Consonant with sentiment indices built from the University of Michigan and Conference Board surveys of consumers, we refer to our indicator as the Home Purchase Sentiment Index (HPSI).
We started with a long list of candidate questions to include in the HPSI. We chose 28 questions as preliminary candidates on the basis of data availability, of a priori judgments, of coverage of important aspects of housing markets, and because of their simple correlations with the four housing market outcome measures during 2010-14.
Nonetheless, in light of common practice and the relatively short history of the NHS, we were inclined to select about five questions for the HPSI.
From the 10 questions in Table 1, on the basis of judgment and statistical evidence, we ultimately selected the six questions in Table 3 for the HPSI. We sought to include questions that would reflect the factors generally judged to drive home purchases: incomes, interest rates, and house prices--whether experienced or expected.
Nonetheless, because we judged that job-loss concerns would be quantitatively important and fairly reliable signals for understanding and forecasting housing market outcomes, we chose to include Q112B in the HPSI.
Figure 1 shows the HPSI (thick black line) and the net percent positive responses to its six component questions for March 2011-December 2014.
Indeed, Q20B was the only HPSI component that significantly improved forecasts for all eight housing market outcomes in Table 2.
Given its six component questions, how did we calculate the HPSI? Simply: The HPSI is the average of the net percent positive responses of each of its six component questions.
The HPSI is an "equally-weighted" index in that we take a simple, rather than a weighted, average of the net percent positives.
We also eschewed seasonal adjustments to the HPSI and to its component questions.
Nonetheless, the components contribute very unequally to the equally-weighted HPSI. Relative contributions can be measured by relative volatility.