Exploring Income and Climate Effects on Home Prices

July 24, 2023July 22, 2023 ryan.renkey

Over the past few years, I’ve come across numerous headlines about the booming housing markets in different US cities. Austin and Boise have especially captured the spotlight, and it seems like every major metro in the South has been featured in articles highlighting the best places to invest in a home. This got me wondering about the factors driving regional price differences between these metro areas and whether we can predict how long these trends will continue.

Analyzing income as a factor in regional price variation

One significant factor that affects home price appreciation in most metro areas is the local income level. In areas with higher incomes, residents can afford to drive up the prices of real estate more comfortably than in metros with lower incomes. While this makes sense intuitively, I haven’t come across a graphic that compares home prices to income levels, so I decided to do it myself. To do this, I collected income information using the latest per capita income figures from the BEA and house prices from the Zillow Housing Value Index for June 2023. My analysis covered 50 of the largest metro areas in the United States, and I conducted a linear regression to determine the influence of income alone on the variation in home prices.

The model reveals a strong and expected correlation between home prices and income, suggesting a suitable linear relationship. By using income as a variable, we can account for approximately 70% of the variation in housing prices across metro areas. However, this approach doesn’t work as well for certain cities, like Chicago, where the predicted home price of $503,000 based on income is significantly higher (57%) than the actual home price of $320,000. I observed similar discrepancies for other cities, such as Buffalo and Pittsburgh, where the predicted home prices were higher than expected, while cities like LA and San Diego had lower predictions compared to their income levels.

Looking at the list of cities, it’s evident that climate may contribute to the price differences. California’s famous Mediterranean climate, characterized by mild and pleasant weather, stands in contrast to the harsh winters and humid summers in the Midwest and Great Lakes regions. Clearly, the local climate plays a role in determining home prices. To objectively compare these diverse climates, I needed a simple measure. Considering factors like sunshine days, rainfall frequency and severity, heat, humidity, and winter severity, among others, seemed relevant, but using too many variables would complicate the model unnecessarily. Therefore, I settled on a single measure: degree days.

What are degree days?

Degree days are a measure of how much the outdoor temperature deviates from a standard base temperature, usually 18°C or 65°F. They are used to estimate the energy demand for heating and cooling buildings, as well as to compare the climate of different locations. You can calculate degree days by subtracting the base temperature from the average daily temperature, and then summing up the results over a period of time. There are two types of degree days: heating degree days (HDD) and cooling degree days (CDD). HDD accumulate when the average daily temperature is below the base temperature, indicating a need for heating. CDD are positive when the average daily temperature is above the base temperature, indicating a need for cooling. For example, if the average daily temperature is 50°F, then the HDD for that day is 65 – 50 = 15, and the CDD is 0. If the average daily temperature is 77°F, then the HDD for that day is 0, and the CDD is 77 – 65 = 12.

Here are the degree days for some US cities along with income data. To find degree days for other US cities you can use this tool from HUD

City	Home Price	HDD	CDD	Total Degree Days	Per Capita Income (2021)	2yr Income CAGR
Los Angeles	940,093	928	1506	2434	$72,637	7.8%
San Francisco	1,199,118	2597	163	2760	$123,711	10.4%
Miami	535,228	141	4090	4231	$73,522	7.5%
Austin	492,046	1509	2982	4491	$71,372	6.0%
Atlanta	379,933	2827	1810	4637	$63,219	6.1%
Seattle	742,571	4615	192	4807	$89,274	7.1%
Las Vegas	419,573	2239	3214	5453	$58,276	6.9%
Washington DC	572,170	4055	1531	5586	$80,822	5.2%
New York	630,382	4754	1151	5905	$85,136	6.6%
Boston	691,525	5630	777	6407	$92,290	6.7%
Salt Lake City	552,637	5631	1066	6697	$61,551	8.1%
Chicago	320,726	6630	702	7332	$71,992	6.9%

Using both degree days and income to explain housing prices

To test my theory, I conducted two regression tests. In the first test, I solely considered the total degree days to determine how much of the variation in home prices could be attributed to this factor and whether it was statistically significant. The results of this test confirmed that degree days indeed have statistical significance. However, it only accounted for about 32% of the variance, and with a 95% confidence interval, it showed that home prices decreased by an estimated range of $61 to $148 for each additional degree day. Since we already established that income was statistically significant, I proceeded to perform a regression with both income and degree days as variables.

In this last regression, where both income and degree days were considered, the model performs exceptionally, explaining a substantial 85% of the housing price variance between metro areas using the adjusted R-squared method. Additionally, both income and degree days showed statistical significance. Now, let’s examine some of the predicted home prices and how they compare to actual homes.

Observations and insights

SF prices might not be as crazy as they initially seem, especially when you take into account the sky-high salaries in the area. Moreover, factoring in the weather could make a compelling case that home prices are actually undervalued!
Considering that degree days explain half of the remaining variance not accounted for by income, only around 15% of the variance might be attributed to other factors. While many articles emphasize the tax benefits of different locations, it’s possible that these factors have less impact on home prices than commonly assumed.
As for large coastal cities like NY, DC, and Boston, it’s unlikely that their price appreciation will surpass the national average. Their home prices are close to or exceed the predicted values that account for income and degree days and the rise of remote work has allowed employees more flexibility in choosing where they live, making expensive cities less attractive when they could relocate and still maintain their salaries elsewhere.
If we assume that all degree days hold equal value, we can anticipate substantial price appreciation in the South and in lower-priced mid-sized cities in the Midwest, such as St. Louis, Indianapolis, and Pittsburgh.
The relatively high salaries in Miami and Austin may serve to sustain the current real estate booms in each city.
The predicted home values indicate that most cities west of Denver might be overpriced. One potential reason for this could be that the model solely relies on income and degree days, overlooking crucial climate and location information that is consistent across much of the West. For instance, the West generally boasts more abundant sunshine and better access to nature compared to the East Coast.
I have yet to test the model by removing home price outliers like SF or LA. When dealing with a small number of outliers, a regression model may attempt to fit them, even if these outliers are influenced by different factors. For example, if SF prices are significantly influenced by foreign buyers, it wouldn’t be sensible to use the same model for SF as we would for a city like Oklahoma City, where foreign buyers are not a significant factor.