Predicting the contribution of artificial intelligence to unemployment rates: an artificial neural network approach

As technological innovations gain the capacity to replace human labour, it is increasingly possible that artificial intelligence can lead to higher unemployment rates. This paper is devoted to forecasting unemployment that is based on artificial intelligence as an input of interest by using an artificial neural network learning process. The simulation is performed based on a sample including 23 of the most high-tech and developed economies, over the period from 1998 to 2016. The proposed artificial neural network with one layer and 10 neurons offers good results in terms of unemployment prediction, with an overall coefficient of determination of 0.912. Artificial intelligence input is a top contributor to the prediction of unemployment, along with foreign direct investment, total population, labour productivity, and lagged unemployment. Inflation and government size register a modest contribution. This suggests that forecasts that include this new variable will be more accurate.