| |
 May 2009 Ahead of the Market? Assessing the Value of the ADP Employment Report
To be sure, only the strongest form of the EMH holds that all information is always reflected in the market. In its weaker forms, there's the possibility that inside information -- or even soon-to-be public information if it's important enough -- can be of value, leaving some hope for the diligent active investor. Everybody eventually gets the same news, but if you get it first or can correctly read the tea leaves to determine what it will be before it's officially announced, you'll have a leg up on the rest of the market. That's why investors are always looking for special information or insights that can give them a timely edge. The irony is, these anomalies only work when few investors are aware of them. When they become too widely known, everyone (i.e. the market) acts on them so no one has an advantage. Oddly enough, there are some broadly followed indicators that have the potential to move the market on a regular basis. This doesn't seem possible, even under the weakest form of the EMH.
Monthly Mover
Anxious investors looking for this edge can take their cue from ADP's National Employment Report which comes out one or two days before the Bureau of Labor Statistics issues the "official" report. It's based on a survey of ADP's business outsourcing clients from across the country and is often seen as a preview of the "official" BLS report. But ADP's report is no secret; thousands if not millions of investors look forward to it and the press gives it extensive coverage. Nevertheless, it moves the market when it differs from the consensus estimate. When that happens, the "official" report that follows is often a non-event. How can this be? If everyone knows about it and looks to it for guidance, why isn't it already reflected in market prices? Why isn't the BLS's report a non-event every month? Probably because there have been months when the ADP report has sent a signal that hasn't been confirmed the BLS. In those instances, the market initially reacted one way on the former, only to reverse course on the latter. To a quant, this raises the question of the extent of the relation between the two reports. Arguably, it's this less than perfect relation that prevents this anomaly going despite its broad acceptance.
Quantifying the Relation Starting with January 2001, we paired the results from the ADP report with those from the BLS. The results are shown on the nearby chart where we've plotted the ADP figures on the horizontal axis and the BLS's on the vertical axis. If the series were identical (i.e., always had the same value) they'd line up in a straight line with a 45-degree slope. That's not the case here, but there is a definite linear pattern. The green line on the chart represents what's known as the "least-squares fit line" or "line of regression". Of all the possible straight lines that can be drawn, it's the one that comes the closest to touching all of the circles. Notice that it's pretty close to that 45-degreee slope. The actual regression equation is shown in the top left quadrant of the graph where y represents the BLS value, x is the ADP value, and the final figure (1.6972) is a constant. All of the values in the equation as well as on the chart are in thousands. Notice that the coefficient (0.9974) which multiplies the ADP value is very close to 1. If the two series always gave identical results, it would be 1.
R-squared can vary from 0-1 and measures the strength of the relation between the two series. At 0.9013, it's relatively high. All of this would make you believe this is a true anomaly; the ADP estimate is an accurate representation of the official numbers a few days hence. Makes you wonder why everyone doesn't act on it rather than why so many do. But those full period numbers don't tell the whole story. This is not unusual in statistical studies and why it's important to look at more than just one time frame. In this case, we looked at 3-year time periods with all (except the last which ended on February 28, 2009) ending on December 31. We calculated the correlation for each of these periods and ended up with the data illustrated in Chart 2. This time the green line shows the trend over the period. As you'd expect from the total period findings, it shows a strengthening relation. However, the individual periods tell a different story. They range between .971 (February 2009) to .579 (2006). Rather than rising at a steady pace as suggested by the trend line, the line itself is overly influenced by the relatively high beginning value (2003) and ending value (2009). In the meantime, the correlation has shown a high degree of fluctuation. Although correlations are never static, even between the most highly correlated data sets, the strongest relations don't show nearly this much variance. Typically higher fluctuations over various time periods indicate that the regression equation is somewhat spurious in that there are features of the dependent series (the BLS statistics in this case) that are not captured by the independent series (the ADP numbers). At times the equation works well (as in 2008 and 2009) but at times it doesn't (2006 and 2007). Essentially the missing factor(s) are not as important in the former, but are in the latter. The trend line is misleading because it's simply averaging over the entire time frame. Here's the problem for anyone hoping to rely on this relation: You never know when those missing factors will come into play. You never know if you're in the midst of a period of high correlation or entering a low one. You might be able to look back over the past several months and compare the results between the two to get a sense, but that's the best you'll ever be able to do. True anomalies must be more reliable than this.
Reading the Tea Leaves And it's not that one employment report is more "correct" than the other. Both reports use sampling techniques to gauge the number of jobs created or lost in the prior month. Neither is capable of looking at each and every corner of the country to count each specific job. Because of this, the two are like thermometers placed in different rooms in a house. Often they'll measure similar results, but at times (such as when the sun strikes one room and not the other), they can have significantly different readings. Neither is more correct than the other. Investors are always looking for that silver bullet to lead to certain gains. Presumably, that's why the EMH holds that all available information is already reflected in market prices. When one event appears to consistently (even over a short period) foreshadow another, it's only natural to view the two as related. Recently that's been the case with the ADP and BLS employment reports. It would be wonderful if their recent relation continued to grow and become more reliable, but as you can see form Chart 2, that hasn't been the case in the past. In the 3-year period ending December 2003 the correlation was only slightly weaker than it is now, but three years later it was at its lowest point. There's nothing to prevent that from following again. In fact, there's probably more reason to believe the correlation will be lower in the next three years if for no other reason than the fact that it can't get much higher. If the correlation does begin to decline -- and we think it will -- the ADP report will slowly lose its market-moving ability. By the time the media stops following it each month, the correlation may be set to rise again. Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||
ROPONENTS OF THE EFFICIENT Market Hypothesis (EMH) will tell you that all of the relevant data is always reflected in the market. In other words, you, as a single investor, can't "beat the market" with your special knowledge, data, or research because no matter what it is, it's already incorporated in the prices of the relevant securities. Pretty depressing, huh?
