Tracking Economic Activity in Real Time

By Julius Probst

The Corona pandemic and the current recession it has resulted in have made it abundantly clear that macroeconomists need to look beyond conventional measures in order to assess the economy in real time.

Last year, I attended a conference at the Danish Central Bank on the use of Big Data to assess the state of the economy. While there is always the worry that more data will simply increase the noise without necessarily providing a better signal, there is little doubt that advances in computing power and better statistical analysis/methods should help us in analyzing bigger and richer data sets. We at Macrobond certainly feel that we are part of the solution, with what can sound like an overwhelming amount of data (170 million time series, though well structured and searchable) available in our platform, and an integrated smart analysis and charting tool that makes that data manageable.   

Another promising aspect of current macroeconomic analysis is the increased use of high-frequency data, which we have also emphasized in recent months (see here and here).

In this post, I want to look at some macro series, including Nowcast estimates and other high-frequency datasets, to see how they have performed during the current economic shock.

The chart below shows the well-known Sahm rule according to which changes in the unemployment rate indicate an upcoming economic downturn.

Macrobond Moment: You can add recession bands in a Macrobond chart by right-clicking on the chart and selecting “Fill range”.

A recession starts when the 3-month moving average of the unemployment rate increases by 0.5 percentage points or more, relative to its low during the previous 12 months.

Historically, the rule has worked well and could have indicated to the Fed already by the beginning of 2008 that a recession was coming. However, with the current Corona shock, the Sahm rule moved too slowly, since a 10% percentage point increase in the unemployment rate happened during the month of April alone. Therefore, looking at higher frequency indicators, like weekly job claims, is  a better way of measuring the economy, at least in the current environment, than using more conventional macroeconomic indicators that have significant time lags, such as inflation and GDP estimates.

This was also a problem in 2008 when the Fed did not act upon the economic downturn until the fall when it increasingly became clear that the economy was already entering a massive downward spiral. And the start of the Great Recession was dated later on to December 2007 according to NBER.

Precisely for that reason, macroeconomists have started to construct Nowcast models, which estimate the economy in “real time”. They are commonly based on dynamic factor models. More specifically, these models use several high-frequency data series related to economic growth as inputs and then extract a factor variable, which serves as an estimate for GDP growth.

The graph below shows the Nowcast estimates from the New York Fed model and the Atlanta Fed model. The NY model also includes the forecast for Q1, which came out at -0.3% and thus significantly higher than the actual GDP value of almost –5% for Q1 (annualized). For Q2, they currently estimate a decline of 35 to 50% of GDP.

However, the Nowcast estimates themselves rely on somewhat more slowly moving series, such as industrial production and the employment report. That is why these models only started to predict such a significant decline for Q2 GDP by the middle of May when it had been already abundantly clear that Q2 would be an economic disaster after the lockdown.

Other leading indicators, such as soft data based on surveys, already indicated a severe economic depression before the Nowcast flashed up. The big shock moment came in February when the Chinese PMI dropped by more than 20 points – I have to say it, an unprecedented decline. And of course, other advanced economies followed suit in the next months as the Corona virus spread to the rest of the world.

Note that the recent recovery in the PMI does not actually indicate that a V-shaped economic recovery will take place. In fact, the index has to rise substantially above 50 for some of the output loss to be recouped by the end of the year. 

Business surveys were also starting to deteriorate by March. However, given the unprecedented shock and speed of the Corona shock, they were not very useful as a leading economic indicator. Moreover, while soft economic data can be a useful guide, ultimately hard data determines economic outcomes in the end.

Market monetarists, for example, are focusing on financial data, including prediction markets and asset markets, to estimate the current stance of monetary policy and economic conditions. As one can see below, the Corona shock immediately translated into much tighter financial conditions as asset prices fell across the board, volatility increased, and therefore also financial stress. And as the chart below shows, the correlation between volatility and financial stress is extremely high.

However, Central Banks have acted faster and with much larger interventions than in 2008. For that reason, financial conditions and especially stock prices have decoupled somewhat from the underlying economic data, which is still quite miserable across the globe.

The new Weekly Economic Indicator by the New York Fed, which we also added recently, is another indicator constructed from component analysis for 10 high frequency time series based on real economic activity, such as consumer confidence, retail sales, and steel and electricity production. The underlying principal component then provides an estimate for real GDP growth.

Using our Principal Component Analysis (PCA) Tool in Macrobond, it is possible to create your own Nowcast model. Inspired by the Weekly Economic Indicator, my colleague Wadsworth has done just that for Canada, using Canadian high-frequency data.

The first principal component of the high frequency data (below) is fed into a regression model that automatically lags PC1 to the point of highest correlation with Real GDP growth (YoY%).  The resulting output is a weekly tracker of the “real economy” that forecasts YoY% GDP growth for Canada.

Macrobond Moment: This document was created using our PCA analysis tool. Using our formula language, we have also created a best lag correlation. The R2 value is dynamic and automatically updated with incoming data.

Given that we are currently adding a lot of high frequency indicators, like Corona virus stats, the Google mobility tracker, etc., if you’re a Macrobond user, you can use our PCA tool to fairly easily build your own Nowcast model that could be more reliable than the more slow-moving Atlanta and New York Fed Nowcasts.

Good luck! 

Disclaimer: We don’t usually have views and opinions about economic and financial states of affairs, (not ones that we express publicly as a company, anyway). We do believe, however, that people can and do appreciate a variety of perspectives. What you’ve just read is the perspective of the author. While we think our writers are very smart, Macrobond Financial does not expressly endorse the views presented here. And, as the old adage goes, you shouldn’t believe everything you read (not without finding the data, performing a few analyses and presenting it in a nice chart). We want to make it clear that we are not offering this information as investment advice. That being said, if you have Macrobond, you can easily check everything that’s mentioned here, and decide for yourself. If you don’t have Macrobond, now you have a great reason to get it.
Scroll to top
Thank you! Your subscription has been confirmed. You'll hear from us soon.
Subscribe to our latest posts