Friday, January 30, 2009

Graph: U.S. housing prices, 1890-2008

This graph shows U.S. housing prices for the first quarter of each year. 2009's first quarter numbers should be released four months from now.

Source: My Housing Graphs Website.

10 comments:

  1. This graph should probably be on a percentage scale or log scale. It makes the recent run up look larger than it was, though it was still quite large!

    ReplyDelete
  2. On a percentage scale, it would look exactly the same. You would just have different units listed on the side.

    A log scale would be irrelevant, because you should be looking at the red line, not the blue line. If you used a log scale, it would screw up the red (inflation-adjusted) line.

    ReplyDelete
  3. I am trying to source housing prices back to 1890, can you give me any ideas for replicating the data series you show here?

    Thanks

    ReplyDelete
  4. The data in that graph comes from Yale economics professor Robert Shiller. See http://www.irrationalexuberance.com

    You may also want to read his book, Irrational Exuberance, 2nd Ed.

    ReplyDelete
  5. I really appreciate your post. It gives an outstanding idea that is very helpful for all the people on the web. Thanks for sharing this information and I’ll love to read your next post too.
    Regards:
    optiontips.in

    ReplyDelete
  6. I'm interested in comparing long term price appreciation for low, modest, expensive properties.  Any suggestions? 

    ReplyDelete
  7. i disagree.  show us the new graph and let us decide.

    ReplyDelete
  8. James, I love your data - thanks for sharing.  I noticed a discrepancy in the above graph and your house price graph from 1970 - today.  The inflation-adjusted house prices (red line) in the graph entitled United States House Prices, going from 1970- 2011, starts at about $150K and tops at about $275K.  In the 1890-today graph it's consistently about $25K less.  Explanation?  I'm guessing you're using a different normalizing factor in both graphs, but I look forward to your response.

    ReplyDelete
  9. Chris, there are several reasons for it. First notice that the blue lines match each other better than the red lines, although they are still slightly different. The blue lines represent nominal values. The reason they are slightly different is because I'm taking the latest median home price from the National Association of Realtors (NAR) and discounting it into the past. NAR price data is notoriously volatile from one quarter to the next, while index data from S&P, FHFA, & Freddie Mac are fairly smooth. So, the volatile end point causes the entire blue line to jump around a little bit every time I update my graph to match the latest NAR median home price. Ideally I would use a less volatile data source, but all data sources have their quirks.

    Second, the red line just takes the blue line and adjusts it for inflation. The graph above is almost three years old, while the graph on my housing graph website is current. Since the U.S. has experienced inflation over the past three years, that slightly changes the values on the red line. Also, I am using different measures of inflation in the graph above than in the one on my housing graph website. The graph above uses the CPI (officially called the CPI-U), because the CPI has data available back far into history. The graph on my housing graph website uses a version of the CPI called the CPI-U Research Series Using Current Methods (CPI-U-RS). The CPI-U-RS only goes back to 1977. Prior to 1977, I use the CPI - All Items Less Shelter on my housing graph website.

    Now, why do I use different measures of inflation? Well, the graph above is directly based on data from Yale University economist Robert Shiller. (See http://www.irrationalexuberance.com.) He used the regular CPI-U, probably because that's the only measure of inflation that is available going back to 1890, so that's what I used in that particular graph. Using the regular CPI-U has problems, though. The way the CPI-U is measured has been improved over time. Prior to 1983 it counted home prices as part of inflation. In 1983 it switched to measuring "owner-equivalent rent", i.e. the price an owner could rent his house for. So in effect, prior to 1983 you are taking the change in home prices and partially dividing it by the change in home prices. The CPI-U-RS uses a consistent way of measuring inflation over time. The CPI-U-RS uses the current method of measuring inflation and applies it back to 1977. So, by using the CPI-U-RS I can avoid partially dividing the change in home prices by the change in home prices for the 1977-1983 period. I wish the CPI-U-RS were available back to 1970, but it's not. To avoid partially dividing the change in home prices by the change in home prices for 1970-1977, I use a version of the CPI that omits shelter (including home prices, rents, and hotel room prices) for that period.

    I hope that explanation made sense. In short, I think the graphs on my housing graph website are more accurate than the graph above.

    ReplyDelete
  10. I should also point out that every time the new values for the S&P/Case-Shiller Home Price Index are released, they make small corrections to their historical data at the same time. Since the graph above is almost three years out-of-date, that might have some effect, but I'm not sure how much.

    ReplyDelete