Nowcasting growth to enrich the factors used for government bond selection

Our multi-factor bond strategy uses well-known factors like value, momentum, low risk and quality to select government bonds and credits with superior risk-adjusted returns. The portfolio construction algorithm aims to efficiently harvest these factor premia while keeping the top-down risk profile in line with the index and ensuring a strong sustainability profile.

We now add a quality measure for the selection of government bonds. This measure uses nowcasting to capture timely signals from the plethora of macroeconomic data that are continuously released.¹

To select high-quality government bonds we thus want to compare the state of the economy across countries. There are some challenges to gaining timely data on the state of the economy, though.

The official data series for economic growth is GDP, a comprehensive measure of economic activity. The drawback of this series is that, for most markets, it is updated only quarterly, with a publication lag of at least a month. Moreover, the data is prone to substantial revisions. This means that a bond selection process in early July would either use first-quarter data, showing which economies were growing strongly four to six months previously, or would need to be delayed by a few weeks until the publication of the first estimates of second-quarter growth – and maybe even longer as these numbers might be revised later on.

Fortunately, many macroeconomic data series are released more frequently and with a shorter publication lag than GDP. However, these indicators typically relate only to specific parts of the economy. To get the full picture of the overall economy, it is necessary to combine several data series.

For this we use nowcasts, daily estimates of current growth based on a wide variety of data series. In early July, many data series give information about May, and some already for June; survey data can shed some light on what could be expected in July. This is an advantage compared to the traditional GDP numbers: by combining data series that are updated faster and more frequently, one can create a comprehensive measure of activity that leads the latest published GDP data by up to several months.

Nowcasting can thus be used to get timely information on the state of the economy. But is this enough to select government bonds with superior risk-adjusted returns? To answer this question, we performed historical backtests. Ideally, we would only use nowcasts that had actually been published at that point in time, but these have a limited history. We therefore extend our dataset by using nowcasts that are carefully reconstructed. As this process requires a large variety of economic indicators, with their exact historical publication dates, there are limits to how far back in time one can go, and the backtest is therefore fairly short.

Figure 2 shows the cumulative performance of a long-short portfolio based on the nowcast. This portfolio has long positions in bonds from countries where growth is moderating and short positions in bonds from countries where the economy is heating up. The portfolio is rebalanced monthly.

We find that the nowcasts generates attractive risk-adjusted returns. Furthermore, this performance is largely uncorrelated with the factors we already use for government bond selection. It thus adds value to our multi-factor government bond selection. Out of prudence, we add this new quality measure with a somewhat lower weight than other measures, to reflect the shorter history over which we can assess its performance. For other factors we have much longer backtests, extending beyond 200 years. For deep-sample evidence on the existence of value, momentum and low-risk factors, please refer to Factor Investing in Sovereign Bond Markets: Deep Sample Evidence.