24.06.2020 - Comments

Are we there yet? Tracking Economic Recovery in Germany

All up-to-date Charts and Data are in our Real Economy Tracker

When economic and financial conditions are changing almost daily, it is not satisfying to wait months until official real economy data (such as GDP or industrial production) become available. For at least three decades, economists have worked on techniques to estimate (or nowcast) the economic activity in (almost) real time. A crucial issue has been the selection of suitable indicators with good tracking characteristics. Internet data has become a helpful source of information. Using internet data, in the following weekly Google-Trends statistics of various search categories, we calculated an index of economic activity for Germany using an approximate dynamic factor model. This index, which tracks the quarter-on-quarter percentage change in GDP, dropped from -0.7% in the last week of 2019 to -11.2% in the first week of June. A slight recovery seems to be in place by the second week of June (-8.7%).

As Figure 1 shows, the most recent official data releases for Germany show a reduction in GDP of 2.2% in the first quarter of 2020, and a slump in industrial production of 25% in April relative to January. The next official releases are due on July 14th (industrial production) and July 30th (GDP). These releases will be flash estimates, which will be revised later. In the meantime, more readily available variables can offer an approximation of the official releases from different angles. A widely used variable is electricity consumption (Figure 2) because it is a necessary input for most economic activity and it is available in almost real-time and in a high (hourly) frequency.

Electricity consumption, however, is not enough to capture the dynamics of all economic activity. Official indicators related to economic activity (such as international trade or labor markets) are released with significant lags and are not available in a daily or weekly frequency and therefore of little use for a Nowcast model. With rising importance of the internet - and with Google having become the dominant search engine (with a market share in Germany of over 90% ) - the search intensity of keywords in Google (collected in “Google Trends”) has been a useful indicator for anticipating official data releases of real economy variables.1 The intuition for applying search statistics is straight forward: An unemployed person, for example, will likely search for keywords such as “unemployment office”, “unemployment support”, “jobs”, etc. Google stores the search intensity of the relevant keywords under a specific category (for instance “Welfare & Unemployment”) and calculates an index which facilitates a comparison over time. The higher the search intensity, the higher the official unemployment rate might be.

Figure 3 shows the official monthly German unemployment rate and the weekly Google search intensity index in the category “Welfare & Unemployment”, both as percentage change relative to the previous 12 weeks. For the time after June 2016, the series move together with spikes in the Google statistic leading spikes in the unemployment rate. This is especially evident in the week of March 15, as the Google index spiked just before the unemployment rate increased in April and May. Various studies have shown that Google-Trends statistics for unemployment are a good predictor for the unemployment rate.2

To construct our economic activity index for Germany, we use various indicators for categories covering economic activities related to travel, sales, labor and mobility, together with “hard data” such as electricity consumption, the “truck toll mileage index” (which measures the traffic of large trucks in Germany), industrial production and GDP.3 To estimate the index we use an approximate dynamic factor model.4 In a nutshell, the estimation procedure assumes that each indicator can be expressed as the sum of two components: one (unobserved) factor, which is common to all the series, and a second factor unique to each time series. The unobserved common factor can be estimated using the available information and can be updated as soon as new information is available. As the common factor captures the co-movement of the indicators selected to track economic activity it can be interpreted as an index for economic activity.

The activity index for Germany is available in a weekly frequency and is presented in Figure 4. It closely follows the dynamics of the official data for GDP and industrial production growth. The index reached a trough during the first week of June and has increased since then, suggesting that a slow recovery has started. The shape and speed of the recovery will depend on multiple variables such as the development of the Covid-19 pandemic, the effectiveness of the many policy measures of the German government and the European Central bank, and the recovery of the global economy. Although recovery appears to have begun, our weekly index suggests that economic activity is still roughly 9 percentage points below its level from 3 months ago.

The predictive power of our index can only be evaluated after the final official data have been released. Meanwhile, we can shed more light on its quality by comparing it with other approaches of a similar type. The German Bundesbank, for instance, releases a weekly economic activity indicator similar to ours. It is presented, together with our index, data for GDP and industrial production in Figure 5. Both indices are highly correlated (90%) and suggest a slow recovery during the last few weeks. However, three main differences should be noted. First, the Bundesbank index captured the recovery one week later than our index. Second, the trough of our index is deeper. Third, the variance of our index is in generally higher than that of the Bundesbank.

The reasons for the differences can be threefold. First, the use of internet data. The Bundesbank relies on internet data only for three unemployment search terms. In our index we use mainly internet data and not only for certain key terms but for whole categories. The use of the faster available internet data could spot the recovery a week earlier. Second, the deeper trough and the higher volatility of our index is most likely related to the greater use of more volatile internet data. Third, different calibrations of the model influence the reading. The Federal Reserve Bank of New York, for example, calibrates its economic activity index for the US to the mean and standard deviation of US GDP growth after 2008, avoiding the higher volatility of the Great Financial Crisis. For Germany we prefer to use the mean and standard deviation for the whole series since 1970, because the current situation, on which we of course focus, is especially volatile. If we would select, for instance, the mean and standard deviation after 2010, our index would be closer to the Bundesbank’s.

There is no manual on how to track real economic activity in real-time, and there are as many estimation procedures as there are economists doing the estimations. Despite these differences, one result seems to hold across all studies: a slow recovery seems to have started in Germany. How fragile it is and how fast it will be, are further, more complex questions that will keep us busy. Only time will tell how accurate all the current efforts in tracking economic recovery are.

1 See Choi, H., & Varian, H. (2012). Predicting the present with Google Trends. Economic record, 88, 2-9.

2 See Askitas, N., & Zimmermann, K. F. (2009). Google Econometrics and Unemployment Forecasting. Applied Economics Quarterly, 55(2), 107. for a study using German data.

3 For a detailed description of the data and methodology see the technical note.

4 For a complete description of the methodology see Duarte, P., Süssmuth, B. (2018). Implementing an Approximate Dynamic Factor Model to Nowcast GDP Using Sensitivity Analysis. Journal of Business Cycle Research 14, 127–141.


Related Articles

Legal notice

The information contained and opinions expressed in this document reflect the views of the author at the time of publication and are subject to change without prior notice. Forward-looking statements reflect the judgement and future expectations of the author. The opinions and expectations found in this document may differ from estimations found in other documents of Flossbach von Storch AG. The above information is provided for informational purposes only and without any obligation, whether contractual or otherwise. This document does not constitute an offer to sell, purchase or subscribe to securities or other assets. The information and estimates contained herein do not constitute investment advice or any other form of recommendation. All information has been compiled with care. However, no guarantee is given as to the accuracy and completeness of information and no liability is accepted. Past performance is not a reliable indicator of future performance. All authorial rights and other rights, titles and claims (including copyrights, brands, patents, intellectual property rights and other rights) to, for and from all the information in this publication are subject, without restriction, to the applicable provisions and property rights of the registered owners. You do not acquire any rights to the contents. Copyright for contents created and published by Flossbach von Storch AG remains solely with Flossbach von Storch AG. Such content may not be reproduced or used in full or in part without the written approval of Flossbach von Storch AG.

Reprinting or making the content publicly available – in particular by including it in third-party websites – together with reproduction on data storage devices of any kind requires the prior written consent of Flossbach von Storch AG.

© 2020 Flossbach von Storch. All rights reserved.