Bond yields and growth

http://www.businessinsider.com/torsten-slok-us-rates-market-significantly-mispriced-2016-2

I came across this week’s graph while putting together my list of the Top 20 Best 2016 Graphs. With this graph, which appeared in February 2016, Deutsche Bank’s Chief International Economist showed a correlation between bond yields (vertical axis) and the underlying US economic growth (horizontal axis). At that time, US bond yields had dropped to 1.8%, while growth was still around 2.5%. Did the bond market sense something that was not yet reflected in the figures, or were US bond yields just overshooting? The economist concluded the latter, and that US bond yields should have actually been 2.3%.

While the graph seems pretty clear, I think it does raise a few questions. For example, the US GDP figure is a quarterly indicator, so you’d only expect to see four points added per year. However, the note says the sample period covers the last 12 months. Furthermore, we know negative growth percentages exist too, so why does the horizontal axis stop at 0%?

The answer is simple: the graph doesn’t show actual GDP data, but the Atlanta Fed’s GDP forecast, which I’ve already discussed several times before (see here and here). A couple of times a week, this indicator is updated to precisely reflect the state of the underlying US economy, with the number of observations per year increasing from four to 150. That also immediately explains why there are no negative GDP growth values: because growth was positive for the duration of 2015 (more or less the year shown).

So growth and bond yields are pretty closely correlated with each other. But is that really true? The next graph may lead you to believe there is actually a strong correlation, and that it might be a good predictor. The last five points represent the most recent data for both the bond yield (now at 2.34%, nearly the same as the figure given by Deutsche Bank at that time) and from the Atlanta Fed (which has now risen to 3.6% for the fourth quarter!).

Atlanta Fed versus 10-year Treasury yield
0dunw3aO26DggLfCLb2akxQEnjHSBE_mVkSoI1powcKgt8TAzCvseYmz0_Sa_ZqO-jcOvPeSjwZ6BIkD7clTynJ4S3kpWAn3M_96c8RFcmGDuPUmIgvzeQjqlBQi_GdpxflnfJS3aHxMc_Al2g
Source: Bloomberg, Atlanta Fed, Robeco.

It looks nice, but that’s because I did my best to make the correlation stand out on purpose by only adding the last five points, rather than all the data in between. As you may have noticed, US bond yields have risen sharply since Trump’s election, so the last few points ended up close to the original line. But if I include all the data in between, the picture looks a lot less convincing.

Atlanta Fed versus 10-year Treasury yield
Nm4aaWouVpZq8RKJHSG2xCss_Tn_Rzx3zlzgjOwVM_y30reeryZLMD1uIpAA3ymUIBfk6mK2wtSBq1nhqTjOkY8wrz4dUvaQKDRljeIGH59cyUL8E9oWV23a3n5jHTO7fIWTU7tz_xICKlAbQA
Source: Bloomberg, Atlanta Fed, Robeco.

The original data are shown in blue and the data recorded since mid-February in red. In these months, there clearly was no positive correlation (higher growth = higher bond yields). In fact, the opposite seems true. In my mind, there are three reasons why this correlation did not persist :

  1. The US economy may be the world’s largest, but that doesn’t mean its financial markets are unaffected by events happening elsewhere. This year, both the ECB and the Bank of Japan were strongly committed to lowering capital market rates further, causing them to drop to a new historic low. Because the international capital markets are interconnected, US interest rates dropped too, in spite of respectable second and third-quarter growth. Incidentally, the fact that the Atlanta Fed now registers quarterly growth of 3.6% has nothing to do with Trump’s election: that bond yields rose to 2.34%, however, does.
  2. The time span is too short. For me, anytime I see correlations drawn based on data that was collected over less than five years, it’s a red flag. I want to see what the correlation does in the long run, particularly if there is enough historical data available. There can, of course, be good reasons why the correlation was less important before a certain period, but then you’d better have a good explanation for that. If the strong correlation is based on data from the last 17 months, I don’t really buy it. In this case, not much additional historical data is available anyway, because the Atlanta Fed’s forecasts only go back to mid-2014.
  3. Nominal versus real. The Atlanta Fed is an indicator that very specifically measures real growth and therefore does not take underlying inflationary dynamics into account. Nominal growth is much more important to the bond market than real growth, as is shown in the next two graphs. These graphs show the realized year-on-year growth since 1981 (horizontal axis), versus the bond yields (vertical axis). The first graph shows the correlation with real growth and the second with nominal growth.

Real GDP growth versus 10-year Treasury yield
xsAjqW_55TDHMjT15WLwNLzq7Qru1_exJkQKEopNUh40L-lT3iS0TQa-sBUProeDafZpSZE77ILo3J9ZXV5gtrFhcn8uK_sKfee27PA0StZJNX9WxDcRbMm-8qSEVVkM6RH3L9pNpCmqroEc4Q
Source: Bloomberg, Robeco.

Nominal GDP growth versus 10-year Treasury yield
Ht33smKugLfr1Zf8T3BL2ZzAFsmQniFdbd5eAVf_By5OcwrOW9S1sWyNATgnq5iMOtXHFSeirYrXgwwvCEbDZJZj6f0ZhXumJU_w_IFYXOnVg0eelveJ04bG0HRE3VuOv5LJr4odMXBR3z84ag
Source: Bloomberg, Robeco.

Clearly, the nominal correlation is much stronger. That said, this correlation obviously does not explain everything, either. For example, in the US, first quarter nominal GDP growth was about the same in 2015 as in 1982 (both round off to 4.5%), but 1982 bond yields were 14%, as compared to the 2% we had to settle for in 2015…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s