Peculiarities of ARDL modeling in sociological time series analysis (the case of economic news in the dynamics of CSI in 2010–2017)

Abstract

The Consumer Sentiments Index (CSI) reflects views of the population of Russia on the economic and financial policy of the country and contributes to the understanding of recessive changes in the economy. Current methodological approach singles out inflation, exchange rate, unemployment, intensity of economic events coverage in mass media as the primary factors that guide consumers in their assessments when “rational” signals arise. The article pays attention to the peculiar features of using the ARDL approach in sociological research based on the example of assessing non-economic factors on CSI in 2010-2017, including “socially significant” factors such as mass media. In autoregressive models with distributed lag (ARDL) it is possible to use the “non-economic” indicators that are difficult to include in classical vector autoregressive models (VAR). The article shows that ARDL modeling improves the interpretation of models in the presence of mixed series, and a two-month lag in the news intensity can demonstrate a decrease in consumer sentiments. The approach used in the current study allowed to identify episodes of desynchronization of the dynamics of macro indicators since the 2010s, which, on the one hand, indicates actual changes in the corresponding indicators, and on the other hand, brings more certainty to people’s understanding of the current situation in the economy and opportunities for making large purchases. Additionally, the article examines the methodological and analytical benefits of the CSI data and describe the specifics of including various “sociological” parameters and indicators into the analysis.