ifoCAST

ifoCAST

With ifoCAST, the ifo Institute has adapted its forecast architecture to modern scientific methods. Each year at fixed points in time, the ifo Institute presents estimates and forecasts of Germany’s current gross domestic product (GDP) growth. These base on numerous economic indicators, and assess their influence on changes in the forecast. In this way, we link our monthly ifo Business Climate Index with a regularly updated assessment of current economic developments. 

Feeling the Pulse of the Economy: Backcast – Nowcast – Forecast

ifoCAST distinguishes three forecast horizons: Backcast – Nowcast – Forecast. The backcast is the estimate of German GDP growth for the respective previous calendar quarter before its official publication by the German Federal Statistical Office. The nowcast estimates the development of German GDP growth for the current quarter. The forecast is the forecast for the following quarter.

ifoCAST - GDP Forecast and Forecast Adjustment Due to New Information

ifoCAST generates forecasts of current German GDP with the help of a statistical model that works with historical correlations. These forecasts do not necessarily have to correspond to the published ifo Economic Forecasts, which also take expert knowledge into account.

Change compared to previous quarter in %, price-, seasonal-, and calendar-adjusted

Forecast uncertainty: 90% interval based on forecast distribution.
Source: ifoCAST; calculations by the ifo Institute, as of:

How to Read the Graphic?

For each forecast point in time plotted on the horizontal axis, the black point shows the respective point forecast for the rate of change in GDP compared to the previous quarter. The gray band describes the forecast uncertainty: the wider the band, the greater the forecast uncertainty. Changes to the point forecast occur between two points in time due to unexpected information not anticipated by the model, in other words, when indicators are published that do not correspond to the model forecast. The colored bars represent the proportion of the forecast change related to each indicator group (in percentage points). Their sum corresponds to the forecast adjustment. The actual realized rate of change in GDP compared to the previous quarter is marked with a triangle or diamond. The GDP flash report (triangle) is usually published by the German Federal Statistical Office 30 days after the end of the quarter (T+30). The detailed report (diamond), which may already reflect initial revisions, is sent 25 days later (T+55).

  • Methodology of ifoCAST
    • The core of ifoCAST is a dynamic factor model with mixed frequencies. The common component is extracted from a variety of economic indicators (e.g., the ifo Business Climate Index or industrial production). The model can simultaneously process monthly and quarterly indicators, as well as map the influence of new information on the change in the GDP forecast.

  • Indicators and Selection
    • Forecasts for the overall economic development should be based on a broad spectrum of indicators. For this purpose, we have compiled a large dataset that currently comprises around 300 variables from various categories (e.g., production, orders, surveys). In addition to indicators for the German economy, information describing the international economy is also included. Using a machine learning algorithm reduces the number of indicators that are included in the dynamic factor model. Currently, the forecasts with ifoCAST are based on a selection of 21 indicators, which are assigned to seven groups for simplified presentation.

      Indicators Used for the GDP Forecast with ifoCAST 

      Indicator

      Group

      Price-adjusted gross domestic product

      National accounts

      Production in manufacturing industry

      Productions

      Sales in manufacturing

      Turnover

      Sales in industry (excluding construction)

      Sales

      Sales in the energy sector

      Sales

      Sales in retail trade (without car dealership)

      Sales

      Sales in the hospitality industry

      Sales

      Sales in wholesale trade

      Sales

      Order intake in manufacturing, total

      New orders

      Order intake in manufacturing, domestic

      Orders

      ifo Business Expectations, industry and trade a

      Survey

      ifo Business Expectations, manufacturing

      Survey

      ifo Export Expectations

      Survey

      ifo order change compared to previous month, manufacturing

      Survey

      ifo Business Expectations, wholesale

      Survey

      ZEW Financial Market Indicator

      Survey

      Notified Vacancies

      Labor market

      Imports (special trade)

      International

      Exports (special trade)

      International

      Global industrial production

      International

      Global trade volume

      International

      a Manufacturing, construction, wholesale, and retail trade.

      Sources: German Federal Employment Agency; CPB Netherlands Bureau for Economic Policy Analysis; ifo Business Survey; German Federal Statistical Office; ZEW Financial Market Survey.

  • Historical Forecast Quality
    • To make valid statements about the quality of the new forecast approach, we evaluated the forecast errors over a longer period of time by means of a forecast experiment for the years 2011 to 2019. For each quarter, a situation was assumed as a forecaster would have found it at the respective forecast point in time. Thus, we use only the data available at that time, estimate the model with it, create a nowcast for the respective quarter, and then move forward successively in time.

      With very few exceptions, the nowcast of the new forecast approach is very close to the realization of GDP as published by the German Federal Statistical Office. For the period 2011 to 2019, the mean forecast error of ifoCAST is 0.03 percentage points, so that no systematic distortions can be deduced. The forecast quality, which is usually measured in terms of the root mean squared forecast error, is 0.31 percentage points for the same period. The standard deviation of quarterly German GDP growth is 0.37 percentage points and thus above the average forecast error of ifoCAST. Compared to simple alternative forecast approaches, ifoCAST provides more accurate predictions.

       

      Info graphic, Evaluation of the nowcast for the period 2011 to 2019

Basic Texts

Article in Journal
Robert Lehmann, Magnus Reif, Timo Wollmershäuser
ifo Institut, München, 2020
ifo Schnelldienst, 2020, 73, Nr. 11, 31-39
Contact
CV Foto, Robert Lehmann, ifo Institut

Dr. Robert Lehmann

Economist
Tel
+49(0)89/9224-1652
Fax
+49(0)89/985369
Mail
You Might Also Be Interested In