New working paper: The topography of growth and decline – Local price dynamics within cities

My latest working paper researches within-city home price dynamics in bullish and bearish residential real estate markets. It contributes to the literature addressing the urban layout of cities by formulating and empirically testing a novel idea for changes in the price gradients across neighborhoods under different market regimes. It finds that the combination of city-wide falling home prices and declining population numbers hurts low-value neighborhoods most, while falling home-prices in cities with robust demographics do not lead to shifts in the within-city distribution of housing wealth. In addition, the paper confirms earlier findings on endogenous home price dynamics.

A new level of spatial detailedness is achieved by combining a high-quality data-set for The Netherlands with estimation techniques borrowed from the geoscience domain. The data comprise of 1.8 million single family home transactions and 0.8 million apartments sold by members of the Dutch Realtor Association (NVM). This is the first paper that estimates home price index surfaces for an entire country based on a spatial error model (SER).

Price index for single-family homes in West Berlin

Next year, Berkeley’s Ashok Bardok, Bob Edelstein and Cynthia Kroll will publish an around-the-world analysis of the current housing crisis. Together with Piet Eichholtz, I had the honor to contribute a chapter on the big outlier in European housing markets: Germany. More on this soon.

In the meantime, I cannot refrain from publishing the first hedonic price index for single-family homes in West Berlin:

Home values take deep dive in West-Berlin
The blue line shows the prices for single-family homes in West Berlin, controlling for quality differences in the homes, while the red line represents home prices in real terms (without inflation). The dashed green line is the path of the Consumer Price Index (CPI).

Based on a data set of single-family homes we estimate a standard hedonic price equation spanning the years 1978 through 2007. The graph above suggests that the Berlin housing market has experienced roughly three phases for the 30 years covered by the index. The first period, between 1978 and 1989, can be characterized by steady growth. The average house price increase over this time period was 1.9 percent (corrected for inflation).

The exuberance in the aftermath of the fall of the Wall fueled price increases, peaking in 1994 with nominal prices being 47 percent above 1989 values. The average annual price increase for this period was 8.1 percent in nominal terms, and 4.6 percent in real terms.

In 1995, house prices decreased slightly followed by a much larger drop in 1996. In 2007, house prices were back at 1989 levels – in nominal terms. In real terms, however, prices had plummeted to 55 percent of their 1994 values, and 84 percent of their 1978 values. Such an extended period of price losses is unprecedented among European capitals (see figure below).

Berlin home prices compared to price dynamics in selected European capitals

Google Streetview, do it yourself.

Google Street view started in 20 German cities today – my home was blurred, however. Oh German privacy paranoia, what a strange  beast you are. I can only recommend Jeff Jarvis’ essay on this phenomenon.

I went downstairs, took a picture and uploaded it at several place. Lets see how long it takes Google to find the picture and to present it next to their blurred view.


Geocoding with QGIS and Google Maps

Translating large numbers of unstructured addresses into exact geographic locations (coordinates) was my competitive edge: Write a little script in Perl that feeds the Google Maps API with (unstructured) addresses, retrieve and store Google’s replies – et voila: Let the GIS analysis begin.

Obviously, I could have asked commercial geocoding services to provide me the locations of my data points. There are some major drawbacks when using “professionals” that I do not like:

  • They are expensive. Actually, I admire them for charging fees  for a service that can be accomplished with free tools in the market.
  • The results are like meatballs: You do not really know what ingredients went into it. Relying on Google maps at least allows a graphical sanity check.
  • They are slow. Imagine, you really really need this extra layer of information on a Friday afternoon to go on with your research… When doing it yourself, you can have the answer within minutes.

The good news is: No scripting is needed anymore to do free and fast geo-coding. The open source GIS software QGIS (it is free and runs on all operating systems!) now has a plugin that does all the interaction with Google’s web service. It is still a bit rudimentary, but, hey, it does its job.

Just check out Steven’s great article on how to get it running on your machine.

How much risk must one be willing to take when starting up a business?

An academic super-hero of mine pointed out a recent paper by Robert Hall and Susan Woodward in the last issue of the American Economic Review. The article is titled “The Burden of the Nondiversifiable Risk of Entrepreneurship” (link here).

One key result is that a “typical venture-backed entrepreneur received an average of $5.8 million in exit cash”. Sounds good, doesn’t it? But read on: “… Almost three-quarters of entrepreneurs receive nothing at exit and a few receive over a billion dollars”. Want to reconsider your lofty start-up ideas?