Table 1

Changes in Not Seasonally Adjusted Money Stock-M (per cent change, not at an annual rate)

seasonally adjusted change is even more sharply focused. From March to April of this year, not seasonally adjusted M1 increased 3.2 per cent. This is .4 of a percentage point faster growth than in the same period of 1976 (nearly 5 per cent faster at an annual rate) and 1.2 per cent (over 14 per cent at an annual rate) faster than the average change for the ten-year period 1966 to 1975. The case for a much greater than seasonal spurt in M1 growth in April is even stronger if comparisons are made only with the first half of the ten-year period.

Table

On the basis of the evidence summarized above, it is clear that, based on either seasonally adjusted or not seasonally adjusted data, there was a real spurt in M1 growth in April of this year. However, since there was also a spurt in M1 in April 1976, does this mean that the seasonal factors for this period are changing? Some shift might be in process, but evidence is far from conclusive, as indicated in Table 2. shows monthly per cent changes in M1, not seasonally adjusted for the period 1965 to date. Instances when M1 growth was strong relative to the average change--roughly .3 per cent, or 3 per cent at an annual rate, greater than the average monthly change--two years in a row are blocked in heavy dark lines. As the table shows, there are eleven periods when not seasonally adjusted M1 growth bulged in the same month two years in a row relative to its average change for that month. In each instance, My growth in the years following the bulge fell back to near its average rate of change for the month. The only exception is the month of September, which shows a change in seasonal

Table 2

pattern beginning in 1965. Thus, there is little in the past pattern of M1 growth that says rapid M1 growth in a particular month two years in succession indicates a changing seasonal pattern. Indeed, the evidence that is available leads to the opposite conclusion.

Even if future revisions of seasonal adjustment factors should moderate the April M1 growth the strength will not disappear, but only be distributed to other months of the year. The underlying strength of M1 in recent months is indicated by growth for the first half of 1977 at about a 6 per cent annual rate and for the second quarter of nearly 8 per cent at an annual rate. If past revisions are any guide, these figures are not likely to change significantly.

The CHAIRMAN. If the gentleman would yield for a moment?
Mr. BLANCHARD. Yes.

The CHAIRMAN. I want to join Mr. Blanchard in that request. I think it would be very helpful if the Federal Reserve would give us as much as you can on that.

I also will ask unanimous consent to put into the record at this point a paper prepared by Dr. Robert D. Auerbach of our staff in collaboration with Dr. Jack L. Rutner of the Federal Reserve Bank of Kansas City on seasonal adjustment distortions.

[The paper referred to follows:]

SEASONAL ADJUSTMENT DISTORTIONS CAUSED BY THE X-11

PROGRAM

(By Robert D. Auerbach and Jack L. Rutner)1

(1) INTRODUCTION

Our intent here is to examine some of the purposes for which seasonal adjustment procedures are employed, to describe a widely-used technique for seasonally adjusting data, and then present informally some of our own findings about defects in the procedure. We then suggest precautions that should be taken in viewing seasonally adjusted data and suggest alternate seasonal adjustment procedures.

(2) NEED FOR SEASONAL ADJUSTMENT IN TESTS OF ASSOCIATION

Suppose the correspondence between changes in the money supply and subsequent changes in the price level were being investigated with, for example, monthly figures, from 1950 to 1976. Employing either the money supply series or the price series in raw form without any prior adjustment has at least two drawbacks. First, each series would generally be rising over time while, second, each series would have periodic movements around their trends, with most of those within 1 year periods called "seasonals". It is doubtful, however, that most investigators would interpret the common trend in the two series to mean that they are necessarily related in a meaningful way inasmuch as most economic time series have been rising since World War II. Thus, the first type of adjustment that would be necessary on the two series would be some form of "trend removal". Although it is not the subject of our study, the need for trend removal gives some insight into the reason for seasonal adjustment.

Most economic time series are systematically affected throughout the year by factors which themselves follow a regular seasonal pattern such as weather patterns, formal schooling, agricultural planting and harvesting, and holiday seasons. Most careful investigators will not necessarily infer a causal relationship between two variables just because they are both affected in a similar manner by these seasonal factors.

What is of use, however, in discovering the relationships between economic series is the knowledge that one series has increased more or less than it normally does in a particular season as a result of a non-seasonal movement in another series. Removing the ordinary recurring seasonal movements in a series allows one to make this kind of determination.

Additionally, the presence of seasonals in two series may make it appear as if both series are related due to common movements when they may not be. Alternatively, if one series has a seasonal while a second does not, it may appear as if both are unrelated when they indeed are. Thus, removing a common seasonal has an advantage, because what remains is two series that are not biased either for or against finding a meaningful cause and effect relationship between them.

1 The coauthors are as follows: Dr. Auerbach is economic advisor to the Committee on Banking, Finance and Urban Affairs, and professor of Economics at the American University. Dr. Rutner is a financial economist at the Federal Reserve Bank of Kansas City. The views expressed in this summary are not necessarily those of the Federal Reserve Bank of Kansas City or of the Federal Reserve System. A technical version of these findings will appear in the coauthors' forthcoming article "The Misspecification of a Nonseasonal Cycle as a Seasonal by the X-11 Seasonal Adjustment Program" in the Review of Economics of Statics (Harvard University Press).