Housing Starts: a Year 2000 Forecast using
Long Term Interest Rates and Total Consumption Spending
 
JASON M. MYERS and SCOTT C. SMITH
College of Business
Western Carolina University
 
Abstract
 
The nation’s housing starts are expected to increase over the next two years. This forecast is based on the Keynesian investment function model, assuming housing is determined by two-year-lagged long-term interest rates and real total consumption spending. Housing starts will continue to increase at three percent a year over the next two years. Investment in housing will rise because interest rates will remain low and consumption expenditures will continue to rise. Continued consumer optimism will support higher levels of both consumption spending and investment in new housing for the foreseeable future. In order to keep the economy growing at its current pace, the Federal Reserve should attempt to keep the price level and interest rates stable to prevent a decrease in consumption spending and maintain consumer confidence. (JEL: E21, E31, and E43)
 
Part 1. Introduction
 
The purpose of this paper is to forecast the nation’s housing starts for the years 1999 and 2000. The explanatory variables are the secondary market long-term interest rates (LTIR) and total consumption spending (TCS). The approach for the paper is based on the Keynesian investment function (Keynes 1936, pp. 135-146). The data used to forecast housing starts are monthly statistics from January 1990 to December 1998. The Keynesian investment function relates investment to interest rates and total income. Housing starts and TCS are used as proxies for investment and income.

Investment in housing is a critical determinant of the direction of the economy. Investment in housing and TCS would be expected to decrease whenever consumer confidence falls, but housing investment would also be expected to fall as TCS rises whenever LTIRs rise. These events would be leading indicators of a recession. Since interest rates could rise unpredictably due to unanticipated inflation, the forecast horizon was restricted to two years. With a two-year forecast, the reliability of the findings increases.

The remainder of the paper is arranged as follows: Part 2. documents the data on housing starts, TCS, and LTIR for the past eight years; Part 3. presents the economic theory of the Keynesian investment function, on which the forecast is based; Part 4. presents the forecast of housing starts for the years 1999 and 2000; Part 5. discusses the economic importance of the forecast; Part 6. addresses effects on economic policy.

 
Part 2. Data
 
Several variables were examined but found to have poor explanatory power in relation to housing starts. Variables employed besides the LTIR and TCS are the consumer price index (CPI), short-term interest rates (STIR), and gross domestic product (GDP). The STIR was not very effective in explaining housing starts since STIR is a better indicator for short-term investments like certificate-of-deposit accounts and treasury bills. Short-term borrowing is more likely to respond to changes in STIR than 30-year mortgage lending. The STIR used was the one-month certificate of deposit rate. The LTIR would be expected to better explain long-term investments like housing. The CPI also had low explanatory power. GDP is reported quarterly and all other variables used in this paper are reported monthly.

The CPI monthly data was taken from the Bureau of Labor Statistics. The monthly data for housing starts, LTIR, STIR, TCS, and GDP were taken from the Federal Reserve Bank of St. Louis Federal Reserve Economic Data (FRED). The FRED variables used for the forecast are GS30 (30 year government securities interest rate), PCEC92 (real personal consumption expenditures), and HOUST (housing starts) and are all real variables. PCEC92 is represented in billions of chained 1992 dollars and both PCEC92 and HOUST are seasonally adjusted at annual rates (SAAR). LTIR is taken from the secondary market and is the thirty-year Treasury bond rate with a constant maturity, and is not seasonally adjusted. Little or no seasonal variation is present in long-term interest rates. The primary source of all of this data is the United States Department of Commerce, the Bureau of Economic Analysis, and the Federal Reserve Board of Governors.

Regressions on the forecast data were performed on two parameters. In the first regression, housing starts is compared to TCS and LTIR for the months between January 1990 and December 1998. A regression was then run with the right hand side variables lagged two years comparing housing starts to TCS and LTIR between the months of January 1992 and December 1998.

Housing starts is used as the forecast target since it is a leading indicator of economic performance. As discussed above several variables were considered. However, only LTIR and TCS had strong explanatory power for housing starts, according to t-statistics and adjusted R-square. TCS is used as a proxy for GDP. Intuitively, GDP or TCS should be significant in forecasting housing starts since consumer spending responds to interest rates through the loanable funds market and reflects consumer confidence. LTIR should be inversely related to housing starts. When LTIR decreases, the amount of housing starts should normally increase. Including both interest rates and TCS as right hand side variables allows a more accurate forecast of housing starts.
 

Part 3. Economic Theory:
The Keynesian Investment Function as a Forecasting Instrument
 
Either TCS or LTIR can change the amount of housing starts each year. When LTIR and TCS act together they increase the accuracy of housing starts forecast. Monthly data for LTIR and TCS are used to forecast the volume of housing starts over the next twenty-four months. The prediction of housing starts will be determined by the Keynesian investment function which is:
 
I = f(i, Y) (1.

This is a simple model that will be used to forecast housing starts. I is investment that consumers commit to buying new houses. The function shows that either variable i or Y influences housing starts (I). The interest rate, (i) is LTIR, the long-term interest rate. Y is TCS, total consumption spending. The Keynesian Investment Function becomes:

 
Housing Starts = f(LTIR, TCS) (2.

Each variable can affect housing starts independently of the other variable. If LTIR continues to decrease, consumer buying power increases. Rising consumer incomes and confidence promote more purchases of new houses (E-mail Trends, 1999). Housing starts can also be an independent function of TCS. As TCS increases, the number of houses people can afford to buy also increases, because the income of one household may come from the consumption spending of another. TCS has increased due to the amount of increased income that American consumers have realized over the last several years. Increased incomes lead to higher levels of consumption at any level of GDP.

A few limitations of this paper should be noted. First, TCS is very volatile and depends highly on other economic factors, such as the unemployment rate and interest rates, which are not captured in the analysis attempted here. Another shortcoming is that LTIRs can be altered anytime the Federal Reserve Board decides interest rates need to be raised to combat inflation.

 
Part 4. Estimates of Housing Starts based on LTIR and TCS
 
The investment function, equation 2, was estimated using monthly data from January 1990 to December 1998. The modified investment function formula using housing starts and lagged LTIR and TCS is written as:  
HST = a0 + a1* LTIRT-24 + a2 *TCST-24

Without lagging the right-hand-side data, regression results are (with t-statistics in parentheses):

HST = -814.81(-4.78) + 0.23(0.02)LTIRT + 0.48(15.79)TCST
 
When the regression was run with a two-year lag, the relation between LTIR and TCS and housing starts appears stronger. Regression results with right hand side variables lagged two years are:
 
HST = -828.07(-3.90) + 7.96(0.71)LTIRT-24 + 0.497(11.47)TCST-24

First, the t-statistics on the two-year lag are better than the unlagged regression. Next, the adjusted R-square for the two-year lagged regression is 61%, which means that there is only 39% variation in housing starts that may be explained by another variable. The Adjusted R-Square on the unlagged investment function was only slightly better (70%). However, the t-statistics for the unlagged regression did not produce as good results as the lagged regression. Forecast housing starts calculated from this equation are given in Table 1.

 
Table 1
Forecast of Housing Starts
Date
Housing Starts
Annualized Percent Change in HS
November-98
1654.00
10%
December-98
1738.00
14%
January-99
1634.38
7%
February-99
1651.84
0%
March-99
1640.24
4%
April-99
1644.29
7%
May-99
1645.05
7%
June-99
1649.23
1%
July-99
1700.98
-1%
August-99
1697.29
5%
September-99
1698.41
8%
October-99
1690.60
0%
November-99
1717.58
4%
December-99
1711.31
-2%
January-2000
1728.50
6%
February-2000
1739.79
5%
March-2000
1740.21
6%
April-2000
1772.38
8%
May-2000
1789.14
9%
June-2000
1803.35
9%
July-2000
1789.55
5%
August-2000
1812.10
7%
September-2000
1814.82
7%
October-2000
1825.35
8%
November-2000
1823.93
6%
December-2000
1832.85
7%
 
This forecast is consistent with the past behavior of the variables. If LTIR remains low and TCS continues to rise, housing starts will continue to rise over the next two years. The continual rise in housing starts can be shown in the following chart plotting actual and forecast housing starts.
 
Chart 1
Housing Starts from 1990-2000
 
 
Part 5. Forecast Implications: Continued Economic Growth
 
Housing starts are a leading indicator of economic activity, acting as a barometer of consumer confidence. Additionally, consumers purchase more houses when their buying power increases. If either LTIR increases or the TCS of the economy increases, the amount of leverage for purchases is decreased. Greater TCS can lead to a reduction in buying power since the Marginal Propensity to Save (MPS) has diminished over the last few years. Since buying houses requires a significant amount of money, consumers must increase their MPS. If MPC increases, consumers reduce their ability to afford houses or other large ticket items. Because of the forecast implications presented in this paper, if housing starts begin to decrease due an increase in LTIR or a decrease in TCS, these indications may provide an early warning of a recession.

The forecast reveals no sign of the economy slowing down. Over the next two years, interest rates will most likely stay at their low levels since inflationary pressure has not increased. Furthermore, TCS will remain high due to an increase in buying power because of low interest rates, reduced unemployment (Email Trends, 1998), and an average increase in salaries.
 

Part 6. Policy Conclusions
 
Housing starts are forecast to rise three percent per year over the next two years. Consumer spending should continue its annual growth of four percent. Interest rates should stay constant over the next two years, remaining between five and seven percent on long term mortgage loans.

If this forecast turns out to be true, the economy is in for a prosperous two years. With interest rates low, consumers can afford to purchase more items including homes. An increase in the MPC provides an abundant amount of employment opportunities for U.S. citizens.

Even though higher wage demands are affecting businesses, many companies have learned to accomplish more tasks with fewer workers. Additionally, businesses are operating with increased profits in comparison to their past performance. What has led to a decrease in unemployment is the changing demand for businesses to provide more personal services to consumers. Even if the economy does not turn out the way the forecast predicts, the worst that is likely to happen is that unemployment may rise back up to its traditional standard of five to six percent.

Interest rates should stay the same over the next two years. If the Federal Reserve Board finds it necessary to raise the interest rates due to inflation, the Board should be cautious about raising the interest rates too high. As it has been stated, an increase in interest rates will increase unemployment, decrease consumer spending and inherently decrease the amount of housing starts.
 

References