Examining the NAHB/Wells Fargo Housing Market Index (HMI)

Special Studies, March 29, 2007

By Anupam Nanda, PhD.

Report available to the public as a courtesy of HousingEconomics.com


The National Association of Home Builders (NAHB) produces the Housing Market Index (HMI) every month to gauge builder sentiment regarding the demand side of the single-family housing market in the US. [1] Over the past few years, the HMI has been used increasingly by Wall Street firms, the Federal Reserve and other government officials, and various economic analysts, as well as by the news media to provide insight as to the health and probable course of the housing market in the near term.

The HMI is based on a survey that has been mailed to a panel of NAHB builder members every month since January 1985. The survey asks builders to rate housing market conditions based on their experiences. About 400 responses are obtained each month. Builders, with their experience and close contact with local market conditions, provide timely information about current housing market conditions as well as how home sales are likely to behave in the future.

The Housing Market Index (HMI) is a weighted average of responses to survey questions asking builders to rate three aspects of their local market conditions:  current sales of single-family detached new homes, expected sales of single-family detached new homes over the next 6 months, and traffic of prospective buyers in new homes.

The NAHB survey asks builders to rate sales and sales expectations as “good,” “fair” or “poor.” Builders also rate traffic of prospective buyers as “high to very high,” “average” or “low to very low”. Three component indexes are calculated by first seasonally adjusting the percentage of responses in the Good/High and Poor/Low categories. Then the formula [(Good/High - Poor/Low +100) [2] is applied to the seasonally adjusted numbers to produce an index. This formula puts each index on a scale ranging from 0 to 100.  The three components are then incorporated into the overall HMI, using weights based on correlations with present and future single-family housing starts.

Trends

The HMI has generally tracked single-family housing starts quite well (Figure 1). For example, both series reached peaks in December, 1998. In periods when the HMI has not moved in concert with starts, it has often moved first, anticipating the change in housing activity. In 1994-95, for example, the HMI fell before the starts series began its downward movement. More recently, the HMI started falling steadily in November 2005, while single-family housing starts were ascending to a record high in January 2006 and didn’t begin contracting until March of that year.

Figure 1. HMI & Single-Family Housing Starts

Sources: Builders’ Economic Council (BEC) Monthly Survey, NAHB Economics; Census Bureau; SAAR=Seasonally Adjusted at Annual Rate

The HMI also tracks single-family building permits quite well (Figure 2). It is useful to look at permits as well as starts because permits are less susceptible to seasonally factors (e.g., weather). Permits also are based on a larger sample than starts and have less month-to-month volatility.

Figure 2. HMI & Single-Family Building Permits

Sources: Builders’ Economic Council (BEC) Monthly Survey, NAHB Economics; Census Bureau; SAAR=Seasonally Adjusted at Annual Rate

It is advisable to be cautious when trying to draw conclusions from a visual inspection of graphs. When responding to survey questions in the last few months, builders may be comparing current performance with record-high performance a year ago. Thus, 1.3 million single-family housing starts or about 1.0 million single-family new home sales may now appear “poor” to builders, who might have considered it “fair” or even “good” in a year like 1995 (when there were 1.08 million single-family starts) or 2000 (when there were 1.23 million single-family starts).

Further Analysis

Given the difficulties involved in trying to draw conclusions by visually inspecting graphs of series that lie on different scales, we turn to statistics for further insight. The HMI correlates strongly with both single-family housing starts and single-family building permits out to six months in the future (Table 1). The correlations of the HMI with starts are above 70% in all cases shown in the table. The correlations of the HMI with permits are also high, especially out four to six months in the future.

Although we’ve shown that the HMI correlates well with single-family starts and permits, it is well known that other variables—such as interest rates—are also highly correlated with starts and permits. This leads to a natural question about predictive power of the HMI. How much does the HMI add to the ability of the NAHB, the Federal Reserve, Wall Street firms, and others to predict housing variables?

We test this with a statistical procedure used previously by Jack Goodman. Writing in 1994, Goodman concluded that the NAHB builder survey did help to predict housing starts and, in fact, was the only “attitude survey” that could help predict housing variables in a meaningful way when compared with three other surveys. [3]

The approach is conceptually simple: We start with a model that predicts starts or permits without the HMI, and then see if adding the HMI improves the model’s ability to predict these variables. Due to the plethora of changes in the U.S. housing market over the past two decades, we also test different time frames. In particular, we’re interested in knowing if the HMI continues to predict housing activity as well in the 21st century as it did 20 years ago

The results, which are somewhat technical, are shown in the attached appendix. In brief, the results show that the HMI contributes significantly to the models and helps them predict new single-family starts, irrespective of the time frame studied.  We find very similar results when we use single-family building permits in the experiment.

Conclusion

The NAHB/Wells Fargo HMI continues to provide early indications of current housing market conditions and has significant power to predict single-family housing starts and permits. The wealth of information contained in HMI is as useful in the post-2000 period as it was earlier in its two decade history.

____________________________

Footnotes:

[1] The NAHB/Wells Fargo Housing Market Index is strictly the product of NAHB Economics, and is not seen or influenced by any outside party prior to being released to the public. HMI tables can be accessed online at: nahb.org/hmi. More information regarding housing statistics is also available at housingeconomics.com.

[2] If all respondents answer “Good/High” then the index is 100. If all respondents answer “Poor/Low” then the index is 0. If equal numbers of respondents answer “Good/High” and “Poor/Low”, the index is 50. Any number over 50 indicates that more builders view sales conditions as good than poor.

[3] Mortgage Bankers Association’s (MBA) purchase applications index, university of Michigan’s question on consumers’ plans to purchase homes, and Conference Board’s question on consumers’ plans to purchase homes.

References:

  1. Builders’ Economic Council (BEC) Monthly Survey, National Association of Home Builders (NAHB), Economics, January 1985 – December 2006.
  2. Emrath, P., (1995), “Housing Market Index”, Housing Economics, June, pg. 11.
  3. Goodman, J. L., (1994), “Using Attitude Data to Forecast Housing Activity”, The Journal of Real Estate Research, Vol. 9 No. 4, pg. 445-453.