5.2. The Unit Root Results
Although it allows the examination of the relationships between variables that are stationary at different degrees, such as [I(0)] and [I(1)], the degree of stationarity of the variables in the A-ARDL model should not be [I(2)] (Sam et al., 2019). Therefore, the degrees of stationarity of the variables in the growth and inequality models were examined by conventional ADF, PP, and ZA unit root tests with structural breaks, which are frequently used in the empirical literature, and the results are reported in Table 4.
Table 4
Test Statistics |
CT | ADF | PP | ZA |
Variables | [I(0)] | [I(1)] | L | [I(0)] | [I(1)] | L | [I(0)] | [I(1)] | L |
LY | -2.583 | -5.477* | 0 | -2.595 | -5.498* | 3 | -5.342** [2001] | ─ | 2 |
LGI | -2.644 | -5.850* | 0 | -2.707 | -5.944* | 1 | -3.727 [2005] | -6.639* [2003] | 2 |
LEQ | -3.078 | -3.783** | 1 | -1.919 | -1.874 | 4 | -4.808 [2006] | -5.901* [2014] | 1 |
LRD | -4.042** | ─ | 0 | -4.175** | ─ | 5 | -5.335** [1998] | ─ | 0 |
LEI | -2.673 | -5.277* | 1 | -2.359 | -4.245** | 3 | -4.680 [2008] | -16.853* [2011] | 1 |
GY | -5.739* | ─ | 0 | -6.211* | ─ | 5 | -5.955* [2002] | ─ | 0 |
GEQ | -2.184 | -5.776* | 0 | -2.001 | -5.076* | 3 | -3.685 [2013] | -7.383* [2013] | 0 |
GTO | -3.684** | ─ | 2 | -5.034* | ─ | 3 | -6.628* [1998] | ─ | 1 |
GRD | -7.261* | ─ | 0 | -7.321* | ─ | 3 | -9.170* [1999] | ─ | 0 |
GEI | -2.641 | -5.932* | 0 | -2.957 | -6.009* | 3 | -6.430* [2006] | ─ | 4 |
FN | -2.205 | -5.196* | 3 | -1.791 | -5.576* | 3 | -3.669 [2001] | -6.511* [2012] | 3 |
FS | -2.422 | -3.261* | 1 | -1.738 | -3.361** | 4 | -4.551 [2008] | -5.331* [2004] | 1 |
RL | -2.692 | -6.127* | 2 | -2.601 | -6.108* | 3 | -5.232** [2003] | ─ | 2 |
TR | -2.533 | -4.831* | 0 | -2.579 | -5.223* | 3 | -4.561 [2003] | -6.882* [2006] | 3 |
The ADF and PP test statistics in Table 4 indicate that the null hypotheses of a unit root are rejected at the 1% or 5% level of significance for LRD, GY, GTO, and GRD variables at level value [I(0)] and for all other variables at first differences [I(1)]. These results indicate that the variables LRD, GY, GTO, and GRD are stationary at level value, while all other variables are stationary when the first difference is taken. When the ZA test statistics in Table 4 are examined, it is seen that if the effects of structural breaks are taken into account, more variables such as LY, LRD, GY, GTO, GRD, GEI, and RL become stationary at level value [I(0)]. However, despite considering the effects of structural breaks, it is observed that LGI, LEQ, LEI, GEQ, FN, FS, and TR variables are stationary in their first differences, not at level values. These results are obtained by rejecting the null hypotheses of unit root at 1% or 5% significance level for LY, LRD, GY, GTO, GRD, GEI, and RL variables at level value [I(0)] and for all other variables at first differences [I(1)]. These results show that the variables in the growth and inequality models are stationary in different degrees and, at most, the first differences. Examining the long-term relationships between the model variables with the A-ARDL is appropriate.
On the other hand, it is seen that the structural break dates determined for the variables by the ZA test generally coincide with the economic contraction periods caused by internal (1999, 2000, and 2001) and external shocks (1997-East Asia Financial Crisis and 2008 Global Financial Crisis) in the Turkish economy. For example, in the growth (LY and GY) variables, which were stationary at their level values, the structural break dates that occurred in 2001 and 2002, the Turkish economy expanded by 6.45% after experiencing a contraction of about 5.75%, according to the World Bank data. Similarly, it is seen that the structural break dates determined to have occurred in the reform indicators in the form of FN, FS, RL, and TR coincide with the dates when the structural adjustment programs following the economic contraction periods in the Turkish economy gained importance and prevalence.
5.3. The A-ARDL Results
After determining that the variables in the growth and inequality models were integrated at most [I(1)] level, at this stage, the optimal A-ARDL model (with a maximum lag length of 2) is determined and estimated with AIC. The F-overall, F-independent, and t-dependent boundary test results of the predicted A-ARDL model are presented in Table 5. In contrast, the results of the diagnostic tests are reported in Table 6.
Table 5
A-ARDL Cointegration Results
Model | Optimum Lags( AIC) | Test Statistics |
F-Overall | F-Independent | t-Dependent |
LY=(LEQ, LGI, LRD, LEI, FN) | (1.1.0.2.2.1) | 8.644* | 10.225* | -6.711* |
LY= (LEQ, LGI, LRD, LEI, FS) | (1.0.0.0.0.1) | 53.513* | 63.854* | -14.555* |
LY= (LEQ, LGI, LRD, LEI, RL) | (1.0.1.0.0.1) | 57.478* | 68.577* | -14.955* |
LY =(LEQ, LGI, LRD, LEI, TR) | (1.0.1.1.0.1) | 5.8394** | 6.827* | -5.568* |
GEQ= ( GY, GRD, GEI, GTO, FN) | (1.0.0.0.0.1) | 6.118** | 5.303** | -4.276** |
GEQ= ( GY, GRD, GEI, GTO, FS) | (1.0.0.0.0.1) | 6.351** | 5.544** | -4.687** |
GEQ= ( GY, GRD, GEI, GTO, RL) | (1.1.2.0.1.1) | 8.811* | 8.289* | -4.546** |
GEQ= ( GY, GRD, GEI, GTO, TR) | (1.1.2.2.1.1) | 9.288* | 8.876* | -5.410* |
Critical Values (Upper Bound) | Resource | %1 | %5 | %10 |
t-dependent | Pesaran et al. (2001) | -4.790 | -4.190 | -3.860 |
F-overall | Narayan (2005) | 6.370 | 4.608 | 3.858 |
F-independent | Sam et al. (2019) | 6.480 | 4.540 | 3.760 |
Table 6
A-ARDL Diagnostic Test Results
Model | Diagnostic Check Statistics |
ARCH | LM | JB | RR | R2 | F | CS(CS2) |
LY=(LEQ, LGI, LRD, LEI, FN) | 2.395 (0.112) | 2.660 (0.102) | 0.451 (0.798) | 0.940 (0.526) | 0.97 | 43.034 (0.000) | S(S) |
LY= (LEQ, LGI, LRD, LEI, FS) | 2.021 (0.166) | 2.267 (0.112) | 0.086 (0.958) | 0.310 (0.581) | 0.94 | 47.276 (0.000) | S(S) |
LY= (LEQ, LGI, LRD, LEI, RL) | 0.988 (0.328) | 0.803 (0.410) | 0.137 (0.933) | 0.009 (0.923) | 0.94 | 44.639 (0.000) | S(S) |
LY =(LEQ, LGI, LRD, LEI, TR) | 0.012 (0.911) | 1.703 (0.202) | 0.172 (0.917) | 0.618 (0.441) | 0.95 | 44.643 (0.000) | S(S) |
GEQ= ( GY, GRD, GEI, GTO, FN) | 0.073 (0.788) | 0.058 (0.944) | 1.341 (0.511) | 0.002 (0.958) | 0.39 | 2.136 (0.080) | S(S) |
GEQ= ( GY, GRD, GEI, GTO, FS) | 0.207 (0.652) | 0.087 (0.916) | 1.557 (0.459) | 0.041 (0.841) | 0.42 | 2.3587 (0.057) | S(S) |
GEQ= ( GY, GRD, GEI, GTO, RL) | 2.570 (0.120) | 0.298 (0.745) | 0.971 (0.615) | 0.196 (0.663) | 0.79 | 6.236 (0.000) | S(S) |
GEQ= ( GY, GRD, GEI, GTO, TR) | 2.408 (0.110) | 0.091 (0.913) | 1.003 (0.605) | 0.339 (0.568) | 0.85 | 7.161 (0.000) | S(S) |
Examining the boundary test statistics in Table 5 reveals that all A-ARDL models described to examine the effects of reforms on growth and inequality are also co-integrated with different degrees of integration. We reach these conclusions by rejecting the null hypothesis and the fact that the F-overall, F-independent, and t-dependent boundary test statistics, which are iteratively calculated (1,000) by Monte Carlo simulations for all defined A-ARDL models, are greater than the upper bound critical table values at a significance level of 1% or 5%. Nevertheless, the F-overall, F-independent, and t-dependent bounds test statistics are significant for all AARDL models, indicating a long-term co-integration relationship between the variables in the growth and inequality models.
Table 6 offers diagnostic test statistics for A-ARDL models, including identification error (Ramsey Reset-RR), autocorrelation (Lagrange Multiplier-LM), heteroscedasticity (Autoregressive Conditional Heteroscedasticity-ARCH), normality (Jarque-Bera-JB), structural stability (Cusum-CS and Cusum of Squares-CS2), F, and R2. Analyzing the diagnostic test statistics reveals that all A-ARDL models with relatively high R2 values indicating the explanatory power of the dependent variables and statistically significant F-statistics also satisfy the diagnostic stability conditions. In all A-ARDL models, it is assumed that there are no identification errors, autocorrelation, or heteroscedasticity, that the remains exhibit a normal distribution, and that structural stability exists. The probability values of the test statistics calculated for the RR, LM, ARCH, and JB diagnostic tests are greater than 0.05, and the CS and CS2 test results are Stable (S).
5.4. Long and Short-Term Estimation Results
The results of the short-term and long-term A-ARDL coefficients of growth and inequality models that are cointegrated in the long run and satisfy the diagnostic stability conditions are presented in Tables 7 and 8, respectively.
Table 7
Long- and Short-Run Estimation (Growth)
Dependent Variable = LY |
| Financial Sector | Fiscal Sector | Real Sector | Trade Sector |
Variables | Coefficient | Std. Error | Coefficient | Std. Error | Coefficient | Std. Error | Coefficient | Std. Error |
Long-Run Coefficients | | | |
LEQ | 0.5765 | 0.0749* (0.000) | 0.4397 | 0.1052* (0.000) | 0.5370 | 0.0701* (0.000) | 0.4914 | 0.0745* (0.000) |
LGI | 0.3327 | 0.0072* (0.000) | 0.3052 | 0.0123* (0.000) | 0.3283 | 0.0057* (0.000) | 0.3127 | 0.0232* (0.000) |
LRD | 0.1178 | 0.0217* (0.000) | 0.0490 | 0.0811* (0.550) | 0.0867 | 0.0501** (0.028) | 0.0038 | 0.0401 (0.925) |
LEI | 0.8619 | 0.0217* (0.000) | 0.9633 | 0.0478* (0.000) | 0.9016 | 0.0631* (0.000) | 0.9961 | 0.0149* (0.000) |
FN | 0.0151b | 0.0036* (0.000) | ─ | ─ | ─ | ─ | ─ | ─ |
FS | ─ | ─ | -0.0120* | 0.0039 (0.005) | ─ | ─ | ─ | ─ |
RL | ─ | ─ | ─ | ─ | 0.0238** | 0.0113 (0.047) | ─ | ─ |
TR | ─ | ─ | ─ | ─ | ─ | ─ | -0.0205 | 0.0167 (0.232) |
C | 24.8669 | 4.2921* (0.000) | 13.7899 | 2.2224* (0.000) | 12.7973 | 2.1752* (0.000) | 20.0247 | 4.1796* (0.000) |
Short-Run Coefficients | | | |
\({\varDelta \mathbf{L}\mathbf{Y}}_{\mathbf{t}-1}\) | -1.8982 | 0.2828* (0.000) | -0.9546 | 0.0655* (0.000) | -0.9549 | 0.0638* (0.000) | -1.4409 | 0.2588* (0.000) |
\(\varDelta \varvec{L}\varvec{E}\varvec{Q}\) | 1.0942 | 0.2690* (0.000) | 0.4198 | 0.2308*** (0.082) | 0.5128 | 0.2266** (0.034) | 0.7084 | 0.2367* (0.007) |
\(\varDelta \varvec{L}\varvec{G}\varvec{I}\) | 0.2989 | 0.0202* (0.000) | 0.2914 | 0.0192* (0.000) | 0.3135 | 0.0221* (0.000) | 0.2983 | 0.0202* (0.000) |
\(\varDelta \varvec{L}\varvec{R}\varvec{D}\) | -0.1212 | 0.0612*** (0.064) | 0.0468 | 0.0598 (0.441) | 0.0637 | 0.0512 (0.229) | 0.0055 | 0.0545 (0.920) |
\(\varDelta \varvec{L}\varvec{E}\varvec{I}\) | 3.0912 | 0.8030* (0.001) | 2.5864 | 0.5241* (0.000) | 2.3252 | 0.5329* (0.000) | 3.5515 | 0.7455* (0.000) |
\(\varDelta \varvec{F}\varvec{N}\) | 0.0289 | 0.0118** (0.027) | ─ | ─ | ─ | ─ | ─ | ─ |
\(\varDelta \varvec{F}\varvec{S}\) | ─ | ─ | -0.0115 | 0.0112 (0.317) | ─ | ─ | ─ | ─ |
\(\varDelta \varvec{R}\varvec{L}\) | ─ | ─ | ─ | ─ | -0.0234 | 0.0150 (0.145) | ─ | ─ |
\(\varDelta \varvec{T}\varvec{R}\) | ─ | ─ | ─ | ─ | ─ | ─ | -0.0296 | 0.0115** (0.018) |
\({\mathbf{E}\mathbf{C}\mathbf{T}}_{\mathbf{t}-1}\) | -0.8982 | 0.2316* (0.000) | -0.9546 | 0.0483* (0.000) | -0.9549 | 0.0464* (0.000) | -0.5409 | 0.2187* (0.000) |
Table 8
Long- and Short-Run Estimation (Inequality)
Dependent Variable = GEQ |
| Financial Sector | Fiscal Sector | Real Sector | Trade Sector |
Variables | Coefficient | Std. Error | Coefficient | Std. Error | Coefficient | Std. Error | Coefficient | Std. Error |
Long-Run Coefficients | | | |
GY | 0.0380 | 0.0113* (0.005) | 0.0384 | 0.0177** (0.041) | 0.2196 | 0.0447* (0.000) | 0.2046 | 0.0244* (0.000) |
GTO | -0.0503 | 0.0031* (0.000) | -0.0415 | 0.0146* (0.009) | -0.1109 | 0.0081* (0.000) | -0.0634 | 0.0097* (0.000) |
GRD | -0.0239 | 0.0031* (0.000) | -0.0211 | 0.0023* (0.000) | -0.0880 | 0.0085* (0.000) | -0.0934 | 0.0110* (0.000) |
GEI | 0.2728 | 0.1903 (0.166) | 0.0031 | 0.1939 (0.987) | -0.2403 | 0.1466 (0.118) | -0.7763a | 0.2454* (0.006) |
FN | -0.2402 | 0.0547* (0.000) | ─ | ─ | ─ | ─ | ─ | ─ |
FS | ─ | ─ | -0.7163 | 0.2323* (0.005) | ─ | ─ | ─ | ─ |
RL | ─ | ─ | ─ | ─ | -2.8519 | 0.0447* (0.000) | ─ | ─ |
TR | ─ | ─ | ─ | ─ | ─ | ─ | -1.7009 | 0.1913* (0.000) |
C | -0.0543 | 0.1750 (0.758) | 0.0171 | 0.1862 (0.927) | 0.0202 | 0.2012 (0.921) | -0.5827 | 0.1687* (0.003) |
Short-Run Coefficients | | | |
\({\varDelta \mathbf{G}\mathbf{E}\mathbf{Q}}_{\mathbf{t}-1}\) | -0.1874 | 0.1468 (0.214) | -0.2293 | 0.1358 (0.105) | -0.2328 | 0.0914** (0.020) | -0.2737 | 0.0803* (0.004) |
\(\varDelta \varvec{G}\varvec{Y}\) | 0.0072 | 0.0089 (0.436) | 0.0088 | 0.0086 (0.318) | 0.0152 | 0.0065** (0.031) | 0.0184 | 0.0064** (0.011) |
\(\varDelta \varvec{G}\varvec{T}\varvec{O}\) | 0.0004 | 0.0040 (0.931) | 0.0005 | 0.0039 (0.902) | -0.0106 | 0.0033* (0.004) | -0.0065 | 0.0028** (0.030) |
\(\varDelta \varvec{G}\varvec{R}\varvec{D}\) | -0.0045 | 0.0044 (0.326) | -0.0048 | 0.0043 (0.281) | 0.0056 | 0.0035 (0.128) | 0.0098 | 0.0037** (0.017) |
\(\varDelta \varvec{G}\varvec{E}\varvec{I}\) | 0.0511 | 0.1277 (0.692) | 0.0007 | 0.1321 (0.996) | -0.0559 | 0.1306 (0.674) | 0.0708 | 0.1525 (0.649) |
\(\varDelta \varvec{F}\varvec{N}\) | -0.0450 | 0.1299 (0.732) | ─ | ─ | ─ | ─ | ─ | ─ |
\(\varDelta \varvec{F}\varvec{S}\) | ─ | ─ | -0.1642 | 0.1591 (0.312) | ─ | ─ | ─ | ─ |
\(\varDelta \varvec{R}\varvec{L}\) | ─ | ─ | ─ | ─ | 0.3781 | 0.1792** (0.049) | ─ | ─ |
\(\varDelta \varvec{T}\varvec{R}\) | ─ | ─ | ─ | ─ | ─ | ─ | 0.1508 | 0.1107 (0.192) |
\({\mathbf{E}\mathbf{C}\mathbf{T}}_{\mathbf{t}-1}\) | -0.1874 | 0.0476* (0.000) | -0.2293 | 0.0552* (0.004) | -0.2328 | 0.0283* (0.000) | -0.2737 | 0.0320* (0.000) |
The short-run and long-run coefficients of the FN variable are calculated as 0.028 and 0.015, respectively, and are statistically significant, as shown in Table 7 of the growth model results. These findings indicate that a 1% increase in financial sector reforms carried out during the review period to regulate/supervise the central banking and financial system, privatization and restructuring of financial intermediary institutions, etc., resulted in growth increases of 0.028% in the short-run and 0.015% in the long-run for Turkey. Consistent with the findings of Lora and Barrera (1997), Falcetti et al. (2006), Khan and Qayyum (2006), Ostry et al. (2009), and Christiansen et al. (2013), which examined the effects of financial sector reforms on growth, these findings indicate that financial sector reforms promote growth in Turkey in both the short- and long-term.
In contrast, the short-run coefficient of 0.046 and the long-run coefficient of -0.012 calculated for the variable FS appear statistically insignificant and significant, respectively. These results indicate that the fiscal sector reforms carried out in Turkey during the review period did not impact growth in the short term and led to a decrease of 0.012% in the long term. These results, which are consistent with the Correa (2002) study on the effects of fiscal sector reforms on growth in the literature, indicate that fiscal sector reforms carried out in Turkey to regulate/audit public revenues and expenditures, managing public external debts, regulating the social security system, increasing the transparency of economic statistics, etc. inhibit growth in the long run. On the other hand, these findings contradict those of Lora and Barrera (1997) and Morley et al. (1999), who studied the effects of fiscal sector reforms on economic growth.
Similarly, the RL variable's calculated short-run − 0.023 and long-run 0.024 coefficients appear statistically insignificant and significant. These results indicate that the reforms carried out in Turkey's real sector during the review period did not impact growth in the short term and led to a 0.024% increase in the long term. These findings, which are consistent with the studies of Morley et al. (1999), Staehr (2003), Falcetti et al. (2006), Djalilov and Piesse (2011) and Duval et al. (2018) on the effects of real sector reforms on growth, indicate that real sector reforms carried out for privatization, regulation of public services, supervision of labor-goods markets in terms of wages, prices, management, etc., and development of anti-poverty strategies, etc.,
The TR variable's short-run (-0.029) and long-run (-0.020) coefficients are statistically significant and nonsignificant, respectively. These findings indicate that the trade sector reforms carried out in Turkey during the review period to liberalize international trade policies resulted in a 0.029% decline in growth in the short run and were ineffective in the long run. This situation, which coincides with the findings of Correa (2002), Djalilov and Piesse (2011), and Christiansen et al. (2013), may be attributable to the relatively small number of trade sector reforms carried out in Turkey during the review period and the inability to channel growth as much as required. These growth model results demonstrate that reforms' short-term temporary and long-term permanent effects on growth vary considerably by sector.
Examining the Table 7 growth model results in terms of the inequality variable reveals that the short-run and long-run coefficients of the LEQ variable are positive and statistically significant for all models. In all models, the coefficients of the LEQ variable are calculated between 0.419 and 1.094 in the short run and between 0.439 and 0.576 in the long run. According to these findings, a 1% increase in inequality, as measured by the Net Gini coefficient, increases Turkey's growth by between 0.419 and 1.094% in the short run and between 0.439 and 0.576% in the long run. Contrary to the studies of Ostry et al. (2021), which found that inequalities reduce growth in 138 low-, middle-, and high-income countries, these results indicate that growth in Turkey during the review period was driven by the expenditures of those who received a larger share of the income distribution, rather than by all income distribution segments.
Examining the growth model results in Table 7 regarding control variables reveals that the long-run coefficients of the LGI, LRD, and LEI variables, the primary growth determinants in all models, are positive and statistically significant and are calculated by expectations. Consistent with the research of Gründler et al. (2020), Ostry et al. (2021), and Botev (2022), these findings indicate that increases in investments, education level, and R&D expenditures have long-term, positive effects on economic growth in Turkey. Despite these absolute long-run effects of LGI, LRD, and LEI variables on growth, it is known that their short-run coefficients and results on growth vary across models and dynamic lags.
The short-run − 0.045 and long-run − 0.240 period coefficients calculated for the FN variable are statistically insignificant and significant when the inequality model results in Table 8 are examined. These results indicate that the reforms carried out in Turkey's financial sector during the review period did not impact inequality in the short term but reduced it by 0.240% in the long term. These findings, which indicate that financial sector reforms intended to regulate the banking and financial sector in Turkey, etc., improve income distribution and reduce inequalities over time, are consistent with the findings of Morley (2000) and Agnello et al. (2012). In contrast, the findings contradict the findings of Furceri and Loungani (2015), Gründler et al. (2020), and Ostry et al. (2021), who contend that financial sector reforms exacerbate inequality.
Similarly, the short-run and long-run coefficients calculated for the FS variable are statistically insignificant and significant, respectively. These results indicate that the fiscal sector reforms carried out in Turkey during the review period had no impact on inequality in the short term and a 0.716% decrease in the long term. These results, which indicate that fiscal sector reforms carried out in Turkey to regulate public revenues and expenditures in a manner that can ensure fiscal discipline, improved income distribution, and reduced inequalities over the long term, are consistent with the findings of Gupta and Jalves (2022) that fiscal sector reforms reduce inequalities in both developed and developing countries. In contrast, these findings contradict Morley's (2000) conclusion that fiscal sector reforms increase inequality in Latin American countries.
The short-run and long-run coefficients of the RL variable are calculated to be 0.378 and − 2.851, respectively, and are statistically significant. These results indicate that the real sector reforms carried out in Turkey during the review period for privatization, management of public services, market supervision, etc., resulted in a short-term increase of 0.378% in inequality and a long-term decrease of 2.851%. These findings, which suggest that real sector reforms in Turkey distort income distribution in the short run but improve it in the long run, are consistent with Campagne and Poissonnier (2017), Brancaccio et al. (2018), and Botev (2022) but contradict Roeger et al. (2021) and Wiese et al. (2023).
The short-run and long-run coefficients for the TR variable, 0.150 and − 1.700, are statistically insignificant and significant, respectively. These results indicate that the trade sector reforms carried out in Turkey during the review period to liberalize international trade policies did not affect inequality in the short term and led to a 0.716% reduction in the long term. These findings are consistent with Josifidis et al.'s (2020) conclusion that trade sector reforms reduce inequality in EU member states and contradict Gründler et al.'s (2020) conclusion that trade sector reforms increase inequality in developed and developing countries. These model results indicate that reforms' short- and long-term effects on inequality vary across sectors. In this context, it was determined that the financial, fiscal, real, and trade sector reforms, which were determined to have variable effects on Turkey's income distribution in the short term, improved income distribution and decreased inequality over the long term.
Examining the results of the inequality model in Table 8 in terms of the growth variable reveals that the short- and long-run coefficients for the GY variable are positive and statistically significant, except for the financial and fiscal sector models in the short-run. In all models, the coefficients for the GY variable are calculated to fall between 0.007 and 0.018 in the short run and 0.038 and 0.219 in the long run. Inequality measured by the Net Gini coefficient increases by 0.007 to 0.018% in the short run and 0.038 to 0.219% in the long run for every 1% increase in Turkey's real GDP growth rate. These results, which are consistent with those of Josifidis et al. (2020) and Botev et al. (2022), which examined the effects of growth on inequality in European Union and OECD member countries, indicate that not all segments of society benefit equally from the increases in income level during the period under review in Turkey, or that inclusiveness is low. In contrast, these findings contradict the findings of Gründler et al. (2020) and Ostry et al. (2021), who studied the effects of growth on inequality in developed and developing countries.
Examining the results of the inequality model in Table 8 regarding control variables reveals that the direction and significance of the short- and long-run coefficients of the inequalities-determining GTO, GRD, and GEI variables vary across models. For instance, while the long-run coefficients of the GTO and GRD variables were calculated as negative and statistically significant in all models, the long-run coefficients of the GEI variable were only calculated as negative and statistically significant in the model with trade sector reforms. Similar results hold for the short-run coefficients of the GTO, GRD, and GEI variables, and it is known that the effects of trade openness, education level, and R&D expenditures on inequality vary across models. These results, which indicate that the increases in Turkey's trade volume, education level, and R&D expenditures during the review period contributed to the reduction of inequalities by improving income distribution, coincide with the studies of Gründler et al. (2020), Ostry et al. (2021), and Botev (2022) in a significant manner.
Tables 7 and 8 show that the lagged error correction term \({\text{E}\text{C}\text{T}}_{\text{t}-1}\) coefficients are negative and statistically significant for all growth and inequality models. In A-ARDL models, the effects of short-run shocks between variables will disappear in the long run, and the variables will return to equilibrium, as demonstrated by these results.