This paper examines housing starts as a leading economic indicator, explaining how the statistic is defined and measured by the U.S. Census Bureau. It reviews six-month forecasting methodologies, including the neural network model used by the Financial Forecast Center and the broad set of macroeconomic data inputs it relies upon. The paper then analyzes the real-world consequences of forecast accuracy, drawing on a June 2004 decline in housing starts to illustrate how discrepancies between projected and actual figures affect builders, lenders, investors, and related industries. The influence of interest rates on both construction lending and household purchasing decisions is also discussed.
The term housing starts refers to an economic indicator representing the number of public and private single-family and multi-family dwellings on which construction begins during a given period. A single-family dwelling that has begun the foundation-digging process constitutes one housing start, while a multi-family dwelling counts as a number of starts equal to the number of individual units it contains. In 1992, the U.S. Census Bureau began including a housing start figure for every project in which an entirely new dwelling or set of dwellings would be built on an existing foundation. The statistic does not include remodels, additions, or commercial buildings converted to residential use. Housing starts are considered a leading indicator of the overall economy because they directly reflect conditions of supply and demand (A.G. Raymond & Co., 2004).
As with most economic issues, organizations and businesses often rely not on past statistics but on forecast statistics for the period they must plan for. It is for this reason that economists must be very careful about the outcomes of their forecasting. For housing starts, these forecasts typically project six months into the future. According to the Financial Forecast Center, the housing starts forecast for the six-month period from July through December 2004 was as follows:
Total New Privately Owned Housing Units Started — SAAR (Thousands of Units)
Jul 2004: 2,087 | Aug 2004: 2,109 | Sep 2004: 1,856 | Oct 2004: 1,413 | Nov 2004: 1,550 | Dec 2004: 2,378
(Updated Tuesday, July 13, 2004; Financial Forecast Center, 2004)
The Financial Forecast Center uses what it describes as a state-of-the-art computer model known as a "neural network" for developing forecast statistics. This system relies far less on human input, thereby reducing the skewing of statistics caused by human factors such as preconceptions and biases. The multi-faceted set of data inputs used by this model includes, but is not limited to, the following (FFC, 2004):
The predictions generated by these models are checked against the historical movement of the relevant index, indicator, or stock as a double-check. If a model's predictions are deemed unrealistic, they are reformulated and rerun (FFC, 2004). It is within this framework of statistics and historical data that the housing starts forecast for the United States is produced.
The importance of housing starts forecasting becomes apparent when considering the broad economic consequences of unexpected changes in the figures. In a July 2004 article responding to an unanticipated fall in housing starts, one journalist captured the immediate impact: "Home construction defied expectations and took its biggest tumble in more than a year during June amid higher mortgage rates, and building permits also fell sharply" (Bater, July 13, 2004).
A number of macro-level decisions in the homebuilding industry depend on these forecasts. The industry as a whole must make supply-and-demand decisions about future building activity. If too little raw material is available, production falls below acceptable levels; if too much is allocated, a surplus creates concern throughout the raw materials industry.
New housing construction is important to the overall economy. Construction results in the hiring of workers, the production of construction materials and equipment, and the sale of large household appliances such as ranges and refrigerators. In addition, when owners or tenants occupy newly built housing, they often purchase new furniture, carpeting, and other furnishings (Frumkin, 1990, p. 129). As this observation suggests, many other industries are tightly linked to the homebuilding trade and can be adversely affected by inaccurate start statistics and erroneous forecasts.
The discrepancy between forecast and actual figures in June 2004 illustrated this risk clearly. Analysts had predicted a modest 1.2% increase in housing starts to a 1.990 million annual rate, yet the Commerce Department reported an 8.5% decrease to a seasonally adjusted 1.802 million annual rate — the sharpest drop since February 2003's 10.7% plunge (Bater, July 13, 2004). This divergence indicated that the economy was not reflecting the positive growth trajectory many economists had anticipated.
"Business-level risks from forecast versus actual data"
"How rate changes affect construction lending and sales"
In other words, people are not as ready for a new mortgage as economists would like to think at this point in the supposed economic recovery. The housing starts statistic, as a leading economic indicator, carries significant weight for a wide range of industries and stakeholders. When forecasts diverge sharply from actual data, the consequences ripple outward from individual builders and lenders to raw materials suppliers, appliance manufacturers, and the broader labor market. Accurate forecasting, sound lending practices, and realistic assessments of household purchasing power are all essential to maintaining stability in the housing sector and, by extension, in the overall economy.
You’re 51% through this paper. Sign up to read the remaining 2 sections.
Sign Up Now — Instant Access Already a member? Log inAlways verify citation format against your institution’s current style guide requirements.