Does the US unemployment rate affect the housing market?
An Exploratory Data Analysis
Introduction
Austin has ranked the top growing cities in the U.S. for consecutive years. Living in Austin for almost a decade, I have witnessed an increasing trend and a strong job market, which drives up the housing price. In 2020, the housing market remains hot despite the pandemic. To prove this trend is universal, I am interested in how the unemployment rate affects the home value and the rental price in the top U.S. metro areas. My first goal for this project is to confirm that the low unemployment rate leads to the growth in home value and rental price. The second objective is to see if the assumption still holds due to COVID-19.
Data Preparation
In my analysis, I used home value and rental price data from Zillow. The data is published on the Zillow research page and has been smoothed, seasonally adjusted. It contains the monthly housing price and the rental price for the top metro areas since 2014. The National Associations of Realtors publishes real estate statistics as well. However, its public data does not offer the level of granularity (i.e., data for each metro city) I need for this analysis. In addition, Zillow’s database is actively maintained by a group of economists and data scientists.
The unemployment rate data is published by the US Bureau of Labor Statistics. The data has been seasonally adjusted. It contains monthly employment, unemployment and the civilian labor force statistics by each US metro area. The data provides the unemployment rate information since 1990.
I have done the following steps to clean the individual datasets and combine them into a final dataset for this analysis.
· Limit the sample size to top 105 US metro areas
· Data type transformation (converting object/float” to “datetime”)
· Data cleaning and reshaping
· Create categorical columns: To find patterns in the data, I classified the 105 metro areas into four regions based on geographic locations — Northeast, Midwest, South, and West. I also labeled cities as ‘Top 25 city.’ Finally, I grouped unemployment rate into low, high, and medium based on quartiles.
Table 1 shows the statistics summary of the final dataset. Table 2 Is the summary which includes the categorical data.
Data Analysis
To test my first hypothesis, I calculated the correlation between unemployment rate, home value, and rental price for the top 100 U.S. metro areas. The data includes observations from January 2014 to August 2020. The correlation heatmap (Figure 1) indicates a positive relationship between home value and rental price with the correlation coefficient close to 1. On the contrary, the heatmap suggests a weak relationship between the unemployment rate vs. home value and the unemployment rate vs. rental price, with a correlation coefficient at -0.091 and -0.063.
To further investigate the relationship between housing price and the unemployment rate, I created a Seaborn pairplot which illustrates the rental price, home value, and unemployment by regions. We see a widely distributed home price and rental price in the West region, followed by the Northeast (see appendix). The South and Midwest, on the contrary, have a relatively dense distribution on housing prices. According to Figure 2, it appears that there is no strong correlation between the unemployment rate and the housing prices, especially for South and Midwest.
I also grouped the final dataset based on the unemployment rate. ‘1’ being the lower quartile, 3 being the upper quartile, and 2 being in the middle. As we can see from Figure 3, we still do not observe a pattern between the unemployment rate and home value, or between the unemployment rate and the rental price.
To test my second hypothesis, which is if the low unemployment rate leads to an increase in home value and rental price since the COVID-19 outbreak, I created a subset of data for 2020 and calculated the correlation for the variables. Again, the correlation heatmap (Figure 4) confirms the positive relationship between the home price and rental value but suggests a weak relationship between unemployment rate vs. home value and the unemployment rate vs. rental price. The pairplot (Figure 5) displays data points in 2020 vs. time period prior to 2020. We saw a distinction of unemployment rate distribution for 2020 and the years before 2020. Nevertheless, the distributions of home price and rental price for both groups are identical.
Finally, if we look at cities in the U.S. individually rather than in aggregate (Figure 6.1), we will find a strong negative relationship between the unemployment rate and the home price and rental price. Subset data of New York City, Los Angeles, and Chicago suggests that rental price increases when unemployment rate decreases (-0.99 correlation coefficient for New York). Similarly, home value grows when the unemployment rate declines (-0.92 correlation coefficient for New York).
Conclusion
The findings above rejected my hypothesis that the unemployment rate is one of the major driving forces for home value and rental price. Pearson’s Correlation analysis concluded that there is no strong correlation between the unemployment rate and the housing value. Although individual cities (e.g., New York, Los Angeles, Chicago) show a strong correlation between unemployment and housing prices, the correlation does not hold when the data is aggregated. Simpson’s paradox could possibly explain the contradictory results. We may not account for other variables that could significantly impact the rising home value in this study.
My hypothesis that a rising unemployment rate could negatively affect home value and rental price does not hold in 2020. Despite the record-high unemployment rate, the average home price in 2020 is still higher than the average home price from 2014–2019.
Due to data availability, this study focused on a single independent variable — the unemployment rate. More factors could contribute to the change in home value. Listed below are factors to consider for future studies.
Demographics: age, income, and population growth of a certain area could affect the demand and the price of its local real estate market.
Interest rate: the low-interest rate could accelerate home purchases. For example, in November 2020, the mortgage rate hit the 12th record low of the year, which favors home buying.
Home inventory: we also need to consider the demand and supply. If there are fewer homes available in the market, the price will go up when the demand remains steady.
More historical data: the dataset contains data from 2014–2020, representing a strong economy since the 2009 financial crisis. Data before 2008 could further validate the relationship between the unemployment rate and the home price.
References
https://ftp.cs.ucla.edu/pub/stat_ser/r414.pdf
2. National Association of Realtors
https://www.nar.realtor/research-and-statistics
