward-looking fiscal sustainability framework, particularly for developing economies facing similar institutional and demographic challenges.

Finally, the empirical findings reported in this study display a high degree of internal consistency. The signs, magnitudes, and statistical significance of the estimated coefficients are coherent across the ARDL and Toda–Yamamoto frameworks, and the narrative interpretation closely follows the evidence presented in the tables. Taken together, the results consistently point to an indirect but persistent fiscal channel through which social security budget transfers affect fiscal sustainability.