Forecasting Recessions in Canada: An Autoregressive Probit Model Approach

We forecast recessions in Canada using an autoregressive (AR) probit model. In this model, the presence of the lagged latent variable, which captures the autocorrelation in the recession binary variable, results in an intractable likelihood with a high dimensional integral. Therefore, we employ composite likelihood methods to facilitate the estimation of this complex model, and we provide their asymptotic results. We perform a variable selection procedure on a large variety of Canadian and foreign macro-financial variables by using the area under the receiver operating characteristic curve (AUROC) as the performance criterion. Our findings suggest that the AR model meaningfully improves the ability to forecast Canadian recessions, relative to a variety of probit models proposed in the Canadian literature. These results are robust to changes in the performance criteria or the sample considered. Our findings also highlight the short-term predictive power of US economic activity and suggest that financial indicators are reliable predictors of Canadian recessions.