GrowthNow

GrowthNow refers to the practice of using current or recent economic data to estimate the GDP growth rate for the current quarter before official figures are published. This provides a more up-to-date understanding of economic conditions and trends, as there usually is a significant time lag between official GDP data releases.

We aim to equip policy-makers and senior executives across Canada with the essential insights they need to make well-informed decisions in a rapidly changing economic environment. This monthly report enables policymakers to respond promptly and implement measures effectively, thereby bolstering the efficacy of economic policies to foster sustainable economic growth and prosperity.

We utilize a mixed data sampling (MIDAS) regression model – introduced by Ghysels, Santa-Clara, and Valkanov (2004) – intended to forecast Canadian Real GDP growth. This approach enables us to model low-frequency variables as a function of higher-frequency variables and their lagged terms. It also offers a more straightforward approach to nowcasting by streamlining specification and theory-based constraints, as opposed to state-space models.

We have chosen to use monthly frequency variables that include GDP at basic prices, retail sales, manufacturing sales, inventories, employment, and commodity prices. We only incorporate one quarterly variable: GDP at market prices, lagged by one quarter.

Once we retrieve quarterly GDP data for the previous quarter, we begin nowcasting GDP growth for the current quarter. We will update our nowcast estimates every time there is a major data release for the variables included in our model. We use Statistics Canada’s advanced estimates for GDP at basic prices and retail sales to give us more information about the current quarter.

Whenever we update the model to account for new data, some high-frequency variables and their lagged terms may be removed from the model, depending on which combination of variables produces the lowest Akaike information criteria.

All nominal variables are adjusted for inflation before we calculate the one-period percentage change. We use growth rates for all variables and their lagged terms as inputs. Our approach does not provide a breakdown of the components of GDP; rather, the model produces an estimate for GDP growth (at market prices) for the current period.

The nowcast results will be published monthly.

The results of our nowcast model for 2023Q1 are displayed in this section.

Our model produced an initial estimate of 0.4 per cent on February 28, 2023. As seen in Chart 1, our nowcast trended upward between April and May, eventually settling at 1.08 per cent on May 20. The official GDP figures were released on May 31, with quarterly GDP growth coming in at around 0.8 per cent.

Sources: The Conference Board of Canada, Statistics Canada.

Despite the discrepancy between our nowcasts for 2023Q1 and the actual figure, our regression is relatively accurate. For the final nowcast of 2023Q1, the root mean squared error of the regression in Chart 2 was 0.2 per cent. This means the average distance between our estimates and the actual values is 0.2 per cent.

Sources: The Conference Board of Canada, Statistics Canada.

We have also back-tested our model to assess its robustness and check how it would have performed in previous quarters. In the last four quarters, the average forecast error was 0.2 per cent.