One Answer to the Curious Case of Residual Seasonality in US Real GDP

Source: CNBC on my TV!

Recently, Steve Liesman from CNBC here in the US pointed out that something was wrong with the US real GDP statistics.  He noted, and here I attach a picture of my TV at home showing his findings, that since 1985, Q1 real GDP growth for the US has been weaker than for other quarters.  Of course this should not be the case given that statistical seasonal adjustments are supposed to take account of any seasonality in the data and automatically adjust for this.  So we clearly have a problem with the US real GDP data, but I believe that Bureau of Economic Analysis (BEA) who produce the data, are going to fix this going back a few years, for their revisions in July.

In a way, this is extremely problematic though, and it stems from the way in which the media in the US reports it's GDP figures.  In Europe and elsewhere in the world, the standard way to report economic growth is by calculating growth as % year over year change in real GDP, which automatically adjusts for any "residual seasonality" in the statistics.  But in the US, GDP figures are reported as "quarter on quarter growth expressed as an annualized rate" - which therefore does rely much more on an accurate adjustment for seasonality in the GDP figures.

Now you are probably thinking - "well who cares?"  Well unfortunately these figures are very important, not only in terms of setting the tone of the US stockmarket, but also in terms of policy measures, such as the adjustments of interest rates by the Fed!  Many of the market commentators saw the revised GDP figures last week with the "Second estimate" of Q1 GDP showing a contraction of 0.7% in (annualized growth in) real GDP as a blow to the recovery and tried to blame this on everything from the port strike on the West coast to the frigid weather in the first quarter. Even commentators said that the economy is too weak for the Fed to move in June to increase interest rates.

But I thought that for this week's blog I would take the real GDP growth figures and re-express them in terms of Year over Year growth. So that's exactly what I have done in the figure below.  This, I would argue is a much better way to judge our economic growth, and when you look at it this way, it is really not too shabby in my view.

Source: BEA.gov; Data calcs: Blog author

Now viewed in this light, a 0.7% contraction, turns into a 2.7% growth rate, which was an acceleration from Q4 of 2014.  Now if you look at the figure above, you'll see that although although consumer spending (C) is drifting in an upwards direction, it is 7.4% increase in private investment spending (I) that appears to have caused the uptick in the GDP % yoy growth data for Q1. Note also that since turning negative in 2010, government spending (G) has also moved into positive territory.

Now what of the international sector.  Well here, if you look at the data, the news isn't good whichever way you report it.  If you use the % YOY method that I use here, you will find that exports fell 22.7% YOY, and imports increased 6.5% YOY.  And in the investment category, if you take out the accumulation of inventories from the figures, investment only increased by 5.2%, which although still impressive, does suggest that business investment still needs to be boosted by consumer spending, which is still quite hesitant.

But from my own perspective, these figures bolster my view that although the Fed probably won't do a rate rise in June or July, they should.  The economy is growing as strongly as it has been at pretty much any time since 2010 when you measure economic growth in the best way possible, by using the %YOY method!  Also, while I know that the strength of the US dollar matters (more on that for another post), the main measure of robust growth in an economy is domestic spending or "absorption".  If the US Treasury and Fed have an exchange rate policy of benign neglect for the US dollar, then the movement of the US dollar should not dictate or effect the direction or timing of monetary policy.

Putting Some Perspective on the March Jobs Figures

The jobs numbers released last month were, by any measure, a tepid take on the performance of the US economy in March at 126,000, and despite the fact that the Bureau of Labor Statistics (BLS) also tempered the first pass at the spectacular job figures for February, which were revised down to a net 264,000 new jobs created, the consensus on Wall Street was that the economy was sagging.  This led to all sorts of speculation that now the Fed wouldn't raise rates until the fall or even into the back end of this year, but I think all this Fedspeak is frankly misguided.  The Fed is simply NOT going to base it's judgement on one month's numbers, and particularly a month that was, by any standards, unseasonally cold and unusual.  So to get this out of the way right at the beginning of this blog, I will state for the record that I still think the Fed will raise rates in June, and September at the very latest, for reasons I have elucidated in previous blogs postings.  My reasoning hasn't changed here, and no slightly weaker sets of data will affect the business cycle and monetary policy "normalizing" arguments.

But I digress.  What I want to address here is the BLS labor data released last month, and the reasons why I think the data will get revised upwards, and even if it doesn't, why it really doesn't merit the type of response it got from the markets.  So first, I know that most of the readers of my blog will recognize that one month's numbers do not a trend make - and this data point, in my view, was simply a blip on an irregular cyclical pattern. Now if we get April and May numbers below 100,000, then I will change my view, but one datapoint in the 100,000s range should cause panic, and particularly not at the Fed, as other data on wage increases suggest that wage increases are now accelerating which points to a tightening labor market, which is not surprising given that we're now at a 5.5% unemployment rate.

US Non-farm payrolls (sa)
2007 to date

The points I would like to make about this datapoint are as follows: when one looks at this data on a longer term basis then there does appear to be a small drop off in employment but this also happened at the beginning of 2012 or 2014, and the economy bounced back nicely from those points. But probably most importantly, the March datapoint was not an outlier, it was just a little disappointing. This is shown in the figure to the left.

US Non-farm payrolls (sa)
1939 to present

So why should it have been disappointing? I think there are clearly 2 reason - first, and probably most importantly, the layoffs that are now occurring in the energy sector given that the oil price is clearly not doing following the "v" shape that some expected.  The "j" (on it's side) shape of price trajectory that we are now seeing means that although the highly indebted small and mid-sized oil companies will have to lay off workers, the larger oil companies will be much better able to tough it out.  These oil layoffs, therefore, will be very much one shot deals, which although now occurring, will not trail beyond around 3 months in my estimation.

The other reason for such a tepid growth figure is that the north of the country once again got a really cold blast this winter, and as I was up in Boston in March (seeing Thomas Piketty speak among other things), so I can personally corroborate that fact!  In fact by some measures that part of the country got record amounts of snow this year. Now the figures are "seasonally adjusted" so do take account of the lower level of labor market activity in different seasons, but the statistical methodology can only account for the average level of seasonality - it cannot detect outliers - so these will inevitably show up in the data, which, I believe, they did.

Now the second point I would like to make is that when you observe the full sweep of non-farm payroll figures going back to the beginning of the series (see the figure above), what is clear is that although the figures in the upturn that we have seen might be slightly lower than those observed in the 1992-2000 period, they are certainly not out of line with figures from previous decades.  In fact the level of net job creation appears to be very much in line with what we observed in the 2003-2006 period and back in the 1960s as well. In other words it is somewhat of a myth that jobs have not been created at the same rate that they have been in the past, as it depends which "past" business cycle you look at!

In fact one of the biggest observations I see from this longer series is that there appears to be several highly irregular cycles at work in the data - and although obviously the negative observations occur mostly during recessions, the positive datapoints also appear to show a lot of cyclical behaviour.  That obviously calls for some analysis - and so I will definitely be using this dataset for a future academic paper.  Stay tuned folks!