We’ve all woken up to it on a Saturday morning. The sound of the jackhammers and bulldozers concurrently destroying the building next to us in addition to our alarm-less wake ups. Construction seem to be everywhere in DC now and can range from your small neighbors row house projects all the way to massive scale projects like Capitol Crossing. Seemingly, the alarm-less wake ups will only continue as housing construction in DC is reaching its highest levels in over five decades. With no sign of the jack hammers abating, we wanted to see if there is any upside for our own home values while dealing with all this noise. In other words, at the current rate of growth for building permits in the District, will our home values see an increase as a result?
Building permit data is sometimes used as a key economic indicator to help determine future changes in home prices. By leveraging OCTOs published building permits issued in DC in the last five years, and created a map to show the construction hot zones across DCs zipcodes.
In taking a closer look at the data we may be able to help determine what future home prices will be, three to six months out from now. This question is bit tricky. While an increase in building permits can show potential increased home prices, other factors such as a boost in housing supply can offset price increases. To answer this question, we focused on just two zip codes within the District, and quantified not only the total numbers of permits pulled, but also the various types of permits.
The team took an initial look at the data by examining building permits and home closing prices from January 2011 through March 2015. In order to focus the results of the analysis, we looked at two economically different areas within DC — zip code 20003 which includes the Navy Yard and Capitol Hill areas, as well as zip code 20011 which includes the Brightwood, Crestwood, and Fort Totten areas. An initial analysis of the average home closing pries in both these zipcodes provide several interesting insights. First, both zip code areas appear to have a clear seasonal cycle to the data; this is not surprising of course. Generally, average sales prices tend be be lower in the winter months and higher in the summer months. These two zip codes are also different: there is a clear upward trend in sales prices in zip code 20011, while the sales prices in zip code 20003 appear to be relatively flat.
With an idea of the shifts in annual home prices for both zip codes, we wanted to get an idea of how the building permits in these neighborhoods changed on a yearly basis. We used 5 different building permit types: construction, postcard, home occupation, shop drawing, and supplemental. Building permits offer foresight into future real estate supply levels and their delay can often choke housing supply levels. Even with such promise, there several caveats to note here. While the data shows that a spike in permits can signal an increase in housing supply in the near future, other factors like supply can offset these gains like we noted above. That is, is that if more houses are built than consumers are willing to buy, housing prices are likely to decline and devalue the real estate market.
In this case we are looking at average sales prices for zip code 20003 and numbers of permits pulled for the 3 most common types of permit: supplemental, construction, and postcard. The first indication of a future increase in housing prices can be seen 6 months in advance, with an increase in both the number of construction and shop drawing permits. Intuitively this makes sense, as shop drawing permits are required to receive approval of shop drawings produced by contractors, suppliers, and fabricators before construction can begin. The quantities of these two types of building permits appear to be very strong 6-month leading indicators of the DC real estate market.
Another indication of increasing values comes 5 months in advance, with an increase in the number of postcard permits pulled for smaller home improvement projects such as electrical and plumbing work, porch and fence repairs, and interior demolitions — improvements that could be in preparation for selling a home. Four months in advance, we can see that an increase in supplemental building permits are strongly related to higher home values. These supplemental permits are required by licensed contractors in order to install supplemental systems in buildings such as electrical, plumbing, mechanical, elevator, and conveyor systems.
Correlation analysis on zip code 20003 above show strong 6-month leading economic indicators for housing price increases, but zip code 20011 appears to have even stronger leading indicators, but they occur only 4 months in advance. 4-month leading indicators are postcard and home occupation permits, and the 3-month leading indicators are postcard, supplemental, and construction permits. Although the analysis of zip codes is not identical, very similar patterns emerge in assessing the future real estate market 4-6 months in advance.
This showed the team one of the values of open data that is often overlooked. Public data can adds significant value in predicting changes in the market. Let’s say we want to make the best possible prediction of real estate values 3 months from now. Our testing models showed that 3 months from now, DC’s real estate market can experience 53% variable in housing prices in relation to building permit totals. That’s incredibly powerful, and consider the possibilities of supplementing this model with other data such as jobs report information, unemployment rates, and changes in neighborhood socio-economics, to name a few. The model results are shown below. The length of the bar indicates the statistical significance of the leading indicators, and the number represents the numerical impact on sales price. When we ran the same testing to see what housing prices would be 6 months out from now, each construction permit pulled will lead to a $1,708 increase in the average home price.
Based on the average price and most significant leading indicators for zip code 20011, let’s shift the data for the leading indicators several months into the future to see if the trends align. This would be further visual proof that building permit leading indicators are in-fact valid. The top chart shows sales price plotted with the Supplemental permits 3-months prior, while the bottom chart shows the Construction permits 6-months prior. The peaks align fairly well in July 2012, February/March 2013, July 2013, and February 2015. The data also similarly aligns through the many of the troughs in the real estate market.
In summary, things are looking pretty promising when it comes to using open data about building permits to predict changes in DC’s real estate market. When we set out to write the article, our thinking was that we could use building permits from distinct points in time to guess future home sale prices 6 months out. While the results look hopeful off the bat, there’s still much left to look at before we can say for sure how to predict our markets future. Perhaps some of our more data science inclined readers would like to help out? Reach out to us on twitter, facebook, or by emailing TheNinja@district.ninja
Our hypothesis was that different types of permits, pulled at distinct points in time, could combine to provide a robust leading indicator of changes in real estate values in the District.
1. In order to make predictions moving forward, say 6 months from now, the results below show that the model can explain at least 30% of the variability in housing prices 6 months from now. Of course the model isn’t quite as good because we don’t have as much information available, but it’s still quite powerful even with only using building permit data and no other economic data.
2. More sophisticated analysis can be done to more accurately predict future sales prices by isolating the impact of not only building permits pulled in previous months, but also home sales prices in previous months.
3. After ‘de-seasonalizing’ the data to reduce the seasonal noise in the data, we examined all the correlations of sales prices with the numbers of permits pulled in the previous 6 months. The correlation matrix is shown below for zip code 20003, which shows the magnitude of positive relationships between the variables.
4. In order to identify any noticeable relationships between home sale prices and permit data, we plotted subsets of that information. By applying more advanced econometric techniques, one can create a model to isolate the individual impact of each variable, and the combined effect of variables, to changes in price. A standard model was applied to the used zip codes.
5. DC building permit data, which was pulled from DC’s Department of Consumer and Regulatory Affairs public website.
6. The following charts were pulled from the article since they did not clearly show whether there is a relationship between the numbers of permits pulled in previous months with sales prices.