Forecasting the NAPM Purchasing Manager's Index
The Purchasing Managers' Index, published monthly by the National Association of Purchasing Management (NAPM), is a highly recognized indicator of the health of the United States economy. Its information is so highly prized that many desire to get the information early. This article looks at several different indexes published by NAPM each month to help in making estimates of future values of the Purchasing Managers' Index. The results indicate that New Orders and a new index based on New Orders and Vendor Deliveries are both consistent leading indicators of the Purchasing Managers' Index by one month.
The Purchasing Managers' Index, published monthly by the National Association of Purchasing Management (NAPM), is a highly recognized indicator of the health of the U.S. economy. A recent editorial in Electronic Buyers' News (EBN) highlighted the importance of this indicator. EBN commented on a Wall Street Journal article that encouraged NAPM to maintain better control of the dissemination of its report. A supposed leak led to a bond price surge on a Friday afternoon when the sharp decline in the Purchasing Managers' Index was interpreted as a sign of slower economic activity. EBN also reported that some periodicals, such as Business Week, have begun publishing estimates of the index. For example, the June 12, 1989, issue of Business Week predicted a May index of 54.5 percent, while the actual figure turned out to be 49.7 percent. Since then, NAPM has altered its procedures and now releases the survey on the first business day of the month following the survey month.
Previous studies have verified the validity of the NAPM survey relative to the national economy. In one recent study, Hoagland and Taylor proposed two new indexes from the data gathered by NAPM. The first, called the Business Barometer, is an unweighted average of the seasonally adjusted data for the production, new orders, employment, purchases, and vendor performance change indexes. A change index records the percentage of respondents answering "higher" to a question minus the percentage answering "lower." With zero as the point of no change, a positive sign indicates an increasing level of activity, while a negative sign indicates a decreasing level of activity. The benefits of the Business Barometer are that (1) it is easier to calculate, since it requires no weighting, and (2) it is less erratic than the Purchasing Managers' Index.
The second index proposed by Hoagland and Taylor is the Forecasting Index for predicting business cycle turning points. The Forecasting Index is calculated by subtracting the seasonally adjusted Inventory change index from the seasonally adjusted New Order change index. When the Forecasting Index is negative, a business decline is likely to develop; when it is positive, a business expansion is likely. Thus, the proposed Forecasting Index should be useful in forecasting the future values of either Hoagland and Taylor's Business Barometer or the Purchasing Managers' Index. The rationale of this index is that a New Orders increase, production in later periods will need to increase to meet these orders, and if inventories are decreasing, production will need to continue or increase to replace the shrinking inventories. Thus, if New Orders are up and Inventories are down, a higher index value is produced than is the case when New Orders are up and Inventories are up.
This article investigates the performance of the Forecasting Index as a forecasting tool for both national and regional data. Alternative indexes based on the research also are proposed for forecasting the performance of the Purchasing Managers' Index.
The purpose of this research is to evaluate the effectiveness of potential leading indicators of the Purchasing Managers' Index, including the Hoagland and Taylor Forecasting Index (HTFI). Since the Purchasing Managers' Index is considered a coincident indicator of the economy, and is more reliable and available earlier than other indicators, accurate forecasts of the Purchasing Managers' Index will produce a good indication of the future course of the economy.
The Purchasing Managers' Index (PMI) is a composite index based on the seasonally adjusted diffusion indexes for new orders, production, vendor deliveries, inventories, and employment, with varying weights applied. A diffusion index consists of the percentage of positive responses plus one-half of the neutral responses. Diffusion indexes are considered to be leading indicators that show both direction and magnitude of change. Diffusion indexes and change indices are directly related: The conversion formula is:
Diffusion Index = 50% + 0.5(Change Index)
The PMI data used in this study came from NAPM's monthly issues of the Report on Business from January 1982 through June 1989. The 90 months of data were divided into an estimation sample of 66 months (January 1982 through June 1987) and a forecasting sample of 24 months (July 1987 through June 1989). Since the NAPM New Orders and Inventory values are not seasonally adjusted, a change index for each was calculated and then seasonally adjusted using the adjustments reported by Hoagland and Taylor. Each index was smoothed using a five-period centered, weighted moving average methodology. The purpose of the smoothing adjustment was to reduce the effects of monthly variations and to present the basic cyclical component of each index. Figure 1 shows the raw PMI and the smoothed PMI values.
Linear regression was used to determine relationships between the variables of interest and the Purchasing Managers' Index.
The first step in this study was to compare the performance of the Purchasing Managers' Index with a published government index. The index chosen was the Series 920 Composite Index of 4 Roughly Coincident Indicators. Since the Coincident Indicator is an activity index and the PMI is a diffusion index, the Coincident Indicator was differenced (by subtracting the previous period's activity level from the current period's activity level) and multiplied by ten to compensate for scale differences. Figure 2 shows the smoothed PMI plotted against the smoothed Coincident Indicator series beginning in January 1982. Comparing the Coincident Indicator series with the PMI produced a correlation coefficient of 0.91. Thus, with this high level of correlation, the Purchasing Managers' Index can be considered a coincident indicator of the economy. The advantage of using the PMI is that it is available two to three months earlier than the government's Coincident Indicator series.
When the PMI was compared with potential leading indicators of its performance, some interesting results were obtained. Figure 3 shows the smoothed Hoagland and Taylor Forecasting Index (HTFI) plotted against the smoothed PMI. Major turns in the PMI are evident in December 1983, May 1985, and October 1987. The HTFI led the turns in 1983 and 1987 by three months. However, the 1985 slowdown in the PMI was greater than the HTFI indicated, and it took place eleven months after the HTFI indicated it should occur. When a regression analysis was computed using the Hoagland and Taylor Forecasting Index to estimate the PMI, the resulting [R.sup.2] was 0.535, which was significant at the 99.9 percent level. The average forecast error was -1.034, with a standard deviation of 1.173. The best [R.sup.2] - the closest relationship - was obtained using the HTFI from seven periods prior to the period desired to be estimated.
Although the HTFI clearly led the PMI, its high variability in lead times jeopardizes its reliability as a forecasting tool. In studying the performance of the HTFI, the key question obviously is, What happened during the 1985-1986 time period that caused the index to underestimate the effect of the slowdown in the economy? In response, one observation is that the use of inventory changes seems to have little value in forecasting work. The NAPM Inventory change index has been negative since late 1979, except for a 10- to 11- month period in 1983-1984.
The only other index reported by NAPM that indicates activity external to the firm, other than new orders, is Vendor Deliveries. Figure 4 compares this index with the PMI; a slowdown in the index is seen during the 1985-1986 timeframe, which coincides with the slowdown in New Orders. Hence, a proposed new forecasting index could be based on New Orders plus Vendor Deliveries (NOVD). Since the Vendor Deliveries index is an inverted index, a positive value indicates slowing deliveries, which generally indicate an expanding economy. Figure 5 shows the new forecasting index, NOVD, plotted against the PMI. The NOVD Index led the 1983 and 1987 turns by one period and was coincident with the 1985 turn. A regression analysis using NOVD, with a one period lead time, yielded an [R.sup.2] of 0.935 and a mean forecast error of 0.519 with a standard deviation of 1.273.
Since the most significant factor in determining future activity is New Orders, a comparison was made using only New Orders to forecast the PMI. Figure 6 depicts the New Orders data plotted against the PMI. At all three turns, New Orders led the PMI by one month. A Regression analysis produced an [R.sup.2] of 0.958, with a mean forecast error of -1.034 and standard deviation of 1.173.
All three of the forecasting indices consistently crossed over the no change line (50%) ahead of the PMI during upswings but lagged crossing over on slowdowns.
The value of these forecasting equations can be illustrated most effectively by analyzing a fairly recent occurrence. In Business Week's June 12, 1989 issue, the publisher predicted that the May Purchasing Managers' Index would be 54.5 percent. Based on the equations developed above, and shown in Table I, estimates could have been made by April 10 that the May index would be between 48.4 percent and 50.6 percent, depending on the leading indicator used. The closest forecast was produced by the NOVD Index at 49.9, compared with the actual figure of 49.7
The Forecast Index as proposed by Hoagland and Taylor, though it does give more advanced warning, is not a particularly good estimator of the next period's PMI. A redefined forecast index, NOVD, produces more accurate results than the original HTFI. However, the best results come from simply using the current period's New Orders data to estimate the next month's PMI. New Orders is a consistently reliable indicator one or two periods in advance of an actual change in the direction of the economy.
Table : Table I
REGRESSION RESULTS FOR PMI
Indicator A B Lead [R.sup.2] F HTFI 41.515 0.5630 7 months 0.535 65.5 NOVD 48.414 0.2500 1 0.935 911.9 New Orders 47.796 0.4295 1 0.958 1434.5
REFERENCES [1.] Paul Hyman, "Talk on the Street: NAPM Report," Electronic Buyers'
News, June 26, 1989, p. 14 [2.] National Association of Purchasing Management, NAPM Insight,
August 1989. [3.] John H. Hoagland and Barbara E. Taylor, "Purchasing Business Surveys:
Uses and Improvements," Freedom of Choice: Presentations
from the 72nd Annual International Purchasing Conference, Oradell,
N.J.: National Association of Purchasing Management, 1987, p. 167;
Theodore S. Torda, "The NAPM Business Survey: A Prophet for
Profit," 70th Annual International Purchasing Conference, May 1985. [4.] Hoagland and Taylor, op cit., 1987. [5.] Ibid. [6.] Ibid. [7.] U.S. Department of Commerce, Bureau of Economic Analysis, Business
Conditions Digest, February 1989.
Alan Raedels, C.P.M., is an Associate Professor at Portland State University. He holds a Ph.D. degree from Purdue University. Dr. Raedels has taught a wide variety of courses and seminars in the field of purchasing and materials management and has published numerous articles in various management journals. Since 1984, he has been Chairman of the Business Survey Committee for the Purchasing Management Association of Oregon.
PHOTO : Figure 1 PURCHASING MANAGERS' INDEX
PHOTO : Figure 2 PURCHASING MANAGERS' INDEX VS. COINCIDENT INDEX
PHOTO : Figure 3 PURCHASING MANAGERS' INDEX VS. HTFI
PHOTO : Figure 4 PURCHASING MANAGERS' INDEX VS. VENDOR DELIVERIES
PHOTO : Figure 5 PURCHASING MANAGERS' INDEX VS. NOVD INDEX
PHOTO : Figure 6 PURCHASING MANAGERS' INDEX VS. NEW ORDERS…