Forecasting unemployment with internet search data: Does it help to improve predictions when job destruction is skyrocketing?

As more and more daily activities take place online, data on internet behaviour is becoming a key information source. In this sense, several papers have explored the usefulness of internet search data in order to improve the nowcasting and forecasting of economic indicators. Special attention has been paid to predicting unemployment. Nonetheless, most of the empirical evidence on this field has focused in countries with low/moderate rates of unemployment. This paper follows this line of research and explores the use of Google Trends data in order to improve the forecasting of the figures of unemployment in Spain. This country reveals as a very interesting case due to the sharp increases in unemployment caused by the economic crisis. With the aim of providing an extensive approach on the Spanish unemployment, we use ARIMA models, also introducing as explanatory variables the Google searches for “job offers” and a business confidence indicator referred to employment perspectives. In this way we combine the time series perspective with qualitative indicators from the supply and the demand sides, leading to a more comprehensive description and also improving forecasting performance. [ABSTRACT FROM AUTHOR]