Dow Jones Industrial Average temporarily peaked on September 20, 2006 and dropped near 80 points on September 21. In the mean while the yield of 10 year US treasury dropped to 4.65% from the previous day's yield of 4.73%. Those moves were attributed to the disappointing results of business outlook survey of September, announced by The Federal Reserve Bank of Philadelphia on the morning of September 21, 2006, apparently fanning the recession fear anew. The indices of Philadelphia Fed are deemed to have only secondary importance by many economists since it only reflects the business conditions and expectations of a regional area. The sharp reactions of financial and equity markets to the regional indices on September 21, 2006 thus have surprised many observers. We have undertaken a detailed study of the indices of Philadelphia Fed and the results are presented here.
The Federal Reserve Bank of Philadelphia is one of 12 regional federal reserve banks, overseeing eastern Pennsylvania (about two-thirds of the state), Delaware and Southern New Jersey. It is a traditional industrial area with agriculture and traditional commerce mixed in. The indices are constructed based on the survey of area businesses. Two indices of the survey are the focal points here, the current index that indicates the current economic condition of the area businesses, and the future index that measures the expected condition 6 months in the future. The indices come out early; for example, September indices are published on September 21, about 10 days earlier than national purchasing manager survey. Though the Philadelphia Fed's survey only covers one region, many market participants try to extrapolate the indices to a national reading by assuming that the whole country in average is just like that region. The monthly data from the survey are very jumpy. To smooth out the readings, we have taken three-month moving averages of the current index and the future index respectively, and have plotted them in the graph at the right. The graph covers the period from Jan. 1991 to Sept. 2006, with the current index plotted in blue dots and the future index plotted in red dots. The data are obtained from the website of The Federal Reserve Bank of Philadelphia.
Apparently we need some nation wide economic statistics to be compared with Philadelphia Fed's indices. We choose real (means inflation adjusted) personal consumption expenditure for that purpose. Personal consumption expenditure comprises about 75% of GDP, and the data are published monthly. Annualized monthly % changes of real personal consumption expenditure are calculated. The monthly % changes of real personal consumption expenditure are also very jumpy, so 3 month moving averages of the monthly % changes are calculated and are plotted as purple dots at the bottom portion of the graph at the right. The personal consumption expenditure data are taken from the website of The Bureau of Economic Analysis. As can be seen from the graph, 3-month moving averages of monthly % changes of real personal consumption expenditure are still very jumpy and are not suited for extracting macroscopic views about the condition of nation-wide consumption if used alone. We have further computed 12-month moving averages of monthly % changes of real personal consumption expenditure and plotted them as green dots in the bottom portion of the graph. It should be noted that the smoothing out of data by taking multi-month moving averages is done with the sacrifice of timeliness. For example, the data point of 3-month moving average of September of 2006 is measuring the average condition of July, August and September as a whole, not the condition of September alone. However, if the data are jumpy, looking at September data alone will not allow us to derive any meaningful conclusion about the current condition of the economy, so it becomes better to use multi-month moving average even at the expense of the loss of some timeliness.
Let us first look at the green dots, 12-month moving average of monthly % change of real personal consumption expenditure. During 1991, it shows clearly the effect of 1990 to 1991 recession. It took from the end of 1991 to near the end of 1992 for the green curve to recover to a plateau, roughly 3.5% a year, and stayed there until 1995. The growth rate of average real personal consumption expenditure dipped slightly through the years 1995, 1996 and the early part of 1997. It was from the middle of 1997 the growth rate shot up toward the 2-year long plateau of about 5%, the heyday of the late 1990 boom. It is interesting to observe that both the 3-month and the 12-month moving averages of monthly % change of real consumer consumption expenditure peaked and turned down around the beginning of 2000, correctly predicting the onset of the most recent recession. This down turn in the growth rate of consumer consumption was about 3 month earlier than the burst of the stock market bubble. The low growth rate of consumer spending persisted for nearly 4 years, and only recovered to 4% range starting from 2004. The 12-month moving average, the green dots, peaked in the summer of 2005, and has turned down slightly since then.
The purple dots, representing 3-month moving averages of monthly % change of real personal consumption expenditure, though are volatile, yield useful information when combined with 12-month moving averages. The purple dots tend to be more volatile during the transition periods, they are, the recovery from the recession of 1991 to 1992, the recovery of late 2002 to 2003, and the period of sinking into the recession of 2000. The huge ups and downs of the purple dots of late 2001 was due to 911 attack, but the ripple effect subsidized rather quickly as shown in the green dots. The effect of hurricane Katrina in the fall of 2005 caused a large dip in the purple curve. It also forced the green curve to oscillate down to certain degree. It looks like that the oscillations due to the hurricane is masking a genuine slow down of the growth rate of real personal consumption expenditure to about annual rate of 3%. Real personal consumption expenditure fell 1.5% (annualized) in August of 2006 from the previous month. If it continues to experience negative growth rates in coming months, both the purple curve and the green curve will experience significant down turns and may indicate a coming recession. However, if the monthly rate recovers, even just to a few % points in coming months, the purple curve will simply go through another zigzag motion, leaving the green curve steady around 2.5% to 3.0% range. The latter situation is widely called the "soft landing" scenario in the circles of financial and equity markets. From the data of real personal consumption expenditure alone, we do not know which scenario will unfold since the data is designed to tell us the past condition, not as a forecasting gauge. That is why we are interested in the gauges like Philadelphia Fed's future index that may serve as a crystal ball to gaze into the future.
The blue dots, 3-month moving average of Philadelphia Fed's current index, correctly reflect the 1990 - 1991 recession and 2000 -2002 down turn. However, the detailed ups and downs of the blue curve do not match the peaks and valleys of the purple curve, the 3-month moving average of monthly growth rate of real personal spending. Especially notable is the slight downward trend of Philadelphia Fed's current index from 1997 to the end of 1999 that is the golden age of the late 1990 boom and is properly reflected in the strong growth of real personal spending. Philadelphia Fed's current index rose steeply from early 2001, substantially ahead of the recovery of the growth rate of real personal spending in 2003, but the current index peaked and started to decline at the middle of 2004, whereas the real personal spending kept the 4.0% growth rate (in average) until the middle of 2005. When real personal spending has started to fall since the middle of 2005, the current index of Philadelphia Fed has started to level off instead.
The indices of Philadelphia Fed are constructed from the survey of nine components. They are "New Orders", "Shipments", "Unfilled Orders", "Delivery Time", "Inventories", "Prices paid", "Prices Received", and "Employment". Area businesses are asked about the current conditions of those nine components, and the result is used to construct the "current" index. They are also asked about the perspectives of the same nine components 6 months in the future to construct the "future" index. If the future index, the red dots, really forecasts the future activity of the current index, the peaks and valleys in the red curve should precede the corresponding peaks and valleys of the blue curve by a few months. There are some notable successes of the red curve in forecasting the coming behavior of the blue curve. For example, the red curve correctly predicted the turn around of the blue curve in 1995 with an advance notice of about 2 months, and the forecast about the turn around in 2001 with an advance notice about 2 months. The red curve peaked in the fall of 2003, preceding the peak of the blue curve in 2004 by nearly 10 months. However, there are also quite a few failures: The peak of 1991 in the red curve came in one month later than the corresponding peak of the blue curve, and the rounded behavior of the peak in 1994 was also behind the drop of the blue curve. In 2005, the bottom of the blue curve came in one month earlier than that of the red curve; since that time the blue curve stagnated whereas the red curve jumped up and then has dropped. Another notable feature of the indices is the gap between two curves. The red curve tends to jump way above the blue one at the early stage of the recovery, and then the gaps fade away not long after as the actual recovery in the personal consumption expenditure takes hold. Also should not be overlooked is the fact that the future index also trended down in average slowly through the boom years of late 1990.
The seemingly complicated behavior of Philadelphia Fed indices, which often contradict with the personal spending data, can be explained readily by the trade deficit and the movement of US Dollar. The area covered by The Federal Reserve Bank of Philadelphia has many manufacturers that are constantly under the pressure of imports. When imports increase and consumer spending booms, those manufacturers suffer mightily and pulls down the indices. Also when Dollar becomes strong against the currencies of the trading partners, the area manufacturers become pessimistic and the future index will fall quickly, whereas as discussed in Article 1, 2 and 2A, booming imports and the run away trade deficit are the base of booming US personal spending. That is why the indices of Philadelphia Fed are often go opposite to the real personal consumption expenditure. The down turn of the future index in the fall of 2003 coincided with the start of the gigantic dollar buying operation of Japanese government to boost US Dollar and to expand US trade deficit.
From this study comparing Philadelphia Fed indices with the monthly change of real personal consumption expenditure, it should be clear that Philadelphia Fed indices cannot be used as a forecasting gauge for the national economy. Equity markets has apparently noticed their error on Sept. 21, 2006, and the markets have rallied since. The yields of long-term treasuries are still at the level of Sept. 21. However, the low long-term treasury yields are the continuation of the phenomenon called "the conundrum of interest rates" and have nothing to do with the indices of Philadelphia Fed. At the time when foreign governments are the major buyers of US treasuries, they have strong incentives to hold down long-term interest rates in US so that US consumers can continue the habit of borrowing and buying the goods of those foreign countries. It should be reminded that those foreign governments are not private entities with profit motives, and those governments have the will and the ability to buy into long-term treasuries when private institutes are scared into buying safe short-term treasuries. Thus the phenomena of the persistent negative yield curve should not be taken as the sign of an oncoming recession, but as an assurance that foreign governments like China and Japan are doing their work to sustain the borrow and spend boom in US, circumventing the power of The Federal Reserve Bank to regulate US economy.