Monday, June 27, 2011

Housing starts as a Pre-WW2 leading indicator

- by New Deal democrat

This is another post concerning leading indicators from before the inflationary era. Previously we saw that BAA corporate bond rates were one such indicator. Now let's look at housing starts.

Unfortunately the only data for housing starts before WW2 comes from the Statistical Abstract of the United States, and is annual. While that limits our ability to judge its effectiveness somewhat, the good news in that regard is that, as Professor Edward Leamer has shown, post-WW2 housing starts typically haven't had their maximum impact on the economy until a year or even 6 quarters out. Thus if we see that housing starts peak the year before a recession, and bottom the year before a recession ends, that is strong supportive evidence that Leamer's theory holds for the pre-WW2 era of deflationary busts.

It is also good to keep in mind that the 1920's housing boom was every bit the equal in population-adjusted terms as the boom of the first part of the last decade.

So here is the data (in 100's):

YearNonfarm housing startsRecession dates
1919 315 -
1920 247 1/20-
1921 449 - 7/21
1922 716 -
1923 371 5/23-
1924 893 -7/24
1925 937 -
1926849 10/26-
1927 810 -11/27
1928 753 -
1929 509 8/29-
1930 330 cont.
1931 254 cont.
1932 134 cont.
193393 -3/33
1934 126 -
1935 221 -
1936 319 -
1937 336 5/37-
1938 406 -6/38
1939 515 -
1940 602 -


Sure enough, while the data isn't perfect, housing starts generally peaked the year before the recession began, and bottomed the year before it ended. As to the 1927-28 recession, note that housing starts declined -88k in 1926, - 39k in 1927, and -57k in 1928, which on a second derivative basis conforms to the hypothesis. Similarly, while the 1937-38 recession took up almost half of each of 2 years, it is impossible to know if the hypothesis works, after increasing 98k in 1936, they only increased 17k in 1937 before accelerating by 70k in 1938, which likewise on a second derivative basis conforms to the hypothesis.

Note also that like our era, the 1920s housing bubble popped several years before the severe economy-wide downturn kicked in.

In other words, while the fact that the data is annual rather than monthly limits our ability to test it somewhat, there is strong evidence that housing starts are an equally valid leading indicator in eras of deflationary busts, as well as the post-WW2 inflationary era.