Forecasting private investment in Finland using Q-theory and frequency decomposition

We look for a forecasting model for private investments in Finland. As explanatory variables, we use different proxies of Tobin's Q and cash flow as well as these series decomposed to different frequency components. The forecasts are produced using OLS and National Accounts and Financial Accounts data. We find that the models that include a proxy of Q, cash flow or both most often result in smaller RMSFE's compared to the benchmark AR-model, but the differences are not statistically significant. Frequency decomposition can improve the model performance, even though the difference is not statistically significant either. The most accurate forecasts are achieved by approximating Tobin's Q with the ratio of firm's market value and the sum of physical and intangible capital stocks.