• We expect existing home sales to increase to between 5.10 million and 5.17 million units (SAAR) in July.
• This would represent an increase between 1.2 percent and 2.6 percent month-over-month.
• The change is driven primarily by continued market support from flat interest rates and modestly improving family incomes.

On Thursday, the National Association of Realtors (NAR) will release data on July 2014 existing home sales. We expect sales of single family residences, condos and coops to increase to 5.1 million units (SAAR) from 5.04 million units in June.

Background

Existing home sales totaled 5.040 million (SAAR) in June, up 2.6 percent from May, but 2.3 percent lower than a year earlier. Sales had been slowly but steadily increasing since late 2010 before dropping sharply in summer 2013 after interest rates rose. They began to recover in spring 2014, regaining about half of the decline recorded between July 2013 and March 2014.

Despite recent gains, there remain considerable doubts about the strength and durability of recent gains. Policymakers and market observers continue to point to sluggish home sales as one of the key restraints holding back a broader economic recovery.

Forecast

We use two models to develop our expectation for incoming existing home sales data:

• A “fundamentals” model than estimates home sales as a function of macroeconomic fundamentals. (More information about this model can be found in our earlier brief.)
• A “historical” model that estimates home sales as a function of past movements in existing home sales as well as recent pending home sales data.

Several additional assumptions are necessary for the “fundamentals” model, which are described in greater detail below.
Overall, we estimate that July 2014 existing home sales should total between 5.10 million (according to the “historical” model) and 5.17 million units (SAAR) (according to the “fundamentals” model)—an increase of between 1.2 percent and 2.6 percent respectively from June.

Explanation

Interest rates were essentially flat in July, providing a continued boost to home sales, but the gap relative to historical rates narrowed slightly. As this gap narrows further, some homeowners considering a move will opt not to since moving would imply assuming a mortgage loan with a higher interest rate. Income growth likely continued to accelerate—albeit from a low base—providing a further impulse to sales, although there is some uncertainty around this estimate.

Risks

The first, and most prominent in our view, risk to our estimate of existing home sales concerns the rate of change of median family income in recent months—a key input in our “fundamentals” model. Since data on median family income are currently only available through March 2014, we must extrapolate forward.

In our baseline forecast described above, we estimate that median family incomes grow at a pace equivalent to the median percentage point change in the annual growth rate (i.e., the change in the change) over the previous four months. Under this assumption, median family incomes would have grown by 4.1 percent in July 2014 relative to their year-earlier value. This is the assumption used in the estimate above. A more conservative assumption is that median family incomes grew in July 2014 and the same pace as in March 2014 (3.5 percent) while a more optimistic assumption is that median family incomes grew at a pace similar to the peak of the last business cycle (5.3 percent).

Using these estimates of median family income growth produces estimates of existing home sales ranging from 5.10 million units (up 1.2 percent) in the conservative scenario to 5.30 million units (up 5.2 percent) in the optimistic scenario.

A second risk concerns the homeownership rate. Our model holds the June 2014 homeownership rate constant at 64.8 percent. Given that month-to-month changes in the homeownership rate tend to be very small—averaging zero percentage points when calculated to the third decimal place over the period 1976 through 2013—this seems reasonable.

However, the homeownership rate has trended downward in recent years: Over the past two years, it has declined by an average of 0.03 percentage point each month. If this trend continues and the homeownership rate falls to 64.5 percent, and it we assume that median family incomes grow at 4.1 percent, our estimate of July existing home sales would be the same as described above. A change in the homeownership rate of this magnitude would not have a measurable effect on sales over the month.

Methodology

We use two models and several sets of assumptions to develop our expectation for incoming existing home sales data.

• The first model estimates home sales as a function of key macroeconomic fundamentals—including the homeownership rate, the homeowner vacancy rate, home sales prices, interest rates, family incomes and the total number of households.
• The second model estimates home sales as a function of historical movements in the time series as well as pending home sales data from recent months. More precisely, we estimate existing home sales using an ARIMAX model incorporating an ARIMA(2,1,0) model and four lags of pending home sales.

For the first model, assumptions are necessary since data for several of the input variables are not available through the most recent month. (Updated data are available for interest rates.) Below we outline the assumptions:

• For the homeownership rate and homeowner vacancy rate, we hold these values constant at their June 2014 levels. These two variables rarely exhibit large month-to-month jumps so this is likely a reliable if conservative assumption.
• For the five-year percent change in the median sales price of existing homes and the number of households, we use a best-fit ARIMA model over the series for 1976 through present to forecast one-month ahead. The models identified are an ARIMA(2,1,3) in the case of the median sales price series, and an ARIMA(0,2,1) in the case of the number of households. These methods yield estimates of 21.3 percent for median sales price (compared to 21.6 percent in June) and 115.230 million for the number of households (compared to 115.097 in June).
• For the one-year percent change in median family income series, there is considerable uncertainty given that the latest reported value is for March 2014. As a result, we must forecast a series that has shown considerable volatility in recent years four-months ahead. Given the uncertainty, we test and report three scenarios above—a scenario where the annual growth rate of family incomes flat-lines at its March 2014 pace, a scenario where family incomes grow at the median pace for the previous four months, and a scenario where family incomes grow at the fastest pace observed during the previous business cycle.