Industrial production is a major macroeconomic indicator due to being one of the main reasons for economic growth. Both developed and developing countries keep up with monthly industrial production index data to ensure targeted economic growth rates. Therefore, raising industrial production results in higher gross domestic product (GDP) levels. Approximating actual capacity to full capacity and enabling firms to produce at lower costs are primary macroeconomic objectives for every country. Thus, increases in capacity utilization are expected to positively affect and increases in producer prices to negatively affect industrial production. Based on this perspective, this study investigates the impacts of capacity utilization rate and producer prices on the industrial production index in Türkiye.

By reviewing the existing literature, studies on this subject are noticed to have dealt with this issue with the assumption of a linear relationship. The present study aims to observe the nonlinear impacts of capacity utilization rate and producer prices on the industrial production index. To do this, the study uses the nonlinear autoregressive distributed lag (NARDL) model, which separates the positive and negative shocks caused by the explanatory factors. The NARDL model is a modified version of the linear ARDL model developed by Peseran

and Shin (1999) and Peseran et al. (2001). This model has no requirements due to all the variables in the model being integrated in the same order; namely, the series can be I(0) (stationary at level) or I(1) (stationary at first difference). Thus, when some of the variables are stationary at first difference while others are stationary at levels one can continue with the cointegration test using the ARDL model, unlike other cointegration tests. Shin et al. (2014) went on to transform the linear ARDL model into an asymmetrical structure. The structure of the estimated NARDL model in the present study is shown in Eqs. 5 and 6. The variables for the logarithms of the industrial production index, capacity utilization rate, and producer price index are respectively abbreviated as LSAN, LKKO, and LUFE.

By utilizing the data obtained from the Central Bank of the Republic of Türkiye, the study constructed a data set covering the period of June 2007-September 2022. Due to the data being monthly, the total number of observations is 189. The empirical analysis section of the study first tests the stationarity of the variables (see Table 2). The augmented Dickey- Fuller unit root test suggests the LSAN and LUFE variables to not be stationary at level and LKKO to be stationary. On the other hand, the Dickey-Fuller generalized least squares (GLS) unit root test suggests the LUFE variable to not be stationary at level and the LSAN and LKKO variables to be stationary. Both tests also show all variables to be stationary at their first difference. Due to the variables being unsuitable for integrating in the same order, ARDL-based models should be used to fulfill the needs.

The NARDL cointegration test results in Table 3 show the F-statistics to exceed the critical values, which implies the presence of a cointegrating relationship among industrial production, capacity utilization rate, and producer prices. The next step estimates the longterm coefficients of the explanatory factors and applies some diagnostic tests. The results in Table 4suggest a 1% increase in the capacity utilization rate to raise industrial production by 1.63% and a 1% decrease in the same factor to reduce industrial production by 1.27%. On the other hand, a 1% increase in producer prices reduces industrial production by 0.06% while a 1% decrease in the same factor has no statistically significant impact. The fact that positive and negative shocks do not affect at the same rate implies that the shocks from capacity utilization rate and producer prices on industrial production have an asymmetric structure. The diagnostic tests regarding the estimated model show no serial correlation, heteroskedasticity, or misspecification problem to be present in the model and the residuals to be normally distributed. The cumulative sum of the recursive residuals (CUSUM) and the cumulative sum of the squares of the recursive residuals (CUSUMSQ) plot tests are depicted in Figure 1 and Figure 2 and show the estimated coefficients to be stable and within the critical bounds of the model. Lastly, to check the robustness of the asymmetry, the previously estimated model was transformed into a stepwise regression model to test the symmetry hypothesis based on the coefficients of this model. The test results reveal both capacity utilization rate and producer prices to have asymmetric impacts on the industrial production index in Türkiye.

The findings from the study briefly suggest that in order to sustain economic growth in Türkiye, policymakers should first raise the capacity utilization rate and prevent any potential decrease in this indicator. Secondly, they should prevent increases in producer prices. However, the findings revealed no statistically significant favorable impact from decreases in producer prices. The ability of the NARDL model to permit asymmetric effects allows for differentiating policy designs against negative and positive shocks.