A Solution to Fisher's Lament
On November 2nd, Dallas Fed President Fisher lamented that monetary policy mistakes might have been made in the past on the basis of erroneous first-published economic data. And today President Fisher might be feeling a little frustrated as the BLS announced that it had undercounted August and September nonfarm payrolls by a net 139,000. Last month, the BLS announced that it might have undercounted nonfarm payrolls through March of this year by a whopping 810,000. (By the way, even adding 810,000 to the count, nonfarm payroll growth in this cycle still was the slowest of any cycle since 1961.)
Of course, if the Fed would stop trying to macro-manage the economy and instead focus on simply increasing its balance sheet at a constant rate of growth equal to that of the population (see "Inflation Targeting - Is Bernanke Really An Expert On The Great Depression?"), it could ignore economic data - erroneous or accurate - resting assured that it was not contributing to credit-induced inflationary conditions in the economy. But I am not that naïve to expect this prescription for monetary policy will ever be filled in my lifetime.
Let me offer an alternative prescription to a Fed bent on macro-managing: Look at data series that never get revised or that have a lower probability of major revisions. Now, just because a series never gets revised doesn't necessarily imply that it sends a strong message about the performance of the economy. But if such a series does regularly send an accurate signal, it might be worthy of the Fed's consideration and, even more importantly, investors' consideration. Two series that come to mind are the ISM manufacturing index for New Orders and initial jobless claims.
The underlying ISM manufacturing data never get revised. The only revisions to the data are updated monthly seasonal adjustment factors. If we restrict ourselves to the seasonally-unadjusted data, then we don't have to worry about the seasonal revisions that take place annually. Chart 1 shows the behavior of the four-quarter moving average of the unadjusted ISM manufacturing index for New Orders along with the year-over-year percent changes in quarterly observations of real GDP. The shaded areas represent periods of recession. I have cut off the period of observation at the end of 2002 on the assumption that most of the revisions to real GDP growth through the end of 2002 already have occurred. The highest correlation between the two series, 0.80, occurs contemporaneously suggesting that ISM manufacturing new orders is a good, although not perfect, coincident indicator of overall economic activity. Chart 2 shows monthly observations of this index - both seasonally adjusted and unadjusted - from January 2003 through September 2006. Notice that both the seasonally-adjusted and unadjusted New Orders indices are now trending lower, suggesting that the pace of economic activity is indeed moderating. Notice also that there is not much difference between the monthly behaviors of each index, as exemplified by the high correlation coefficient of 0.86.
Now, let's consider unadjusted initial jobless claims. Why initial jobless claims? Because jobless claims are not based on a sample that is massaged with assumptions about how many new small businesses came into existence in a given month and how many folded. Rather initial jobless claims are a "universe" not a sample - no assumptions about anything. Moreover, the initially unemployed have an economic incentive to "get in line" and be counted. As far as I know, the BLS does not legally prosecute or monetarily penalize businesses who send in inaccurate staff counts each month. And let's not even get into the "knock-knock" household survey of employment.
Chart 3 shows quarterly observations of year-over-year percent changes in unadjusted initial jobless claims and year-over-year percent changes in unadjusted nonfarm payrolls from Q1:1968 through Q4: 2002. Again, I am assuming that most of the revisions to nonfarm payrolls through the end of 2002 already have occurred. The highest absolute-value correlation between the two series, 0.72, occurs when year-over-year growth in nonfarm payrolls is lagged by two quarters, suggesting that initial jobless claims are a leading indicator of nonfarm payrolls and economic activity as a whole. (That's probably the reason the Conference Board placed initial jobless claims in its index of Leading Economic Indicators rather than its index of Coincident Economic Indicators, which includes nonfarm payrolls.) The sign on the correlation is negative as one would expect. Rising initial claims for unemployment insurance would logically be associated with falling employment.
Let's take a look at the weekly year-over-year percent change in initial jobless claims starting in the first week of 2003 (see Chart 4). In the week ended July 3, 2004, initial jobless claims were down 27.5% vs. year ago. Irregularly, the year-over-year decline in weekly jobless claims became diminished starting in the second half of 2004 through the early summer of 2005. After Hurricane Katrina hit the U.S. in late August, jobless claims temporarily skyrocketed. This event admittedly distorted the cyclical economic signal of initial jobless claims, sending the year-over-year percent change temporarily higher for a few weeks in the late summer of 2005 and temporarily lower for a few weeks in the late summer of 2006. But notice that in the latest week, October 28, 2006, after most of the Katrina distortion is over, initial jobless claims are now 2.1% above their year-ago level. Similar to the recent behavior of the ISM manufacturing New Orders index, the behavior of initial jobless claims suggests that the pace of economic activity is weakening.
In sum, if the Fed must meddle, perhaps it should be guided by more reliable coincident and leading indicators of economic activity such as the ISM manufacturing New Orders index and unadjusted initial jobless claims than unreliable first-published (or even second- and third-published) random numbers such as monthly nonfarm payrolls. Again, even if the Fed is guided by these unreliable economic reports, there is no reason why investors need to be.