Measuring seasonality in new housing starts

Andrey Kamenov

Andrey Kamenov, Ph.D. Probability and Statistics

The second half of the year traditionally sees more new housing starts. In fact, in this post, we’ll see that seasonality here is somewhat more complex. Is your job is related to the construction industry? Then you most likely have to take this into account.

The U.S. Census Bureau provides microdata for its Characteristics of New Housing survey. Any user would probably encounter several technical difficulties using the data. The most important is sampling the data. According to our estimates, it slightly reduces the effect of seasonality, but luckily it still allows us to compare different samples.

Another issue to consider is the difference in the total number of days in each month. Of course, it’s really only a major issue in February, where it may account for a 10 percent dip alone. But to be consistent, we adjust the monthly rates as if each month had exactly 30 days in it. This way, we may take at least one factor out of the equation and focus on deeper trends in the observed numbers.

To analyze the data, we use the X-13-ARIMA-SEATS algorithm developed by the Census Bureau. Many researchers consider this technique to be state-of-the-art for general seasonality detection problems.

Let’s see where this approach takes us. Click on the chart below to see the larger interactive version pop up.

New housing starts by month

Now, while it may be quite interesting to scroll through the numbers, it still doesn’t say much about any factors that come into play here.

Seasonality in new housing starts by division

It also appears that some states have more pronounced seasonal fluctuations. Judging by the standard deviation of monthly distributions, the Midwest is on the top of the list. The northeastern states also have higher-than-average numbers. The fluctuations here are at least twice as large as those in the southern states and on the West Coast.

Let's see if we can find the underlying cause. It appears that the major factor contributing to the difference we see on the map is related to the housing category. While the numbers from May to October are generally higher across all divisions, one particular type of construction projects stand out here.

Here's the chart which focuses only on the three divisions with the highest numbers on the map (30 percent or more). It shows the number new housing starts in each category.

Seasonality by category

Seasonality by housing category

It appears that in these three divisions there is a significant peak in the number of houses built for sale. Many of these are started between June and November. The highest numbers are observed in October.

Source(s): 

Discuss this article on our forum with over 1,900,000 registered members.

About Andrey Kamenov

Andrey Kamenov

Andrey Kamenov, Ph.D. Probability and Statistics

Andrey Kamenov is a data scientist working for Advameg Inc. His background includes teaching statistics, stochastic processes and financial mathematics in Moscow State University and working for a hedge fund. His academic interests range from statistical data analysis to optimal stopping theory. Andrey also enjoys his hobbies of photography, reading and powerlifting.

Other posts by Andrey Kamenov:

Leave a Reply

Your email address will not be published. Required fields are marked *