NAHB analysis of the most recent Quarterly Sales by Price and Financing published by the U.S. Census Bureau reveals that conventional loans financed the most new home sales in a quarter since 2007. The share of sales backed by conventional loans in the third quarter was 71.8%–the largest share since Q4 2018, but well below the housing boom peak of 91.7% (Q4 2006).
FHA-backed sales made up 17.4% of new home sales in the third quarter of 2020, a three-percentage-point decline, quarter-over-quarter. Likewise, the share of VA-backed sales fell two percentage points to 6.2%. The number (11,000) and share (4.6%) of cash sales rose slightly.
The combined loss of FHA and VA market share—five percentage points—coincided with a five-percentage-point gain in the share conventional loan sales. As FHA and VA market share increases, conventional loan market share typically falls and vice versa.
On a four-quarter moving average (MA) basis, the shares of VA-backed and cash sales are at or near post-Great Recession lows. Conventional loan share was the only category to see an increase (MA), climbing one percentage point to 69.4%.
Although cash sales make up a small portion of new home sales, they constitute a larger share of existing home sales. According to estimates from the National Association of Realtors, 18% of existing home transactions were all-cash sales in September 2020, unchanged from August but up from 17% in September 2019.
Different sources of financing also serve distinct market segments, which is revealed in part by the median new home price associated with each. In the second quarter, the national median sales price of a new home was $324,900. Split by types of financing, the median prices of new homes financed with conventional loans, FHA loans, VA loans, and cash were $349,700, $252,900, $343,100, and $307,800, respectively.
It is worth adopting some caution associated with the Census data as they are estimates based on sample statistics. The statistical error associated with the FHA, cash, and VA sales estimates from this data set are relatively high, meaning that although the data are presented as single numbers, the true values may be substantially different.