Printer Friendly

Forecasting in the 90s: it's not like it used to be.

Forecasting in the

90s: It's Not Like It

Used To Be

Most state and many local governments rely on forecasting models for establishing their annual budgets. These models rely on techniques which establish statistical relationships between "independent" and "dependent" variables, relationships obtained through regression or other time-series techniques which link economic and demographic changes with their fiscal outcomes on the basis of historic data. Most other forecasting techniques, including incremental, average cost, moving average and per capita approaches all share some degree of reliance on historic data.

Forecasting techniques which rely on historic data will maintain their accuracy to the degree that historic relationships remain valid over the forecast period. While such models can be powerful forecasting tools, even under "normal" economic circumstances they require annual or semi-annual updating and review to maintain the accuracy of the relationships.

Failure of historically based models to yield an accurate or reliable forecast can be attributed to either one or both of the following: the inputs (often referred to as "drivers") were wrong, or there was a fundamental change in the relationship between certain, drivers and their dependent variables (often referred to as outputs, or in this case, the forecast).

Given the widespread budget shortfalls which are being experienced by both state and local governments throughout the U.S., it is reasonable to conclude that forecasting models in some jurisdictions have not yielded accurate results. Indeed, the list of jurisdictions experiencing budget shortfalls includes many state and local governments which have devoted substantial resources to their forecasting efforts.

Some of these shortfalls may not have been the result of the models, but may have been caused by politicians and legislators failing to heed their results. Other forecasts may have yielded overly optimistic results if they relied on national or regional forecasts of economic activity. Most blue-ribbon economists and forecasting firms have been wrong in predicting a "soft-landing" and a quick recovery, followed by a return to prerecession, business-as-usual growth rates along with the attendant consumption, growth, spending and borrowing characteristics. Inaccurate assumptions contributed to unreliable forecasts in 1990 and 1991, but they do not explain everything.

Forecasting models (econometric and otherwise) which rely on 1980s-era data may have an inherent expansionary bias. The decade of the 1980s was a period of major expansionary tendencies in employment, households, debt, income, construction, retailing and consumer purchasing, to mention a few. Because of the extended nature of the 1980s growth cycle, it was not possible to verify the downside accuracy of forecasting models during this time. Now that the recession has taken hold, it seems clear that expansionary forces at work in the 1980s are unlikely to continue in the near- to mid-term.

Many forecasting models implicitly assume that the same elasticities exist on the downside as on the upside. That is, that revenues will increase or decrease proportionately with changes in the key driver variables. This assumption is now being tested by reality. Other factors appear to be at work. Decreases in discretionary spending, other belt-tightening by consumers and emerging economic and demographic factors have combined to cause divergent or differential effects under recessionary conditions. Some of these conditions may extend well into the recovery period.

This will necessitate re-estimation and validation of relationships between key forecast drivers and their budgetary outcomes. Two of the most critical drivers which have the most immediate effect on state and local governments involve construction and personal income.

During the construction boom of the 80s, many state and local governments saw healthy increases in their tax receipts from construction activities. Much of the activity fueling this boom, however, was speculative in nature and is unlikely to recur at levels experienced during the 1980s. A significant portion of tax receipts, including income, sales and business taxes, and fees, was realized from construction and related spending. The oversupply of most types of real estate, the credit crunch and potentially slower rates of in-migration and household formation may affect the fundamental relationships that underlie forecasts for revenue sources which are sensitive to construction activity.

Downturns in the construction industry also affect receipts from property and real estate transfer taxes related to existing properties. Rates of appreciation for existing properties have slowed or posted real declines in many areas. Decreases in rates of appreciation or real declines in valuations have often been greater than corresponding declines in personal income and employment. Residential property values are ultimately linked to changes in personal income and consumer debt levels. Commercial property values, likewise, are closely linked to employment. Readjustments in employment in the services and retail sector will affect revenues sensitive to these variables. Similarly, changes in consumer spending patterns and personal income will affect property values and related taxes.

Sales and business tax receipts are also highly influenced by consumer spending. The most recent three years of data on retail sales activity suggests that a smaller portion of personal income is going toward the purchase of consumer items than previously. Rather, more is being diverted to the paying down of debt, higher taxes and increased health costs. According to figures compiled by the Federal Reserve Bank of Cleveland, consumers reduced debt by $20.3 billion between 1990 and 1991. Total private borrowing (excluding financial institutions) decreased from a peak of $600 billion in 1988 to approximately $200 billion in 1991.

Reduced consumer borrowing is often an indication of lower consumption. In the current recession consumers are not only borrowing less, but they are diverting more income toward savings or paying down debt. Forecasting models which rely on personal income and related variables to make projections of sales and business taxes and do not account for this shift in behavior will tend to overestimate receipts from these revenue sources.

Data compiled by the Federal Reserve Bank of St. Louis illustrate this. During the expansionary period between 1987 and 1989, disposable personal income increased at an annual rate of 3.3 percent, while retail sales increased 1.7 percent. The strong positive relationship was reversed, however, during the recessionary period when personal income managed an increase of eight-tenths of 1 percent, yet retail sales plummeted by 3.5 percent (annualized, seasonally adjusted).

Demands for government services are also dynamic and the convergence of factors leading to declines in revenue sources have affected the expenditure side as well. Many state and local governments have experienced above-average rates of growth in programs for health care, human services, prisons, law enforcement and environmental protection, amongst others. Higher outlays have occurred as a result of mandates, increased at-risk populations and other demographic and social changes.

What can be done to correct forecasting models for these sources of error? First, drivers obtained from national forecasts or economic consulting firms should be subject to validation by local experts and officials. Consensus estimates of key inputs, such as employment, wage rates, population, housing starts, should be made for at least three scenarios: high, medium and low. The forecasting model should be run for all three allowing comparison of results.

Second, state and local governments which have invested substantial effort in development of forecasting models may need to develop subsidiary models for analysis of leading indicators based on local or regional data. These indicators could be developed specifically to monitor changes in activity which signal local and/or regional recessionary or recovery tendencies. Development of local indicators will provide analysts and forecasters with regionally sensitive data. These indicators may include such items as local employment rates, bank deposits, building permits, net new business formations, utility hook-ups and vacancy rates, and other locally sensitive data. These data can then be used to construct or modify national or regional drivers.

Finally, existing models can be adjusted to reflect fundamental changes in relationships between variables. To begin to isolate these changes, the model may be run for the historic period with actual data being used as input drivers. Equations which result in unacceptable variances with actual budgetary data should then be reformulated. The results can be used to calibrate or weight existing relationships, or to revise the model to obtain relationships more reflective of recent experience.

There is a tendency to invest models with heroic expectations. The numbers produced can greatly influence policy debates and program activities. During times of economic expansion, forecasts tend to underestimate revenues. While underestimation may constrain budget growth for that year, it creates a surplus for the following year. But in times of economic contraction, the consequences of over-estimation can be much more severe. Jobs may be eliminated and vital services can be placed in jeopardy. If our economic circumstances are reflective of more than just transitory, cyclical change, forecasting models will require recalibration only for an interim period. Yet if the circumstances are reflective of fundamental economic restructuring, more substantial modifications will be necessary. Either alternative requires examination of the new data to determine if existing relationships are still valid.
COPYRIGHT 1992 Government Finance Officers Association
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:evaluation of historical forecasting model
Author:Siegel, Michael
Publication:Government Finance Review
Date:Apr 1, 1992
Previous Article:BenefitLine: fast efficient service for retirement system participants.
Next Article:Declining federal transfers.

Related Articles
The forecasting process: guidelines for the industrial engineer.
Rural electric power requirements forecasts: detecting and correcting for weaknesses and bias.
Business Cycles, Indicators, and Forecasting.
Financial forecast model: council members learn by doing.
Forecasting and Empirical Methods in Finance and Macroeconomics.
Taking the Guesswork Out of Forecasting.
Outlook for 2005.

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