The purpose of nowcasting is to produce forecasts of the present and the near future. Nowcasting models supply information on the current state of the economy for quick use. The models can significantly shorten the release delays of the statistics describing the economic situation.

Statistics Finland started publishing nowcasting models by introducing quicker experimental turnover estimates for main industries in February 2022.

Unlike the longer-delay publications of the past, the quicker turnover estimates are based solely on a sales inquiry sample of 2,000 large enterprises, for which data are collected each month. This shortens the release delay to about 20 days from the end of the statistical reference month (T+20). The delay is thus ten days shorter than in the previously used experimental turnover estimates for main industries or in the currently used turnover estimates (T+30) published 30 days from the end of the statistical reference month.

The nowcasting models for the latest turnover estimate month for main industries are based on both time series models and machine-learning methods making quick production of information possible.

The key idea behind the methods is to make efficient use of many variables by relying on dimension reduction. We also use guided regularisation machine-learning methods for the effective estimation of the forecasting models. Read more about the methods used in the process in the methodological report on creating turnover estimates for main industries.

The first of the quicker turnover estimates for main industries was published about six months ago for the main industries of industry, construction, trade and services. In this blog, their annual changes are compared with the corresponding calculations of official turnover statistics, which are published 45 days from the end of the statistical reference month (T+45).

In addition to turnover estimates, official turnover statistics rely on the Incomes Register of the Finnish Tax Administration and data on value added tax payments and employer contributions of self-assessed taxes.

Price rises have helped many companies to boost their turnover this year

The exceptional year has been reflected in positive annual changes in each main industry. Enterprises have recovered from the emergency caused by Covid-19, and price rises have also helped enterprises to increase their turnover compared with 2021. (Figure 1)

In total industry, the quick T+20 forecasts produced with the experimental turnover estimate have resulted in annual changes that are very similar to T+45 annual changes. The most significant differences were noted in the most recent publication (for June). (Figure 2)

In construction, there are major differences in annual changes between T+20 and T+45 release delays. 

In wholesale and trade, there are only minor annual differences between T+20 and T+45 release delays, except for the most recent month.

In the turnover of the other services activities total, there are differences between release delays in the most recent three months. (Figure 5)

Especially in services and construction, differences in business cycles between small and large enterprises may be reflected in the uncertainty of the most recent month. The sales inquiry covers the largest enterprises of the industries, even though a substantial number of small enterprises operate in construction and services. However, at the same time, the purpose of the forecasting methods is to model trends in the enterprise group as a whole, including the enterprises left out of the sales inquiry and of the most recent month of the calculations.

The period reviewed in this blog is short, but the preliminary results are promising. Moreover, models become more capable over time, and their performance is expected to improve. The usability of new machine-learning methods will also be studied, and work will be carried out to make them more suitable for the purpose. Time will tell how turnover estimates for main industries will evolve.

The author works as Senior Statistician in the Services and industry group of Statistics Finland.

Source:

Bańbura, M., Giannone, D., Modugno, M., & Reichlin, L. (2013). Now-casting and the real-time data flow. Handbook of Economic Forecasting, 2(Part A), 195–237.