Construction of an Economic Activity Indicator for Turkey

ABSTRACT

In this paper, a monthly economic activity indicator is constructed for the Turkish economy for the period of 1988–2020. A dynamic factor modeling framework is utilized in the process. The variables are first categorized into five types as activity (hard data), activity (survey-based data or soft data), trade, employment, and financial variables. After determining the candidate variables for each category, data selection is finalized by using the hard-thresholding method. Results indicate that our monthly economic activity indicator is successful in detecting the past recessionary periods of the Turkish economy and providing timelier information about the course of economic activity.

KEYWORDS:

• Economic activity
• dynamic factor model
• hard-thresholding
• real-time analysis

JEL:

• C22
• E32
• E37

Notes

1. Here, $x_t$ represents the vector of observable monthly variables. For quarterly stock and flow variables, $s_t$ and $f_t$ should be considered, respectively.

2. Seasonally adjusted data are taken from their official sources and CBRT database.

3. For the data sources see Çelgin and Akbostancı (2021).

4. Bai and Ng (2008) use three threshold values for the significance level, which are 0.10, 0.05, and 0.01, in selecting variables by the hard-thresholding method. Since 0.05 is typically preferred, we select it as the threshold value.

5. For both of the variable selection exercises, t-statistics are listed in tables and the selected variables are written in bold and italic font.

6. An important question to consider is whether the selected variables’ relationship with GDP persists when crisis periods are excluded, such as 2001 and 2009. To answer this, a variable selection exercise is repeated by taking values of the variables for 2006–2019, excluding the 2001 crisis from the sample, and for 2010–2019, excluding the 2009 crisis from the sample. It is concluded that the same variables are selected for the estimation for these periods.

7. We also estimated the indicator with the same nine variables excluding GDP. The performances of the indicator with and without GDP are indiscernible; thus, those results are not reported here.

8. For a detailed comparison of the indicator developed in this paper with those of Aruoba and Sarikaya (2013) and Çakmaklı and Altuğ (2014), see Çelgin and Akbostancı (2021).

9. The actual contraction in the second quarter of 2020 was realized as 10.3%.

10. To select the autoregressive lags and lags of the economic activity indicator, we run our model up to the fifth lagged values of both the dependent variable (annual growth rate of GDP) and the indicator for 1989Q2-2019Q4. We then select the model with the lowest BIC value. We also consider the significance of the variables.