Endless worries about a eurozone disintegration and potential growth slowdowns across the developing world, but at least there’s been a streak of surprisingly not-terrible economic indicators in the US heading into the new year.
Or: feels a lot like December 2010, doesn’t it?
Unfortunately, there are some methodological reasons for why the positive indicators and the sense of déjà vu should make you sceptical.
That chart is of Nomura’s economic surprise index for the US, and it shows how macroeconomic indicators in the last two years have surprised in both directions relative to consensus forecasts.
The similar path of the index in 2010 and 2011 is immediately obvious and has an explanation — one that should dampen at least some of the optimism around the uptick in these indicators recently.
In two notes, the first from late October and the second from last week, Nomura explains how the severe contraction in the US economy at the end of 2008 and early 2009 was captured by some economic indicators as new seasonal trends (emphasis ours):
The standard empirical techniques used to seasonally adjust US economic data – such as the Census Bureau’s X11 and X12 programs – have interpreted some of the sharp contraction in the fourth quarter of 2008 and the first quarter of 2009 as a change in “seasonal” patterns. As a result, current techniques for seasonal adjustment tend to boost data in the fourth and first quarter of the year, relative to previous patterns, then depress data in the second and third quarters.
That bias has influenced recent data, including last week’s Philly Fed and Empire State regional manufacturing indexes. But it affects a wide range of indicators.
The standard narrative of the last year was that the US economy was showing signs of revival in late 2010, only to be derailed by the increasingly dramatic sovereign debt problems in Europe, the Japan earthquake and tsunami, and the spike in oil and commodity prices. Since the autumn it’s shown signs of an accelerating recovery, if not an overwhelmingly impressive one.
That’s not wrong, but Nomura’s analysis shows that problems in the techniques used to account for seasonal trends have contributed to an exaggerated sense of how well we were doing at this time last year and, relatedly, to the forecast errors on the part of macroeconomists who got a little carried away.
Could be happening again:
The apparent “bias” in seasonal adjustment also helps to explain the pattern of forecasting errors over the last two years. Figure 5 shows the average market “surprises,” weighted by their standard deviation, for five-month indicators — Philadelphia Business Outlook Survey, the NY FRB Empire State Survey, the Chicago PMI, and the ISM manufacturing and non-manufacturing survey – as well as the estimated seasonal “bias” for these series. The correlation between forecast errors and the estimated seasonal “bias” since the beginning of 2010 is 0.66.
Analyzing other, more complicated data, is more difficult. For example, important aggregates, such as retail sales data available from the Census Bureau, are adjusted at a highly disaggregated level with great attention paid to very specific calendar and trading day effects. Moreover the Bureau of Economic Analysis has stopped publishing the unadjusted data for the underlying components of GDP.
More complicated data also may have been affected by this problem. For example, the average seasonal adjustment factors for retail sales excluding autos, a key input for GDP, have been revised in ways that may have been influenced by the recent recession. These changes imply that the current seasonal adjustment factors may tend to overstate the growth nominal retail sales in the fourth quarter by about 2 percent, at an annual rate, and understate growth in the second quarter by about 1-1/2 percent on the same basis. These differences are large enough to have a notable impact on our assessment of economic trends.
And if you want one bit of compelling evidence to bolster Nomura’s case, here’s a chart comparing the performance of the ISM manufacturing production index against the Fed’s measure of industrial production. Unlike the other indicators Nomura mentions, the Fed measure was adjusted earlier this year to account for the earlier flaws in its seasonal adjustment techniques:
They’re headed in different directions, and the Fed’s reading for November disappointed and cut against the trends evidenced in other indicators when it came out last week. Now we have a reason why.
Up next, writes Nomura, you can expect exaggeratedly strong readings from the Chicago PMI later this month and the next ISM manufacturing survey at the start of January.
We draw two, very simple conclusions from these notes.
One is just that methodological issues, a favourite topic of ours, matter an awful lot and continue to be under-appreciated. A push for better data is always welcome, and we can’t think of a good reason for the Bureau of Economic Analysis, for instance, not to start publishing the unadjusted data mentioned above.
But the second and more important point is that this is another reason to temper any burgeoning excitement over the seemingly favourable trends in the US economy over the last few months (in addition to the obvious ones: Europe, political stasis, etc..).
Nomura does mention one silver lining from this research, which is that it means the recovery has perhaps been smoother and less “stop-go” than we previously thought. Not to be completely dismissive, but “smoothe” has nothing to do with “robust” — and the problem with the plodding recovery is precisely that it’s been plodding, not that it’s been unsteady.
Optimists take note, and you’ll find the Nomura research in the usual place.
Related link:
The case for GDI: Q&A with Jeremy Nalewaik – FT Alphaville