Printer Friendly

NABE presidential address: business economists, forecasting, and markets.

Business economists are again concerned

about the current status and the future of

the profession. A review of the past

forecasting record provides mixed results: GNP

and inflation forecasts better than "no

change," interest rate forecasts less

successful and corporate earnings estimates

consistently too high. But business

economists also possess a variety of skills useful

in nonforecasting activities: assesment of

markets and policy issues, as well as

obtaining data and analyzing it. The paper

concludes with a number of suggestions for

dealing with the business economists' image


ALTHOUGH OTHER OBSERVERS might reach a different conclusion, it is my belief that business economists are plagued with a new round of anxiety about the standing of our profession and, to be frank about it, new concerns about job security. It is fair to note, of course, that several other professions have suffered through their own traumas in recent years as well. I wish to explore the validity of these self-doubts, in particular the part that emanates from our forecasting prowess and performance.(1) Then I will offer both a few suggestions about how we can deal with the image problem that remains and also a few suggestions about how we should conduct our affairs.


Before we delve into the forecasting issues, however, we should emphasize that there is good news inherent in some of our industry facts: Some evidence suggests that the profession is doing better than the undercurrents of doubt reveal.

For one thing, membership in NABE has been on an upswing. After rapid growth in the 1960s and early 1970s, regular membership (i.e., nonstudent) reached a peak in 1983 of 3,491 and then declined for four years to an interim low of 3,059. Membership has since recovered to 3305 at the end of 1990 and has held at about that level in 1991.

In addition, the interest of the media in analysis of economic and business affairs has never been greater. Our profession is constantly sought out to provide insight into major international, national, and industry developments.


I will not attempt a thorough evaluation of our forecasting performance; that task has been undertaken by others.(2) I will draw upon their work, update the record to recent years, and add a few observations of my own. I will draw primarily upon consensus forecasts -- the mean or median of a group of forecasters -- because my focus is on the performance of the profession. The examples will have to be selected from variables that are primarily macroeconomic because these are the only ones for which a fairly long and consistent record of forecasts is available.

GNP and Inflation Forecasts. Let's begin with what are probably the most visible forecasts, at least from the viewpoint of the general public -- GNP (adjusted for price change) and inflation (here as measured by the Consumer Price Index).

A charge often made is that economists have had major difficulty with their predictions near turning points, i.e., forecasting the onset of recession (often failing to predict its arrival) and the ensuing recovery (typically underestimating its strength). A defense that can be offered, of course, is that the average of cycles in the postwar period shows fifty months of expansion and eleven months of recession, so more than 80 percent of the time has been spent in the expansion phase. If a forecaster warns of recession too frequently, the evidence against this view will be known quite soon. (Paradoxically, right now users of forecasts seem more willing to excuse an erroneous forecast of an imminent recession than one that mistakenly calls for continued advance in economic activity.)

But because forecasts are prepared and used on a regular basis, turning points are only part of the story. The usual forecast questions are: How much real growth (or decline) is the economy going to experience? How much inflation (or deflation) are we going to have? Historical records of consensus forecasts are available for several groups, and for current purposes we will look at two of them: our own NABE forecast (prepared for release at our annual meeting, based upon July/August forecasts) and Bob Eggert's Blue Chip (BC) forecast (taken from the October edition of his monthly survey). The difference in timing between the two surveys is one to three months, depending upon the forecaster. This might seem to be a small matter, but if economic conditions are changing rapidly, the extra information supplied by several additional observations on key indicators can be important.

For the annual average change in real GNP, the following table provides summary statistics on the errors of the two panels:

Table 1

Error Statistics for Real GNP Forecasts

Made for the Year Ahead for 1977-90
Statistic NABE Blue Chip
Average absolute error 1.1% 0.9%
Root mean squared error 1.8% 1.5%
Theil's U Statistic 0.5 0.4%

The average absolute errors are near 1 percent and look reasonably favorable,(4) given the observed range in the actual result (i.e., -- 2.6 percent to + 6.8 percent). Relative to the average absolute change of 3.2 percent, these results may not seem impressive -- a 34.3 percent error for NABE and 28.1 percent for BC -- but there must be some recognition of volatility. Some analysts prefer the root mean squared error (RMSE) because the squaring operation places a heavier weight on the large errors. The accuracy is somewhat worse on this statistic, as illustrated in the table.(5)

Another way to grapple with the issue is to compare these forecasts with other methods. Others have undertaken this task in a much more systematic fashion than I intend to,(6) but one error statistic permits us to do so directly in a simple way. The Theil U Statistic(7) compares the root mean squared error of a forecasted change with the naive alternative of a "no change" forecast; when U equals O the forecasts have been perfect, and when U equals 1 the RMSE of the forecasted change is the same as the RMSE of assuming no change. The BC and NABE panel forecasts of GNP show U Statistics of 0.4 or 0.5 -- hardly perfect estimates, but distinctly better than assuming there would be no change in real GNP each year.(8)

Table 2 provides accuracy statistics for predictions of the annual average percentage change in the CPI.

Table 2

Error Statistics for CPI Forecasts

Made for the Year Ahead for 1980-90
Statistic NABE Blue Chip
Average absolute error 1.5% 1.3%
Root mean squared error 1.9% 1.8%
Theil's U Statistic 0.3 0.3

The errors made in forecasting the change in the CPI are larger than those for GNP, as measured by the average absolute error and by the root mean squared error. Once again, we should note the range over which the actual results varied from 1980-90 in order to obtain an impression of the difficulty forecasters were facing: The average rise was 5.5 percent, but the high was 13.5 percent in 1980 and the low was 1.9 percent in 1986, a range of 11.6 percentage points, which constitutes enormous volatility. Errors in the 1.5-2.0 percent zone are not great, but the path that was traced out by the CPI was a very treacherous one. The Theil U Statistic seems encouraging at readings of 0.3 for the two consensus estimates. Remember that 0.0 would be a perfect forecast, and 1.0 is equivalent to the RMSE of always forecasting no change. Clearly, the panels do much better than a "no change" forecast.

It may be argued that "no change" is the wrong naive model for a variable that virtually always rises, such as the CPI. Instead, if we use "same change as last year" as a baseline for comparison, the average absolute error of this naive procedure is 1.8 percent and the RMSE is 2.1 percent; thus the panels are still more accurate.

Improvement or Deterioration? Are we getting better or worse results from our forecasting efforts on GNP and inflation? The answer seems to depend upon the particular time period when the analysis is performed. The typical article in the press on forecasting accuracy is prepared after a big miss, such as occurred in 1981-82; the charge in these instances is that forecasts are getting worse.(10) But other evidence implies that the true situation is different. One example is the tabulation by Bob Eggert(11) given in a presentation at the NABE Annual meeting in 1980, which was then updated in September 1982. Looking back over the period 1969-81, Eggert found average absolute errors for real GNP and inflation fairly close to the ones shown earlier for more recent years. Even more favorable results are cited by McNees,(2) where evidence of improved accuracy is presented on an international basis. On the inflation front, forecasters clearly had difficulty with the run-up in the 1970s and the rapid deceleration in the early 1980s. Since then, accuracy has improved.

On a real-time basis, my own observations about the past year would suggest that the profession did a reasonably good job of assessing how the economic landscape would look. At our 1990 annual meeting in Washington, forecasters had reduced their estimates of GNP growth markedly (even though the median did not yet show a decline), and we were warning that the probability of recession had risen sharply in the wake of the invasion of Kuwait and the surge in oil prices. As the months progressed, the NABE panel predicted that the drop in GNP would be small by historical standards and that the bottom would be reached in the second quarter. We also began to show a deceleration in inflation. The panel participants were almost unanimous (and still are) that the next expansion will be slow in comparison to prior upturns. This guidance, taking the elements collectively, has been about right -- or so it seems, based upon data that are currently available.

Interest Rate Forecasts. Perhaps the second most visible set of macroeconomic variables that economists are called upon to predict is interest rates. Indeed, for those who work on Wall Street or in banks, these forecasts often matter as much if not more than do those for GNP and inflation, because substantial portfolio gains can be realized if accurate forecasts are made and then used by managers. But the outlook for interest rates commands much broader attention, because housing affordability and construction, as well as business capital spending decisions, depend on certain interest rates. Even politicians have been known to advocate specific courses of action for monetary policy because of the importance the public is thought to attach to interest rates!

When looking at financial variables, we should recognize right away that the forecasting challenge is quite different. Financial markets are discounting mechanisms, which means that they are always reflecting new forecasts, and business economists regularly influence how these markets move. Thus, if these markets work well, it should be hard to forecast exactly what will happen over any extended period.

Annual average interest rate forecasts are collected in the NABE and BC surveys cited previously. But I have elected to use a different set of predictions that are made by a panel of forecasters every six months and published in The Wall Street Journal.(13) The participants have been asked since 1982 to predict the yields on three-month Treasury Bills and thirty-year Treasury Bonds on a specific date six months hence. The following table provides the same accuracy statistics used in evaluating the GNP and inflation forecasts:

Table 3

Error Statistics for Interest Rate Forecasts

Made for Six Months Ahead by

The Wall Street Journal Panel (1982-91H1)
 3-Month 30-Year
 Treasury Bills Treasury Bonds
Average absolute error 1.03% 1.08%
Root mean squared error 1.38% 1.34%
Theil U Statistic 0.91 1.07

The average absolute errors are in the vicinity of 1 percent for both T-bills and T-bonds, while the RMSE's are in the 1.3-1.4 percent range. There was a period of sizable misses in the early part of the 1980s, but since then the performance has been better. The U Statistic implies that the RMSE of the bill forecasts is a little better than the assumption of no change, but the bond predictions are slightly worse.(14)

How have business economists done in estimating the direction of interest rates? Here the record is not very good: Of the nineteen trial periods for which The Wall Street Journal tabulation is available, the average estimate had the direction of Treasury Bill rates right seven times, wrong eleven times, and in one case no change was predicted. On a percentage distribution, the result is 37 percent right, 58 percent wrong, and 5 percent indeterminate. For Treasury Bonds, there were five correct estimates of the direction and fourteen incorrect (a percentage split of 26 percent right, 74 percent wrong).

What accounts for this poor showing? I suspect it mainly has to do with the nature of financial markets, as mentioned earlier. At any point in time, most forecasters will have taken into account the quantifiable elements that are known or can be inferred (about domestic output, inflation, and monetary policy, as well as their foreign counterparts). Judgments will vary to some extent, but often there is reasonable agreement on the direction of these variables. Then over the forecast horizon (six months in this case), there will be surprises in the actual outcomes or new factors that emerge (omitted variables, in a modeling sense) and turn out to play a significant role. Of course, perfect models and judgments would have encompassed these surprises, but such perfection is rarely realized.

One way to counter these difficulties, at least in part, is for both forecasters and users of forecasts to be aware of this record and its probable cause. Betting against the consensus on the direction of interest rates often has proven to be rewarding. More careful thinking about what could surprise the consensus in the next six months might be fruitful. Two caveats should be noted: (1) The consensus is right sometimes -- albeit in a minority of the cases examined here; (2) if members of the panel start to become contrarians themselves, the rule will be foiled. It would even be hard to define what the consensus prediction means if each forecaster used as his modus operandi the rule of predicting the opposite direction for rates from that of the panel average.

Earnings Forecasts and Investment Performance. One final forecasting arena will be considered, in part because it provides a comparative perspective. For roughly the past eight to ten years IBES(15) has been collecting earnings estimates for the Standard and Poor's 500 group of companies on two bases: top-down estimates, which are prepared for the index as a whole by economists/strategists at brokerage firms; and bottom-up estimates, which are prepared by security analysts for the individual companies they follow. The former are based upon macroeconomic considerations (e.g., GNP growth, inflation, productivity, foreign economic prospects), while the latter reflect analysis of individual company revenue and cost prospects within the general economic environment.

These estimates have had a consistent tendency to be too high; I guess optimism reigns on Wall Street. But there is also a significant difference between the performance of the two kinds of predictions. (An estimate is collected each month, but we used the convention of selecting the figure projected in October for the following year.) In seven of the eight years, the top-down forecast has a smaller error than did the bottom-up. A more systematic evaluation of the accuracy can be made from the summary statistics provided in the following table:

Table 4

Error Statistics for S&P 500

Earnings per Share for One Year Ahead, 1983-90
 Bottom-Up Top-Down
Statistic (Analysts) (Strategists/Economists)
Average absolute error $4.39 $2.70
Root mean squared error $5.00 $3.20
Theil's U Statistic 1.77 1.13

The average absolute error and the RMSE are substantially lower for the top-down forecasts than for the bottom-up predictions. The U Statistic, however, suggests that no one has very much to boast about here, because the results of both groups are above 1.0; of course, the extent of the deviation from 1.0 is much greater with the bottom-up figures.

These results illustrate how difficult it is to forecast earnings. For the top-down forecasters, fairly small changes in real volume, inflation, and other macroeconomic conditions can result in big movements in earnings. For bottom-up forecasters, relatively small errors in the projection of revenue or cost often amount to big errors in their difference, i.e., profits.(16)


I view the record we have just examined as mixed -- decent results over the past ten to fifteen years on GNP and inflation, but poorer ones on financial variables such as interest rates and earnings. Maybe these results just serve to validate the old saw: "Forecasting is difficult, especially when it is about the future."

But we have to take this subject seriously, for one very important reason: Virtually every significant decision made in the private sector and in the public sector is based upon forecasts. In some cases the forecasts may be implicit, i.e., not developed formally or spelled out in any detail. In many other cases, though, numerical estimates play a central role. Forecasting is a means to an end -- that of making better decisions.

Given the importance of these private- and public-sector decisions, it is not surprising that the demand for economic forecasts has never been greater than it is today. Although everyone will have a guess as to how the future environment will look, business economists have the opportunity to provide crucial input to the process. We bring to the endeavor a set of skills that permit us to play a central role, including:

1. A theoretical framework for identifying and

evaluating some of the forces that will determine

economic outcomes.

2. The theoretical framework permits us to see how

one portion of the economic system affects others.

These interrelationships are important for

estimating the sequence of responses to exogenous shocks,

policy changes, and international developments.(17)

3. An understanding of the process by which

economic policies, especially monetary and fiscal, are


4. Formal models and forecasting tools that may be

used to obtain quantitative answers that account for

the impacts of multiple (and sometimes competing)

forces. We can then adjust these estimates to reflect

our judgments about missing or difficult-to-quantify


5. Access to and familiarity with data and information

sources that can provide a foundation for empirical


6. Access to and knowledge of outside experts who

can help in assessing the prospects.

In short, to the important task of generating useful forecasts, the business economist has a great deal to contribute.


While our role in formal forecasting is critical and likely to remain so, business economists possess a wide variety of other skills that may be productively used to aid in private-sector decisions. These non-forecasting activities may be referred to as economic analysis.

A central question facing all companies is the assessment of markets, both existing and new. Business economists can help determine what causes demand to move higher or lower -- both those factors that relate to the environment (such as economic growth and its components, interest rates, and exchange rates) as well as variables that are at least partly under management control (such as prices and advertising/promotion). Market analysis may be an even more important endeavor in the years ahead, given the trend toward reduced trade barriers and the opening up of hitherto command economies in Eastern Europe and the Soviet Union.(18)

A second area of analytical endeavor relates to evaluating the impact of policy changes, both upon the macroeconomic environment as well as upon the specific sector and industry conditions a firm faces. Changes in monetary policy, as they affect interest rates and eventually economic activity and inflation, are an obvious area of continuing focus. Fed watching/anticipating is particularly important in some companies.

But the policy assessment arena is much wider. What are the probable budget changes and their impacts upon demand? Certainly companies that supply military goods or health-care services will have a major interest in these questions today. Potential changes in tax policy will affect a wide range of companies, both in derived demand for their output and in their profit and loss statements. The likely effects of international policy changes, e.g., GATT negotiations, the North America Free Trade Agreement talks, and Europe 1992, are a growing area of emphasis for many of our companies. Business economists have a great deal to add to the discussion of developments in these arenas.

We are also a group of professionals who know how to deal with both a scarcity of specialized information and the "overload" of conventional data. We know where to find data on newly opening markets and countries. At the same time, with our computer skills we can help to process efficiently and to present large quantities of data to aid in the decision-making process.

Moreover, we have a framework that can help in thinking clearly about some of the vexing and contentious questions of our time. The application of theory will help to clarify some of these issues, but in other cases, empirical estimates will be our main contribution. In many instances, to be sure, the quantitative magnitudes will be subject to considerable controversy. But as economic analysts we can lend clarity to the often-confused discussion one hears of these issues. For our own good, we should emphasize to our managements the role we can play in these nonforecasting tasks as well as in forecasting itself. We may also note that our profession made important contributions to the analysis of major issues in the past decade, including the budget deficits to which U.S. fiscal policy was leading, the impact of low saving on real interest rates, and the prospect that supply/demand forces would prevent oil prices from going only up.


In spite of the fact that business economists do a wide variety of other things, the forecasting function is going to remain a central task for us in the years ahead, given the important decisions that may be affected and the capabilities of our profession. There are some steps we can take to help ourselves, which I will touch on next. Let's split them into the substantive versus the stylistic aspects, even though in a few cases my suggestions have elements of both.

Substantive Improvements. There are some fundamental parts of the forecasting process on which we can improve:

1. Better Theoretical Base. Our models, be they

formal or informal, should be based upon the best

theoretical constructs we can find. Ideological

divisions within the profession will not go away, but

business economists must focus on methodologies

that give correct and practical answers rather than

be chained to a framework that just happens to be

favored by their favorite professor. Unfortunately,

macroeconomic theory is in a state of considerable

flux right now, so it is hard to establish the one

right system to employ. But there are some clear

dimensions in which we must improve, including:

more attention to international forces (growth,

inflation, capital flows, exchange rates); and greater

use of expectational information and theories.

2. Statistical Skills Should Be Enhanced. Forecasters

need to be conversant with the principal techniques

and aware of the advances in areas such as model-building

and solution, autoregressive moving-average

processes, and vector autoregressions.

NABE members are aware of the need to improve

their capabilities here. In surveys of interest in

training courses, our members singled out

advanced training in forecasting and other statistical

applications as their principal areas of focus. We

expect some new courses to be available during the

next year.

3. Emphasis on Key Business Variables. We must be

very careful to orient our forecasting efforts toward

magnitudes that have a direct and important impact

on our companies' businesses. The main objective

must be contributing to superior-decision making,

which in turn will boost profits, by identifying key

causal forces and then providing useful forecasts

and analysis. Too often there has been a tendency

to focus on variables that are traditionally used by

forecasters or to use time-honored approaches that

are already available. Evidence already suggests

that business economists have been moving in

exactly this direction. A recent NABE survey found

that we are spending less time on macroeconomic

issues and more time on industry/financial matters

and on international questions.

4. Identify Uncertainties. Although there is always a

hope that near-perfect forecasts can be made and

that the uncertainties in decisions can be

eliminated, we know that the world does not and will

not work this way. Consequently, business

economists need to emphasize that our forecasts are

conditional upon certain assumptions about policy

and exogenous variables. When there is a shock to

the system -- such as the invasion of Kuwait -- we

must be ready to help evaluate the potential

impacts. Much more attention should be given to what

will happen if alternative assumptions are used. By

doing simulation exercises, we can help our

companies assess the levels of risk involved in their

decisions. Thus, many of us should write more

memoranda or reports about what will happen if

unexpected policies or external shocks begin to

emerge. To be effective on this front, we need to

read widely (often on subjects besides economics),

and we need to pay attention to a wide range of

developments, especially those overseas.

5. Use of Consensus Estimates in Macroeconomic

Forecasts. There are several ways in which we can

benefit from explicit use of consensus forecasts. On

variables such as overall activity and inflation, the

recent record of the consensus has not been too

bad. Forecasters should be aware of how their own

estimates vary from the median (or the average) of

forecasters and be ready to explain the fundamental

reasons for differences. Simply telling our

audiences that the "consensus must be wrong" does not

contribute to understanding of a forecast.

6. Use of Consensus Estimates in Interest Rate

Forecasts. There is a different way that we might

productively use consensus estimates of interest rates.

These markets are generally efficient, and

participants have already taken account of available

information, both coincident and leading. As we saw

earlier, the direction of interest rate changes has

often been the opposite (especially on long-term

rates) of what forecasters had projected. We should

alert users of our forecasts to this pattern and urge

them to consider what surprises are likely to occur

over the prediction horizon. In short, for financial

market forecasts it often pays to be a contrarian and

to go against the consensus, at least in direction.

Stylistic Improvements. There are additional changes we can make to improve the standing of our profession. Many of these have to do with the way we present forecasts, what we claim for them, and how we deal with the errors that inevitably will be made. Here, of course, I am clearly offering my own, and probably biased, views of how we should proceed:

1. Better Communication. A general plea for

improved communication skills has often been made;

indeed, a number of NABE annual meetings have

included workshops on communication. We need

to be aware of our customer needs and the

pressures upon them, to indicate clearly what our

forecasts can be expected to provide, to state what they

are conditional upon, and to aid in the process of

managing the inherent uncertainties.

2. Education in Thought Process. We should attempt

to acquaint our managements with the thought

process involved in forecasting, which may be more

important than any point estimates. Such a thought

process will present a natural avenue for others to

provide useful input and to become participants in,

rather than just users of, forecasts.

3. Motivation to Be Right. It is my view that most of

the audiences we are trying to reach want to hear

from forecasters who are motivated to be right, as

opposed to controversial or simply attention-grabbing.

In our work as economic analysts, we may

develop evidence that hints at wildly unexpected

or unusual outcomes, but we should be careful to

identify those positions as such. When users of

forecasts hear an economist say: "You will never hear

a consensus prediction from me," they have a right

to know if this forecaster is willing to be wrong just

to avoid being labeled as in the consensus.

4. Acknowledgement of Errors. Given the hugely

complex and dynamic systems for which we are

trying to forecast the outcomes, errors are

inevitable. I think most users of forecasts understand

and accept this fact. Where their tolerance is likely

to be stretched, however, is when forecasters

refuse to acknowledge or to analyze the source of

their errors. There is also little sympathy for the

occasionally asserted views: "The data will be

revised to show that our earlier forecast was right"

or "If the Federal Reserve had only run monetary

policy properly, my forecast would have been

right." These and similar excuses are not likely to

score any points with our audiences, and business

economists would be better off not to assert them.


Forecasting is bound to remain an essential function of our profession. Important decisions are dependent upon our forecasts, and we bring critical skills to the task. Some of the results we have achieved are favorable, others are less so, but all professions trying to predict future events are in the same predicament. When sizable forecast errors occur, there will be new rounds of complaints. I suspect we will just have to accept this fact as an occupational hazard. We do need to broaden and deepen our skills and to be more effective communicators -- both of what we can do and of how we can help users of forecasts to function effectively in the face of uncertainty.


(1)I am certainly not the first to point to the anxiety within the business economics profession or to cite the role played by forecasting difficulties. See: Don R. Conlan, "In This Autumn of Our Discontent, a Word to the Wise; Otherwise . . .," Business Economics, January 1983, pp. 10-14; A. Nicholas Filippello, "Where Do Business Economists Go from Here?" Business Economics, January 1985, ppp. 12-16; Marina v. N. Whitman, "New Directions for Business Economists," Business Economics, January 1987, pp. 51-55. (2)See, for example: Stephen McNees and John Ries, "The Track Record of Macroenomic Forecasts," New England Economic Review, November-December 1983; Peter L. Bernstein and Theodore H. Silbert, "Are Economic Forecasts Worth Listening to?" Harvard Business Review, September/October 1984. (3)Data revisions constitute a complicating factor in assessing the accuracy of GNP forecasts. The historical record of the forecasts, of course, remains unchanged once they are made and published. The GNP figures are revised two times for each quarter after the preliminary estimate is reported; such quarterly revisions represented no problem because the annual totals are comprised of the final quarterly totals. But in July of each year, the Bureau of Economic Analysis revises the figures -- both annual and quarterly -- for the three prior years (mainly to reflect more complete and detailed source data). Every five to eight years an even more comprehensive "benchmark" revision is made, in which changes in definition, classification, or methodology may be introduced, along with adjustments to reflect the availability of complete data. Sometimes the resulting annual real GNP changes are altered by a full percent or more. For example, the real GNP change in 1978 was originally reported as +3.4 percent. Another benchmark revision is due to be released in November 1991. (4)In this tabulation the error statistics show the Blue Chip forecasts to be slightly better than those of the NABE panel on the average absolute error and root mean squared error bases and about the same on the Theil U Statistic. The composition of the panels is one factor, although we suspect that a fair number of forecasters participate in both surveys. As noted in the text, there is also a difference in the timing of the two tabulations, which may be another part of the explanation. We might have tried to obtain either NABE forecasts made a bit later or BC forecasts made a little earlier in order to put the two panels in the same position. But our objective here is less to run a race between consensus estimates than to draw some general conclusions about the performance of business economists in forecasting, for which these representative samples will serve the purpose. (5)In light of the revisions in the historical data noted previously, we recomputed the accuracy statistics based upon the original report for the year (usually as of March of the following year) and also after the first July revision. In spite of the revisions, which for some years were large, there is remarkably little impact on the error statistics. Only the RMSE is altered, and the change is quite small. (6)Here again, see the articles by McNees and Ries and by Bernstein and Silbert cited in footnote 2. (7)The U Statistic is defined as follows: where p= the predicted change and a= the actual change. (8)The naive model of "no change" may seem very unrealistic. Perhaps a more natural assumption would be that the recent growth rate continues, or "same change as last year." This alternative naive model performs much worse than the forecast panels, with an average absolute error of 2.2 percent and RMSE of 2.6 percent (using the originally published data for each year; see footnote 5). (9)The BC forecasts of the CPI do not begin until 1980, so we use the period 1980-90, for which comparable estimates for BC and NABE are available. Here as with GNP, BC forecasts are drawn from the October survey while NABE estimates were made in late July/August for the annual meeting. Also, for two years, 1981 and 1982, NABE asked only for a forecast of the GNP deflator, but in all other years the CPI was collected. The actual CPI changes on an annual basis are computed from the average of seasonably adjusted monthly data. (10)See, for example, Robert A. Bennett, "Economists Missing the Mark," The New York Times, December 12, 1984, p. D1. (11)Robert J. Eggert, "Retrospective on Consensus Forecasts by the National Association of Business Economists," updated September 22, 1982. (12)Stephen K. McNees, "The Accuracy Keeps Improving," The New York Times, January 10, 1988, Business Section, p. 2. (13)These interest rate forecasts are published in The Wall Street Journal in early January and early July of each year; Tom Herman graciously supplied the historical data. One reason to use these forecasts of interest rates at a specific point in time is that they are closely related to portfolio management decisions. Average interest rates for a whole year can mask a great deal of intraperiod volatility. (14)The forecasts examined here are all drawn from the experience of the 1980s. Relative to earlier periods, there was much greater volatility (i.e., higher standard deviations) in the past decade. A number of changes contributed to this pattern and made financial market forecasting even harder than it would have normally been. One was the major shift in inflation -- first the big acceleration in the 1970s, due in part to the oil shocks, and then the sharp slowdown as the 1980s unfolded. Another was the Federal Reserve's shift to a monetarist approach, where specific money supply targets are employed as opposed to a primary focus on an interest rate target. With a quantity-of-money objective, there will be more variability in interest rates, especially at the short end of the yield curve. (15)IBES, or Institutional Brokers Estimate System, a division of Lynch Jones and Ryan, has collected and disseminated this information. The bottom-up estimates have been collected since 1979, but the top-down figures only go back to 1983. Don Conlan and George Miller at Capital Research Company supplied some of the early figures and called my attention to some of the earnings forecast results. (16)Another reason for earnings estimate errors is the special charges companies take periodically to dispose of businesses or to revalue assets. Both top-down and bottom-up forecasters have trouble with this, because the adjustments tend to be one-time events, largely at the discretion of managements. (17)I have in mind here the same point that Don Conlan developed in his 1982 NABE Presidential address about our ability to think in circles (to be distinguished from the conventional usage of the term circularity). He argues: "What we bring to the business decision-making process is an ability to take linear patterns of reasoning, run them through a loop-feedback system, and see if they come out the other end with the same set of answeres. Where divergences arise between the linear and the circular ways of reasoning, we can help avoid costly mistakes and produce better decisions." (18)Jay N. Woodworth emphasized the growth of international opportunities for business economists in his 1989 Presidential speech, "Different Economic Systems: Vast New Opportunities for Business Economists," Business Economics, January 1990, pp. 10-17.

Richard D. Rippe is Senior Vice President and Chief Economist, Prudential Securities, New York, NY. This Presidential Address was presented at the 33rd Annual Meeting of The National Association of Business Economists, September 22-25, 1991, Los Angeles, CA. The author is grateful to Don Conlan, Martin Feldstein, and Jay Woodworth for comments and suggestions and to John Canally for excellent statistical assistance.
COPYRIGHT 1992 The National Association for Business Economists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:National Association of Business Economists
Author:Rippe, Richard D.
Publication:Business Economics
Article Type:Transcript
Date:Jan 1, 1992
Previous Article:The Adam Smith address: education, labor force quality, and the economy.
Next Article:Economic growth and policy in the nineties.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters