A Two Year Forecast of the Unemployment Rate
College of Business
Western Carolina University

U.S. unemployment is expected to be between 4.2% and 4.4% for the next two years. This forecast is based on the Keynesian aggregate expenditure model and the Phillips curve using U.S. civilian population, real exports of goods and services, and inflation as explanatory variables. The correlation between unemployment and the three variables was very strong. The forecasting regression's adjusted R-square indicated 95% of the variation in unemployment is explained by variation in the remaining three variables (inflation, exports, and population). Workers and employers should expect current favorable job conditions to stay about the same over the next two years. (JEL: E24, E66)

Part 1. Introduction

This paper forecasts the quarterly unemployment rate for the years 1999 and 2000. The sample period for the data is from the first quarter of 1992 to the fourth quarter of 1998. This sample period runs from the end of the 1990-91 recession to the end of the available data. The forecast horizon is two years into the future. The explanatory variables are U.S. civilian population, real exports of goods and services, and inflation. This forecast is based on the Keynesian aggregate expenditure model and the Phillips curve. This hybrid model is appropriate because of its ease of use and it is readily understandable by economically literate individuals.

An accurate prediction of the unemployment rate is important because it can help government agencies and firms determine if there will be a shortage of workers and if more workers will need to be trained for the future. This forecast of unemployment will also be useful for people planning to enter the workforce in the next two years. The forecast horizon of two years was chosen to keep the forecast as accurate as possible. Going further into the future with three variables would have increased the risk of inaccuracy.

The rest of this paper is organized as follows: Part 2. presents the data used to forecast the unemployment rate and compute the inflation rate; Part 3. presents the theoretical basis for the approach adopted in forecasting the unemployment rate; Part 4. presents forecasts of unemployment for 1999 and 2000; Part 5. evaluates the importance of the forecasts for the economy; and Part 6. discusses conclusions for economic policy.

Part 2. Data

Half of the variables are taken from the Federal Reserve Bank of St. Louis Federal Reserve Economic Data (FRED). The remaining variables were inflation and population. Population was taken from the U.S. Department of Commerce, Census Bureau. Inflation was calculated from the Consumer Price Index (CPI) data taken directly from the FRED database.

The measures of Real Exports of Goods & Services, Civilian Unemployment Rate, and Consumer Price Index For All Urban Consumers are FRED variables EXPGSC92, UNRATE, and CPIAUCSL. The estimates of the United States civilian population are given monthly and the numbers are in thousands (consistent with the 1980 and 1990 decennial enumerations). The variable EXPGSC92 is given in billions of chained 1992 dollars at seasonally adjusted annual rates (SAAR) while UNRATE is seasonally adjusted and given as a percent of the total population.

The final variable CPIAUCSL, is also seasonally adjusted and tabulated from a base year of 1982-84 =100. The primary sources of these data sets are the U.S. Department of Commerce, Bureau of Economic Analysis (EXPGSC92), the U.S. Department of Labor, Bureau of Labor Statistics (UNRATE and CPIAUCSL), and the U.S. Bureau of the Census (U.S. Civilian Population). All data were calculated monthly except for the Real Exports of Goods and Services, which was calculated quarterly.

Because of this difference, all of the remaining data sets were converted to a time frame consistent with the data set of Real Exports of Goods and Services so that all data would correspond to identical time frames.

The data set for the U.S. Civilian Population was converted by taking the value given for the last month of each quarter (March, June, September, and December) and then substituting that value for the quarter that corresponded with that month.

The data set for inflation was calculated from the CPI by taking the current month's CPI figure minus the corresponding past year CPI figure and dividing the result by the preceding year's CPI figure. Next an average was taken of each month within a quarter (January, February, March for the first quarter; April, May, June for the second; July, August, September for the third; and October, November, December for the final quarter) and this was used as the quarterly figure for inflation.

For the quarterly Civilian Unemployment Rate, an average was taken of each month within a quarter. This is similar to the last step used to compute the inflation data.

Additional data manipulation was conducted on both the resulting quarterly data for unemployment and inflation to make the data sets more similar to the other two data sets. These two variables were both multiplied by one hundred so the resulting regression would provide more usable numbers for our calculations.

It is believed that all four variables used to predict unemployment are good predictors for several reasons. According to macroeconomic theory, inflation and unemployment are inversely related through the short-run Phillips Curve, which implies that one can be used to predict the other. Exports were used because it is believed that exports are a negative function of unemployment. When foreign demand for U.S. exports is high, U.S. GDP rises and unemployment falls, and vice versa. As for the population, it is known that unemployment is a portion of total population. For this reason, population was assumed to be a good predictor of unemployment.

It should also be clear that past unemployment could be used to predict future unemployment as well. Trends can be seen from this data set that may foretell the possible future rates of unemployment. The regressions, which were conducted on these data sets, support these assumptions and facts. All data used in this paper is readily available from the web sites of the Census Bureau and the Federal Reserve Bank of St. Louis.

Part 3. Economic Theory:
the Keynesian Aggregate Expenditures Model and the Phillips Curve

This forecast assumes imports will not have an effect on the unemployment rate. It assumes U.S. civilian population, real exports of goods and services, and inflation determine the unemployment rate. Using the Keynesian aggregate expenditure model to help project the unemployment rate for each quarter of 1999 and 2000:

Y = GDP = C+I+G+X = C+I+G+(EX-IM),

Where Y is real GDP, C is consumption spending, I is investment spending, G is government spending, and X is net exports (EX-IM). (EX-IM) corresponds to X or net exports and equals exports (EX) minus imports (IM).

Simplifying by dropping some terms, we obtain:

Y = f(EX)

The aggregate production function is written as:

Y = f(L, K)

Where L is labor and K is capital. Simplified that gives us:

Y = F(L)

Since GDP is a function of labor, GDP is also a function of the employment rate, making GDP a function of the unemployment rate. This allows the equation to be re-written as:

U = f(Y).

Where U is unemployment. Since Y = f(EX) and U = f(Y), we can write:

U = F(EX).

The Phillips Curve predicts the relationship between inflation and unemployment. According to this theory, inflation and unemployment are inversely related, at least in the short run. As a result, it can be assumed that unemployment is a function of inflation:

U = f(Inflation).

Also, unemployment is a subset of the total civilian population. Therefore, it is assumed that population can also be used as a predictor of unemployment.

U = f(P),

where P is population.

Combining all these items of which unemployment is an unspecified function, the hybrid model of unemployment can be written as:

U = f(EX, i, P),

where EX is exports, i is inflation, and P is population. After lagging the right-hand-side of the equation two years (eight quarters), the forecasting equation is written as:

Ut = A0 + A1(Ut-8) + A2(Pt-8) + A3(EXt-8) + A4(it-8), (1.

The right-hand-side data were lagged eight quarters to make it possible to forecast unemployment two years into the future.

Part 4. Empirical Results:
A Short Term Projection of U.S. Unemployment 1999-2000

The unemployment forecasting equation 1 was estimated with 1992.1-1998.4 quarterly data. The least squares regression method was used. The results from the regression are shown in Table 1.

Table 1
Regression estimate of Unemployment Rate Forecasting Equation 1: 1992.1-1998.4
Explanatory Variable Estimated Coefficient T-statistic
Intercept 73.21 5.04
Pt-8 -0.00026 -4.51
EXt-8  0.00206 1.22
it-8 -0.348 -1.88
R2 = 0.9491 F (zero slopes) = 69.92 Prob F = 1.62x10-09
The adjusted R2 is 0.949, indicating approximately 95% of the variation in unemployment is explained by variation of the three explanatory variables (exports, inflation, and population). The F-statistic is 69.9, indicating strong rejection of the null hypothesis of zero slopes. Although the t-statistic for EXt-8 was lower then an absolute value of two, it was left in the forecast equation because a regression with un-lagged right-hand-side data showed it as a good predictor. Inflation (it-8) was left in because the absolute value of its t-statistic is nearly two.
Table 2
Forecast of U.S. Unemployment Rate, 1999.1 2000.4
(Percent of U.S. Labor Force)
Quarter Forecast
1999.1 4.31%
1999.2 4.46%
1999.3 4.39%
1999.4 4.44%
2000.1 4.44%
2000.2 4.24%
2000.3 4.06%
2000.4 4.12%
This forecast projects the unemployment rate to average 4.4% in 1999 and 4.215% in 2000 taken from Table 2. These unemployment rates are very favorable for the economy. This forecast suggests the economy will not have any significant downturns that would warrant large-scale layoffs.
Part 5. Steady as She Goes

Many individuals and groups can use forecasts of unemployment for estimating and making decisions about a broad range of activities. This paper forecasts unemployment to remain at levels that are essentially the same as current unemployment. And since the United States is enjoying an economic expansion and low unemployment, this forecast is favorable in that it predicts a continuation of low unemployment at least two years into the future. The quarterly forecasts over the following two year period show unemployment in a range between a high of 4.46% in the second quarter of 1999 (June 1999) to a low of 4.06% in the third quarter of 2000 (September 2000). As a result of this forecast government spending on unemployment programs should remain relatively unchanged from current levels. This means current markets for employment should remain strong and individuals looking for future employment should not have much trouble finding work. It also implies tight labor markets are not expected to loosen up over the next two years.

This forecast is based on several assumptions. Some of these assumptions are the movement of the American workforce away from the production of goods to the production of services, problems associated with the year 2000 computer bug, and the increasing reliance on the internet and computer related training, research and development, and networking.

Individuals who have a concentration of skills that focus solely on manufacturing may soon find themselves unemployed or retraining for new careers as a result of these assumptions. Further, individuals who are computer literate and easily adaptable will most likely have an overabundance of job and advancement opportunities. The year 2000 problem will ultimately lead to more jobs in the computer industry to trouble-shoot, fix, replace, and/or maintain systems that may be affected.

Short-term volatility in businesses with heavy reliance on information systems may leave many workers out of work until the systems are fixed so that production can continue or paychecks can be processed. In addition, these estimates of unemployment are based on the entire United States population and therefore some areas may enjoy lower unemployment while other areas may have much higher unemployment.

It may become necessary to adapt to the changing economy by acquiring training or retraining on computer applications and new skills in order to maintain high worker productivity, labor demand, and employee value added. Another possible course of employee action would be to evaluate individual industries to find ones that are stable or growing.

It is expected that for the long term, businesses that focus on services and/or computers are more likely to continue expanding into the future while businesses that manufacture items such as cars may find themselves downsizing and moving their production to cheaper countries. Companies to watch would include internet companies such as Amazon.com and Yahoo!, and software companies like Microsoft. Companies such as Levi Strauss, that has just announced large-scale restructuring and layoffs, moving production to foreign countries, are companies for workers to stay away from in the future.

Part 6. Policy Conclusions

Civilian unemployment is projected to remain basically the same over the next two years at approximately 4.30%.

Based on the forecast, employers can expect the difficulty of finding qualified workers to remain the same. Employers will need to provide incentives that will give them advantages over other companies with similar positions to fill. Employers will also need to market employee incentives to distinguish them from other employers. Finding and keeping good employees will remain a difficult task over the next two years.

Workers who may need to change jobs or enter the job market for the first time can most likely continue to expect very good conditions. Based on this forecast, workers should be able to demand better wages and have better job opportunities. The job market is definitely leaning in their favor and it should continue to do so. Workers should have a lot to look forward to in the next two years.

As stated earlier in Part 5, this forecast assumes no significant change in government spending over the next two years, which seems likely given projected budget surpluses. The government is not expected to increase or decrease spending.

An increase in government spending could cause the economy to overheat, leading to inflation. Federal Reserve policy responses might also threaten prolonged prosperity, causing the current expansion to overheat (if interest rates were lowered) or bring about a recession (if interest rates were raised).

The Federal Reserve is not likely to raise or lower interest rates. However, if the Federal Reserve does adjust the interest rates the change should be minimal. The reason for this assumption is that if interest rates are lowered more investment and lower unemployment will result and the threat of inflation will develop. Raising rates would raise unemployment and slow investment, which would slow the economy's current expansion.