Assessing the Link Between the Misery Index and Dollarization: Regional Evidence from Türkiye

Abstract

This study analyzes the relationship between macroeconomic distress and financial dollarization in Türkiye using annual regional panel data for 26 Nomenclature of Territorial Units for Statistics 2 regions over the period 2005–2021. Macroeconomic distress is captured using the misery index, computed as the compound of inflation and unemployment rates, while the share of foreign-currency-denominated deposits in total deposits measures financial dollarization. Applying second-generation panel econometric models that account for regional heterogeneity, we investigate both long-run equilibrium relationships and short-run interactions. Panel cointegration tests show a long-run connection between macroeconomic distress and dollarization. Short-run effects estimated using a Panel Vector Error Correction Model and a Cross-Sectionally Augmented ARDL framework point to bidirectional causality. Long-run coefficient estimates obtained via Dynamic Ordinary Least Squares indicate an apparent asymmetry. Increases in dollarization exert a substantial and economically significant effect on macroeconomic distress, whereas the long-run impact of distress on dollarization is comparatively modest. The findings suggest that dollarization functions not only as a response to macroeconomic instability but also as a structural element that intensifies inflationary pressures and labor market distortions over time. Focusing on regional patterns rather than national aggregates, the paper provides new evidence on the spatial dimension of the dollarization–instability link.

1. Introduction

Macroeconomic stability is essential for reducing uncertainty in economic decision-making, allowing economic agents to form coherent long-term expectations. Prior research consistently demonstrates that price stability and low unemployment are fundamental pillars of economic welfare. Deteriorations in these indicators negatively impact both real economic activity and social well-being (Fischer, 1993; Barro, 1995; Lucas, 2000; Anghel et al., 2017; Kukaj, 2018; Kalinová & Kroutlová, 2023). In this context, the misery index (MI), initially proposed by Okun (1970) as the sum of unemployment and inflation rates, has become a widely used composite indicator that reflects the combined societal burden of macroeconomic distress. In recent years, renewed episodes of high inflation and persistent unemployment have contributed to a resurgence of interest in the MI (Clemens et al., 2022; Sánchez López, 2024).

Beyond being a summary measure of welfare loss, the MI has been used as a diagnostic tool to evaluate broader social and economic conditions and identify areas requiring policy intervention (Cohen et al., 2014; Büyüksarıkulak & Suluk, 2022; Akinlo, 2024; Gakuru & Yang, 2025; Osuma & Nzimande, 2025). The index also reflects the extent of macroeconomic uncertainty households and firms experience. Higher inflation and unemployment heighten perceptions of economic risk, diminish confidence in future income and price stability, and lead to more risk-averse actions. As a result, economic agents often adjust their financial portfolios in response to deteriorating macroeconomic conditions. One prominent behavioral response in such environments is the substitution of assets denominated in domestic currency for foreign currency instruments to protect purchasing power, thereby fostering financial dollarization (Mishkin, 1996; Reinhart et al., 2003; Mecagni et al., 2015). From this perspective, increases in the MI can be interpreted as a main driver of declining trust in the domestic currency and, consequently, rising dollarization.

Another aspect of this relationship is its potentially bidirectional nature. While higher inflation and unemployment shocks erode confidence in the domestic currency and promote dollarization (Ize & Levy-Yeyati, 2003; Mecagni et al., 2015), rising dollarization levels can further worsen macroeconomic instability. By weakening the effectiveness of monetary policy transmission, increasing exchange rate pass-through, and generating balance-sheet vulnerabilities, dollarization can intensify inflationary dynamics and adversely affect labor market outcomes (Reinhart et al., 2003; Yiğiteli, 2022). In this respect, dollarization does not respond only to macroeconomic distress but may also reinforce conditions such as high inflation and unemployment that constitute the MI. This process indicates that the link between MI and dollarization extends beyond a unidirectional causal relationship and instead shows a two-way interaction at the core of macroeconomic instability and expectation formation.

Although there is substantial research on dollarization and its macroeconomic consequences, empirical studies that explicitly examine the interaction between dollarization and macroeconomic distress using the MI as a composite measure remain limited. Existing research has mainly concentrated on individual indicators such as inflation, exchange rate volatility, or monetary aggregates, often overlooking the joint effect of inflation and unemployment on financial behavior. Moreover, evidence on the bidirectional relationship between macroeconomic distress and dollarization, especially at the subnational level, is scarce.

Given this background, the purpose of this study is to examine the bidirectional relationship between macroeconomic distress, measured by the MI, and financial dollarization using regional data from Türkiye. Türkiye provides a particularly informative case, as it has long been characterized by high and volatile inflation, pronounced exchange rate fluctuations, labor market pressures, and pervasive dollarization behavior (Özkul, 2021). The existence of significant regional differences within the country makes subnational analysis essential for understanding how macroeconomic distress and dollarization interact across different economic environments. Using annual data for 26 Nomenclature of Territorial Units for Statistics-2 (NUTS-2) regions over 2005–2021, we construct the MI as the sum of regional inflation and unemployment rates and measure dollarization as the share of deposits denominated in foreign currency in total deposits.

We employ a range of advanced panel econometric techniques to capture both long-run equilibrium and short-run adjustment, considering cross-sectional dependence and heterogeneity. Long-term relationships are examined using Pedroni, Kao residual-based, Johansen Fisher panel, and Westerlund cointegration tests. Short-run interactions are analyzed through a Panel Vector Error Correction Model (PVECM) and the Cross-Sectionally Augmented ARDL (CS-ARDL) approach, which explicitly accounts for unobserved common factors. Dynamic Ordinary Least Squares (DOLS) estimation helps determine the magnitude and trajectory of the MI’s long-run effects on dollarization.

This paper makes several significant contributions to the literature. First, we provide empirical evidence of the bidirectional relationship between macroeconomic distress and dollarization, highlighting feedback effects that amplify economic instability. Second, by employing regional NUTS-2 data, we capture within-country variation that is often obscured in national-level studies and offers novel insights into subnational adjustment processes. Finally, the application of second-generation panel techniques ensures reliable inference when cross-sectional dependence and structural interlinkages are present.

The paper is organized as follows: Section 2 establishes the theoretical framework linking MI and dollarization. Section 3 describes the data and the econometric methodology. Section 4 and Section 5 present and discuss the results, respectively. While Section 6 focuses on policy implications, the paper closes with concluding remarks.

2. Theoretical Linkages Between the Misery Index and Dollarization

2.1. The Misery Index as an Indicator of Macroeconomic Distress

The macroeconomic stability literature emphasizes that deteriorations in fundamental indicators such as inflation and unemployment play an essential role in shaping economic agents’ expectations and behavior (Friedman, 1977; Lucas, 1976). Accelerating inflation or rising unemployment both generate immediate welfare losses for households and firms and intensify uncertainty regarding future income, prices, and macroeconomic conditions. As uncertainty rises, expectations become more fragile, and risk perceptions rise, influencing both consumption and portfolio decisions.

Within this framework, the MI (Okun, 1970) offers a broad indicator of macroeconomic distress. Compared with single-indicator measures, the MI captures the joint burden imposed by price instability and labor market slack, reflecting both current welfare losses and forward-looking uncertainty. From the standpoint of expectations theory (Muth, 1961; Fama, 1970), economic agents continuously update their beliefs based on observed and anticipated macroeconomic conditions. When inflation and unemployment rise simultaneously, the perceived trustworthiness of the domestic currency as a means to preserve value deteriorates, creating conditions for behavioral responses such as currency substitution and precautionary shifts toward foreign-currency assets (Barajas & Morales, 2003). But recent empirical evidence also points to a reverse channel, suggesting that adverse currency conditions, such as exchange rate movements, can translate into higher macroeconomic distress as reflected in the misery index, especially during periods of elevated uncertainty (Sánchez & Arias Guzmán, 2025).

2.2. From Macroeconomic Distress to Dollarization

The dollarization literature commonly identifies high and volatile inflation as a key factor of currency substitution and financial dollarization. Mishkin (1996) contends that price volatility weakens trust in the domestic currency, reducing its effectiveness as a store of value and encouraging economic agents to shift their portfolios toward assets denominated in foreign currencies. Similarly, Honohan and Shi (2001) emphasize that inflation dynamics have a significant and direct impact on money demand, increasing the ease of substitution between domestic and foreign currencies in inflationary environments.

Empirical studies also support this relationship. Reinhart et al. (2003) show that economies experiencing chronic inflation and recurrent currency crises often develop persistent dollarization patterns, characterized by hysteresis effects in which past instability shapes future financial behavior. Edwards and Magendzo (2003) likewise emphasize that repeated currency crises can entrench dollarization by generating hysteresis, making foreign-currency usage persistent even after macroeconomic conditions improve. Studies such as Tweneboah and Alagidede (2019) and Zhao et al. (2025) document that high inflation rates reinforce dollarization as a rational response to weakened macroeconomic conditions. Meanwhile, Neanidis and Savva (2009) identify inflation as a key short-run determinant of dollarization in transition economies. Earlier evidence from Latin America likewise indicates that dollarization is more prevalent in countries with sustained episodes of inflation (Guidotti & Rodriguez, 1992).

Country-specific studies for Türkiye reach similar conclusions. Kara et al. (2017) show that inflation uncertainty has persistent effects on foreign currency demand, strengthening deposit dollarization. Özer (2022) identifies inflation as an important determinant of dollarization in Türkiye. Aktaş and Aydınlık (2022), using regional data, find that inflation exerts the strongest influence on deposit dollarization across NUTS-2 regions. Overall, these findings suggest that macroeconomic distress, as captured by inflation and unemployment, creates clear incentives for economic agents to protect their wealth through foreign-currency holdings. Accordingly, increases in the MI are expected to be associated with greater dollarization.

Recent evidence, however, suggests that the relationship between inflation and dollarization may be asymmetric. Gil and Perez (2025) show that in developing economies with a history of high inflation, financial dollarization can remain persistent even after inflation rates decline. They argue that past inflation-driven instability generates lasting fears of renewed inflation, leading economic agents to maintain foreign-currency positions despite improvements in price stability. This dollarization hysteresis mechanism helps explain why reductions in inflation may have only a limited long-run effect on dollarization.

2.3. Dollarization as a Feedback Channel to Macroeconomic Distress

Dollarization is not only a consequence of macroeconomic instability but also a factor that can amplify it. Higher financial dollarization lowers the effectiveness of monetary policies by reducing central banks’ control over domestic liquidity and interest rates. Mengesha and Holmes (2015) present evidence that rising dollarization contributes to higher inflation, whereas Park and Son (2023) argue that lower dollarization levels enhance the effectiveness of monetary policy instruments in stabilizing prices and supporting economic growth.

In addition, Reinhart et al. (2003) emphasize that dollarization rises exchange rate pass-through, strengthening transmission of exchange rate fluctuations into domestic prices and thereby destabilizing inflation expectations. This process can lead to reinforcing effects, whereby dollarization fuels inflationary pressures that further undermine confidence in the domestic currency. Many empirical studies support this view, including Bailey (2005), Karacal and Bahmani-Oskooee (2008), Ghalayini (2011), Yılmaz and Uysal (2019), and Özkul (2021).

Dollarization may also affect labor market outcomes through exchange rate and financial channels. Currency depreciation associated with dollarization can tighten domestic liquidity conditions, raise interest rates, and suppress investment and production, ultimately exerting downward pressure on employment (Stryker, 1999). Acar Balaylar (2011) finds that rise of real exchange rate negatively affects employment in Türkiye’s manufacturing sector. Yiğiteli (2022) shows that higher dollarization and real wage growth are associated with rising unemployment across Turkish NUTS-2 regions. Similarly, Galindo et al. (2007) document adverse employment effects of liability dollarization, and Yeyati (2006) highlights balance-sheet effects that constrain economic activity and increase unemployment. These channels suggest that dollarization can directly influence the inflation and unemployment components of the MI.

3. Data and Methodology

3.1. Data and Variable Definitions

The objective of this research is to examine the empirical relationship between macroeconomic distress and financial dollarization in Türkiye using regional panel data. The analysis employs annual data for 26 NUTS-2 regions over the period 2005–2021. This data structure enables empirical analysis to explore the long- and short-run interactions between macroeconomic distress and financial dollarization at the regional level in Türkiye.

Macroeconomic distress is measured using the MI (Okun, 1970), which is commonly used in the literature as a composite indicator capturing the welfare losses associated with price instability and labor market slack. MI is calculated using annual regional inflation rates based on the consumer price index (CPI) and regional unemployment rates.

Financial dollarization is measured by deposit dollarization (hereafter DOL), defined as the ratio of foreign-currency-denominated deposits to total bank deposits. This measure is commonly used in empirical studies as an indicator of residents’ preference for foreign currency as a store of value. Foreign-currency deposits are measured in nominal terms, consistent with standard practice in the DOL literature.

Regional inflation (CPI-based) and unemployment data for NUTS-2 regions are obtained from the Turkish Statistical Institute (TURKSTAT, 2025). Data on foreign-currency deposits and total deposits are sourced from the Banks Association of Türkiye (BAT, 2025). The sample period is restricted to 2005–2021 due to the availability of regional CPI data in Türkiye, which began in 2005 and was discontinued after April 2022. Descriptive statistics for the variables used in the analysis are given in

Table 1. Statistic summary.

Variable	Mean	Median	Max	Min	Std.Dev	Skewness	Kurtosis
MI	22.25199	19.755	66.86	10.73	9.285485	1.881185	6.756742
DOL	0.312139	0.30119	0.629239	0.056906	0.10599	0.35954	2.868256

.

3.2. Econometric Framework and Cross-Sectional Dependence

The empirical analysis relies on panel time-series techniques that accommodate both regional heterogeneity and dynamic interactions across time. Prior to implementing unit root and cointegration procedures, it is necessary to determine whether the regional units exhibit cross-sectional dependence.

In regional panel datasets drawn from a single country, economic regions are commonly exposed to shared macroeconomic disturbances, centralized monetary and fiscal policies, and comparable structural characteristics. As a result, the assumption of cross-sectional independence is unlikely to hold. Failure to account for such interdependencies may invalidate first-generation panel unit root tests and lead to biased test statistics and weakened inferential power (Baltagi & Hashem Pesaran, 2007).

To formally evaluate cross-sectional dependence, the Breusch–Pagan LM test, Pearson LM test, Pearson CD test, and Friedman chi-square test are employed. The outcomes of these tests are presented in

Table 2. Cross sectional test.

Test	Statistic	d.f.	Prob.
Breusch-Pagan Chi-square	3022.058	325	0.0000
Pearson LM Normal	104.7675		0.0000
Pearson CD Normal	50.6248		0.0000
Friedman Chi-square	272.1629	16	0.0000

.

The test statistics consistently reject the null hypothesis of cross-sectional independence at conventional significance levels, providing strong evidence of interregional dependence. This study therefore adopts second-generation panel methodologies, which explicitly account for cross-sectional dependence and mitigate the size distortions and spurious inference associated with first-generation models (Baltagi & Hashem Pesaran, 2007).

3.3. Panel Unit Root Methodology

Given the presence of cross-sectional dependence, the stationarity properties of the variables are investigated using second-generation panel unit root tests. In this context, the Cross-Sectionally Augmented Dickey–Fuller (CADF) and Cross-Sectionally Augmented IPS (CIPS) tests are applied.

The CADF test, developed by Pesaran (2007), modifies the conventional ADF regression by incorporating cross-sectional averages of both the level and first difference of the series. This augmentation allows the test to control for unobserved common factors that simultaneously affect all regions. For each cross-sectional unit, the CADF regression is specified as:

$$\mathsf{\Delta }{y}_{it}=a_i+b_i{y}_{i,t-1}+c_i{\overline{y}}_{t-1}+d_i\mathsf{\Delta }{\overline{y}}_t+{ϵ}_{it}$$

(1)

where ${\overline{y}}_t$ denotes the cross-sectional average and the coefficient $b_i$ provides the unit-specific unit root test statistic. The inclusion of cross-sectional averages captures the effects of common shocks affecting all regions. The lagged dependent variable allows for unit-specific persistence.

As the CADF test produces individual unit root statistics for each region, the CIPS test (Pesaran, 2007) obtains a panel-level statistic by averaging the individual CADF statistics:

$$\mathrm{CIPS}={\displaystyle \frac{1}{N}}\sum _{i=1}^{N}CAD{F}_i$$

(2)

The CIPS test extends the IPS test of Im et al. (2003) by explicitly allowing for cross-sectional dependence. It provides reliable inference in panels of varying dimensions and offers a consistent framework for stationarity testing in the presence of common factors (Breitung & Pesaran, 2008; Choi, 2015).

3.4. Panel Cointegration Analysis

To assess whether a stable long-run relationship exists between the MI and DOL, a set of complementary panel cointegration tests is employed. These tests differ in their treatment of heterogeneity and cross-sectional dependence, thereby allowing for a more comprehensive evaluation of long-run dynamics.

First, the residual-based panel cointegration tests proposed by Pedroni (1999, 2004) are implemented. These tests permit heterogeneity in both intercepts and slope coefficients across regions and generate seven statistics that capture within-dimension and between-dimension cointegration dynamics.

Second, the Kao (1999) cointegration test is applied as a more restrictive alternative, imposing homogeneous slope coefficients across cross-sectional units. This test is based on an ADF-type unit root test applied to the residuals obtained from the estimated long-run relationship.

Third, the Johansen Fisher panel cointegration approach is employed by combining individual Johansen trace and maximum eigenvalue statistics using Fisher’s aggregation method (Maddala & Wu, 1999). Therefore, the Johansen-Fisher test is frequently preferred, especially in systems-based cointegration analyses, due to its multivariate nature.

Finally, the Westerlund (2007) panel cointegration test is utilized. Unlike residual-based methods, this approach directly evaluates the presence of an error-correction mechanism and is particularly robust to cross-sectional dependence when bootstrap procedures are applied.

Employing both residual-based and system-based cointegration techniques ensures that the long-run relationship between MI and DOL is examined from multiple methodological perspectives.

3.5. Short-Run Dynamics: PVECM and CS-ARDL

After establishing the existence of cointegration, short-run interactions and adjustment dynamics are analyzed using a PVECM and the CS-ARDL framework.

The PVECM extends the conventional vector error correction model to a panel setting by allowing for heterogeneous short-run coefficients across regions while incorporating an error-correction term that captures deviations from long-run equilibrium (Holtz-Eakin et al., 1988; Pesaran et al., 1999; Johansen, 1995). The two-equation PVECM system is specified as:

$$\mathsf{\Delta }M{I}_{i,t}={\alpha }_{1i}+\sum _{k=1}^p{\beta }_{1ik}\mathsf{\Delta }M{I}_{i,t-k}+\sum _{k=1}^p{\gamma }_{1ik}\mathsf{\Delta }DO{L}_{i,t-k}+{\lambda }_{1i}EC{T}_{i,t-1}+{\epsilon }_{1i,t}$$

(3)

$$\mathsf{\Delta }DO{L}_{i,t}={\alpha }_{2i}+\sum _{k=1}^p{\beta }_{2ik}\mathsf{\Delta }DO{L}_{i,t-k}+\sum _{k=1}^p{\gamma }_{2ik}\mathsf{\Delta }M{I}_{i,t-k}+{\lambda }_{2i}EC{T}_{i,t-1}+{\epsilon }_{2i,t}$$

(4)

A statistically significant and negative error-correction coefficient indicates long-run causality (Engle & Granger, 1987; Johansen, 1995).

In addition, the CS-ARDL approach, an adaptation of Pesaran’s (2006) Common Correlated Effects approach to ARDL models, is used to estimate both short-run and long-run relationships while controlling for unobserved common factors through cross-sectional averages. This method is particularly suitable for panels characterized by cross-sectional dependence and heterogeneity (Chudik & Pesaran, 2015). Because it maintains heterogeneity in panel datasets, this method allows different long-run coefficients for each unit and offers strong forecasting performance in situations where economic structures are diverse and common shocks are prominent (Eberhardt & Bond, 2009; Kapetanios et al., 2011). The CS-ARDL models are specified as:

$$\mathsf{\Delta }M{I}_{i,t}={\varphi }_i\left(M{I}_{i,t-1}-{\theta }_{i}DO{L}_{i,t-1}\right)+\sum _{k=1}^p{\lambda }_{ik}\mathsf{\Delta }M{I}_{i,t-k}+\sum _{k=0}^q{\delta }_{ik}\mathsf{\Delta }DO{L}_{i,t-k}+{\psi }_i{\overline{MI}}_t+{\phi }_i{\overline{DOL}}_t+{\epsilon }_{i,t}$$

(5)

$$\mathsf{\Delta }DO{L}_{i,t}={\varphi }_i\left(DO{L}_{i,t-1}-{\theta }_{i}M{I}_{i,t-1}\right)+\sum _{k=1}^p{\lambda }_{ik}\mathsf{\Delta }DO{L}_{i,t-k}+\sum _{k=0}^q{\delta }_{ik}\mathsf{\Delta }M{I}_{i,t-k}+{\psi }_i{\overline{DOL}}_t+{\phi }_i{\overline{MI}}_t+{\eta }_{i,t}$$

(6)

The panel averages $({\overline{MI}}_t,{\overline{DOL}}_t)$ included in these models act as a control mechanism against unobservable common factors (Chudik et al., 2016). While the ${\theta }_i$ coefficient expresses the long-run relationship between variables, a statistically negative and significant ${\varphi }_i$ coefficient indicates long-run cointegration (Chudik & Pesaran, 2015). The ${\lambda }_{ik}$ and ${\delta }_{ik}$ coefficients of the lagged difference terms for MI and DOL allow for the evaluation of short-run effects within the same model framework.

3.6. Long-Run Estimation: DOLS

To estimate the magnitude and direction of long-run effects between the MI and DOL, DOLS estimation is employed. The DOLS estimator addresses potential endogeneity and serial correlation by augmenting the cointegration equation with leads and lags of the first differences of the explanatory variables (Stock & Watson, 1993).

The long-run impact of DOL on macroeconomic distress is estimated using the following DOLS specification:

$$M{I}_t=\alpha +\beta DO{L}_t+\sum _{k=-q}^q{\gamma }_k\mathsf{\Delta }DO{L}_{t-k}+u_t$$

(7)

To examine the reverse relationship, an alternative DOLS specification is estimated in which DOL is treated as the dependent variable:

$$DO{L}_t=\alpha +\theta M{I}_t+\sum _{k=-q}^q{\varphi }_k\mathsf{\Delta }M{I}_{t-k}+{\epsilon }_t$$

(8)

Estimating these equations separately allows for a direct comparison of long-run effects in both directions. Due to its strong small-sample properties, DOLS is widely used in both panel and time-series cointegration analyses (Kao & Chiang, 2001; Mark & Sul, 2003).

4. Empirical Results

This section presents the empirical findings on the relationship between the MI and DOL in Türkiye’s NUTS-2 regions. The analysis proceeds sequentially, beginning with panel unit root tests, followed by panel cointegration tests, short-run dynamics, and long-run coefficient estimation.

4.1. Panel Unit Root Test Results

The analysis first evaluates whether the variables exhibit unit root behavior. Considering the cross-sectional dependence documented in preceding section, stationarity is assessed using second-generation panel unit root tests, specifically the CIPS and CADF procedures. The outcomes of these tests are summarized in

Table 3. Panel unit root tests with cross-sectional dependence.

Variable	Test Level	Test Name	t-Stat	p-Value
MI	Level	Pesaran (CIPS)	0.00000	≥0.10
	Level	Cross-sectional ADF	0.00000	≥0.10
	1.Difference	Pesaran (CIPS	−3.73484	<0.01
	1.Difference	Cross-sectional ADF	−6.03309	<0.01
DOL	Level	Pesaran (CIPS)	0.00000	≥0.10
	Level	Cross-sectional ADF	0.00000	≥0.10
	1.Difference	Pesaran (CIPS)	−3.39495	<0.01
	1.Difference	Cross-sectional ADF	−5.63172	<0.01

.

The test outcomes indicate that, at levels, both MI and DOL fail to reject the null hypothesis of a unit root. This conclusion is supported by the p-values obtained from both the CIPS and CADF tests, which exceed conventional significance thresholds. Accordingly, neither variable can be considered stationary in level form.

When the first differences of the series are examined, the null hypothesis of a unit root is strongly rejected at the 1% significance level. Thus, both MI and DOL become stationary after first differencing. These results suggest that the variables are integrated of order one, I(1), thereby satisfying the prerequisite for conducting panel cointegration analysis.

4.2. Panel Cointegration Test Results

Given that both series are integrated of the same order, the analysis proceeds by examining whether MI and DOL are linked by a long-run equilibrium relationship. To ensure the robustness of the findings, multiple panel cointegration tests are implemented, namely the Pedroni and Kao residual-based tests, the Johansen Fisher panel cointegration approach, and the Westerlund panel cointegration test. The results are reported in

Table 4. Long-run panel cointegration test results.

	Test	Statistic	Prob.
Pedroni Residual Cointegration Test	Panel v-Statistic	5.318796	0
	Panel rho-Statistic	−3.321586	0.0004
	Panel PP-Statistic	−3.233614	0.0006
	Panel ADF-Statistic	−5.64428	0
Kao Residual Cointegration Test	ADF	−6.200608	0.0000
Johansen Fisher Panel Cointegration Test	Fisher Stat	106.1	0.0000
Westerlund Panel Cointegration Tests	Fisher Stat	−2.8913	0.0000

.

The evidence reported in Table 4 leads to the rejection of the null hypothesis of no cointegration under all test specifications. This outcome confirms the existence of a long-run cointegrating relationship between MI and DOL. Accordingly, the two variables exhibit long-run comovement despite the presence of short-run disturbances.

4.3. Short-Run Causality: PVECM Results

The short-run effects between MI and DOL are examined using PVECM. This framework allows for the detection of short-run causal relationships and incorporates the long-run equilibrium adjustment process. The PVECM results are presented in

Table 5. PVECM results.

Dependent Variable	Excluded	Chi-Sq	Df	Prob.
MI	D(DOL)	80.24474	10	0.0000
	All	80.24474	10	0.0000
DOL	D(MI)	44.07318	10	0.0000
	All	44.07318	10	0.0000

Note: The lag length is determined according to the Akaike Information Criterion (AIC), and the optimal lag structure of the VAR model is selected accordingly.

.

The results indicate that, in both model specifications, the test statistics are statistically significant at the 5% level. This implies the existence of short-run bidirectional causality between MI and DOL. In other words, short-term changes in DOL affect the MI, and short-term changes in macroeconomic distress also influence DOL behavior.

4.4. Short-Run Dynamics: CS-ARDL Results

Short-run interactions are further examined using the CS-ARDL framework, which explicitly accounts for cross-sectional dependence and unobserved common factors. The corresponding estimation results are presented in

Table 6. CS-ARDL result.

Dependent Variable	Excluded	CD Statistic	Prob.
MI	D(DOL)	54.05	0.0000
DOL	D(MI)	−2.10	0.0355

.

The CS-ARDL estimates reveal statistically significant short-run effects in both directions. When MI is treated as the dependent variable, the Wald test for short-run changes in DOL produces a significant CD statistic (CD = 54.05, p = 0.0000), indicating that short-run fluctuations in DOL have a significant causal effect on the MI.

Conversely, when DOL is treated as the dependent variable, short-run changes in MI also produce statistically significant effects (CD = −2.10, p = 0.0355). These findings confirm bidirectional short-run causality between MI and DOL and reinforce the PVECM results.

4.5. Long-Run Coefficient Estimates: DOLS Results

After establishing the existence of both long-run cointegration and short-run interactions, long-run coefficients are estimated using the DOLS methodology. The corresponding results are reported in

Table 7. Cointegrating regression long-run estimation dynamic least squares (DOLS).

Dependent Variable	Variable	Coefficient	Std. Error	t-Statistic	Prob.
MI	DOL	74.64377	4.363748	17.10543	0.0000
DOL	MI	0.006707	0.000715	9.374467	0.0000

.

In both model specifications, the estimated long-run coefficients are statistically significant at the 5% level. When MI is the dependent variable, a one-unit increase in DOL leads to an average increase of approximately 74.64 units in the MI in the long run. This finding suggests that greater DOL amplifies macroeconomic instability by intensifying inflationary pressures and labor market distortions.

When DOL is treated as the dependent variable, a one-unit increase in the MI leads to a positive but relatively small increase in DOL (approximately 0.0067 units). This asymmetry implies that while rising DOL strongly worsens macroeconomic distress, increases in macroeconomic instability induce only a limited long-run increase in DOL.

Overall, the DOLS results confirm a statistically significant, bidirectional long-run relationship between MI and DOL and reveal that the relationship is asymmetric. DOL plays a more dominant role in driving macroeconomic distress than MI does.

5. Discussion

The results provide clear evidence of a long-run cointegration relationship between macroeconomic distress, measured by the MI and DOL across Türkiye’s NUTS-2 regions. The existence of cointegration, reported consistently by the Pedroni, Kao, Johansen Fisher and Westerlund tests, indicates that macroeconomic instability and DOL evolve jointly over time despite short-run fluctuations. This finding supports expectations-based and portfolio substitution theories, which argue that sustained inflationary pressures and broader macroeconomic stress erode confidence in the domestic currency and encourage foreign currency holdings (Mishkin, 1996; Honohan & Shi, 2001).

The short-run dynamics estimated using both the PVECM and CS-ARDL frameworks point to bidirectional causality between MI and DOL. This feedback channel demonstrates that short-run increases in DOL intensify macroeconomic distress, whereas short-run deteriorations in inflation and unemployment conditions simultaneously reinforce DOL behavior. Such bidirectional adjustment processes are consistent with the hysteresis and feedback effects emphasized by Reinhart et al. (2003), where past instability shapes current financial behavior and weakens the effectiveness of stabilization policies.

A key contribution lies in the asymmetric nature of the long-run relationship, as revealed by the DOLS estimates. Although both directions of causality are statistically significant, the magnitude of the long-run coefficients differs substantially. Increases in DOL have a larger impact on the MI. This asymmetry suggests that although macroeconomic distress triggers currency substitution, the macroeconomic consequences of DOL (through weakened monetary transmission, increased exchange rate pass-through, and financial fragility) are considerably more severe and persistent.

These findings are consistent with international evidence indicating that high and persistent inflation is a main driver of DOL (Mishkin, 1996; Honohan & Shi, 2001; Reinhart et al., 2003; Tweneboah & Alagidede, 2019; Zhao et al., 2025) as well as with country-specific evidence for Türkiye showing a unidirectional causal relationship from DOL to inflation and employment at the national level, in line with our findings of asymmetric effects (Özkul, 2021).

Moreover, the bidirectional short-run causality and long-run asymmetric effects observed in the analysis are in line with empirical studies documenting that DOL amplifies labor market fragility and balance-sheet constraints, thereby shaping unemployment patterns and real economic adjustment (Yeyati, 2006; Galindo et al., 2007; Acar Balaylar, 2011; Yiğiteli, 2022). Overall, the results confirm that DOL is not merely a passive response to macroeconomic instability, but rather a factor that intensifies economic distress over time.

6. Policy Implications

The empirical findings yield several important policy implications for economies characterized by persistent inflation volatility and partial DOL, such as Türkiye.

First, the substantial long-run impact of DOL on macroeconomic distress implies that stabilization strategies focusing exclusively on inflation or unemployment reduction may be insufficient if DOL is not addressed simultaneously. Policies aimed at restoring confidence in the domestic currency (e.g., credible disinflation strategies, transparent monetary policy frameworks, and institutional arrangements that enhance policy predictability) are essential for weakening the link between DOL and macroeconomic instability.

Second, bidirectional short-run causality indicates the risk of self-reinforcing cycles between macroeconomic instability and DOL. Short-term macroeconomic shocks can rapidly translate into increased foreign currency demand, which in turn exacerbates inflationary pressures and labor market distortions. Coordinated fiscal and monetary policies that mitigate inflation volatility and reduce uncertainty may therefore help dampen these feedback effects. In this context, macroprudential measures that discourage excessive foreign currency holdings can help reduce short-term exchange rate fluctuations.

The introduction of the exchange-rate-protected deposit scheme (KKM) in late 2021 illustrates the policy relevance of these findings. KKM emerged as a non-standard stabilization tool in response to rapidly rising DOL and sharp exchange rate depreciation, aiming to curb foreign currency demand without tightening monetary policy. Consistent with the bidirectional short-run causality identified in this study, the scheme was initially effective in stabilizing the exchange rate and slowing DOL through expectations management and portfolio substitution. This experience implies that the acceleration of DOL is sensitive to short-run policy interventions, especially when confidence effects dominate asset allocation decisions.

However, the longer-term consequences of KKM also underscore the asymmetric effects documented in the analysis. By transferring exchange rate risk from private agents to the public sector, the scheme mitigated immediate currency pressures at the cost of reinforcing inflationary dynamics and weakening monetary policy transmission. Persistent inflation eroded real incomes and labor market conditions, thereby sustaining the macroeconomic distress that, in the long run, incentivizes DOL. In this sense, KKM exemplifies the limitations of de-dollarization strategies that rely on temporary guarantees rather than durable disinflation and employment stabilization, reinforcing the study’s finding that the effects of adverse macroeconomic shocks are more persistent than gains achievable through short-run policy interventions.

Third, the asymmetric long-run relationship highlights that even moderate reductions in DOL can generate disproportionately large gains in macroeconomic stability. Although increases in the MI result in only modest long-term growth in DOL, decreasing DOL could significantly help reduce inflationary and employment pressures. Policies that promote the attractiveness of domestic currency instruments (e.g., deepening local currency financial markets and improving risk-sharing mechanisms) may therefore yield significant welfare gains.

Finally, the regional aspect of the analysis shows that uniform policies may not be the most effective approach. Regions with greater sensitivity to DOL may require targeted interventions, such as region-specific credit policies, employment programs, or investment incentives. Incorporating regional differences into macroeconomic policy design can improve effectiveness and reduce adjustment costs.

7. Conclusions

This study investigates the relationship between macroeconomic distress and DOL in Türkiye using annual regional panel data for the period 2005–2021. Employing second-generation panel econometric techniques that explicitly account for cross-sectional dependence and heterogeneity, the analysis provides evidence of both long-run equilibrium relationships and short-run dynamic interactions between the MI and DOL.

The findings demonstrate that macroeconomic distress and DOL are cointegrated in the long run and exhibit bidirectional causality in the short run. Remarkably, the long-run effects are asymmetric: DOL has a substantially greater impact on macroeconomic distress than the reverse. This outcome underscores the critical role of DOL not only as a response to instability but also as a structural transmission channel that exacerbates economic distress by weakening monetary policy effectiveness and heightening vulnerability to shocks.

By focusing on regional variations, the study provides the first subnational evidence on the spatial dimension of the DOL–MI relationship in Türkiye, moving beyond results drawn from national-level analyses. Overall, the findings indicate that regional disparities matter both for the emergence and the persistence of DOL, reinforcing the need for region-sensitive macroeconomic policy frameworks and sustained disinflation strategies to achieve durable macroeconomic stability.

Future studies could extend this research by employing alternative measures of DOL, such as credit DOL, or by exploring nonlinear and threshold effects. Comparative cross-country studies may further illuminate how institutional quality and policy regimes shape the interaction between DOL and macroeconomic instability.