Federal Reserve Economic Data

The FRED Blog has discussed the relationship between single-family housing starts and completions and also how changes in overall housing market prices are measured. Today we build on the topic of housing by comparing trends in the type of residential constructions erected.

The FRED graph above shows data from the US Census and the US Department of Housing and Urban Development (HUD) on the number of new, privately owned, completed housing units since 1968. There are three size types: single-family buildings (the blue area), buildings with 2 to 4 separate dwellings (the red area), and buildings with 5 or more dwellings (the green area).

The data are shown in a stacked area graph to highlight the relative amounts of each type of housing structure. We also changed the data frequency from monthly to annual to observe the trends more easily. So, what does the graph show?

The number of single-family structures, as a proportion of all types of housing structures, is clearly larger than the number of multifamily structures. This might reflect a preference for single-family housing, but we can’t say for sure because the data do not capture the exact number of individual dwellings in large multi-unit housing.

However, the data do show a trend in the construction of multifamily buildings with 2 to 4 units. That housing trend even has a name: the “missing middle.”

The term was coined to reflect the fact that construction of small-scale and affordable multifamily dwellings has decreased over time. “Middle” housing surged in the early 1970s during a boom in the overall construction of multifamily housing. During this period, nearly half of all new homes were multifamily. This type of housing became relatively less and less popular—as revealed by the shrinking red area in the graph.

Recent research coauthored by Raphael Bostic from the Atlanta Fed notes that small and medium multifamily properties, defined as buildings with 2 to 49 units, comprise over 20% of the US housing stock. This housing segment contains the largest percentage of the lowest-income households and the majority of rental units across the country. You can learn more about this topic here.

How this graph was created: Search the alphabetical list of FRED releases for “New Residential Construction” and select “Table 5. New Privately-Owned Housing Units Completed.” Select the three series naming the number of units per structure and click “Add to Graph.” Use the “Format” tab to change the graph type to “Area” and the stacking option to “Percent.”

Suggested by Zach Wallace-Wright and Diego Mendez-Carbajo.

The Federal Reserve began a tightening cycle in March 2022 by increasing the so-called federal funds rate. But much earlier, in November 2021, Fed officials announced the end of their extremely accommodative policy during the pandemic. With this announcement, financial markets immediately priced-in an increase in the cost of borrowing in anticipation of a future higher policy rate path to combat inflation.

The FRED graph above shows three different ways of capturing the impact that tighter credit conditions have on the financial burden of servicing mortgage debt. This burden is usually calculated in terms of a household’s interest payments for a mortgage as a fraction of that household’s disposable personal income. These data series tend to move with interest rate changes, following the path of inflation shown by the purple dashed line, which is the year-over-year percent change in the consumer price index for all urban consumers (CPI), excluding food and energy, often referred to as CPI-Core.

The green line in the FRED graph shows US mortgage debt service payments as a percent of disposable personal income, as reported by the Board of Governors of the Federal Reserve System. Notice that, since the pandemic, it has been relatively flat, around 4%—which is 300 basis points lower than it was in 2007. Despite the increase in interest rates, disposable incomes have grown at the same speed and the financial burden according to this measure has remained unchanged.

The red line calculates a similar ratio using the interest rate for a 30-year fixed-rate mortgage, outstanding one-to-four-family residential mortgage balances, and the same series for personal income as for the green line. The path of this red line, until 2021, is very similar to the green line’s. But one can see the impact of the policy tightening as the increase in interest rates and outstanding balances offsets the increase in disposable income. Around 2021, this ratio was about 2%, but recent policy tightening has increased mortgage rates to around 7%, causing the ratio to more than double since 2021.

What would the costs have been if rates had not risen? The blue line shows the calculation of a hypothetical counterfactual of the previous constructed variable shown by the red line: Here, the interest rate does not change over time, but is set to a constant value of 3%. (This is the level of mortgage rates in the summer and fall of 2021.) You can see that this series remains flat, below 2%, since the growth in personal disposable income is roughly the same as outstanding mortgage balances. Considering this rough “back of the envelope” calculation, one can argue that household costs of servicing new loans are double what they would have been without recent monetary policy tightening.

Despite these higher costs, why do we still see a very strong economy and high levels of consumption spending? The historical series of mortgage debt service payments as a percentage of disposable income indicates that current mortgage-servicing costs are very low compared with the levels in 2006-08—in fact, they’re nearly half the size.

How this graph was created: Search FRED for the “consumer price index for all urban consumers excluding food and energy” and change the units from “Index” to “Percent change from a year ago.” Add the rest of the series and perform the calculations as follows: Click “Edit Graph,” open the “Add Line” tab, and search for “Households and Nonprofit Organizations; One-to-Four-Family Residential Mortgages; Liability, Level (Billions)”; then click “Add.” Now search for and add “Personal Income (Billions)” and the 30-year fixed-rate series, which shows as percent. Each of the series is ordered alphabetically: a, b, c, and d. To calculate our new variable of mortgage burden, in the formula box type ((b*(d/100))/c)*100. The numerator and the denominator are in billions, and the interest rate is transformed from 5% to 0.05 by dividing by 100. The measure is shown in percent terms, which is why the formula multiplies the ratio by 100. The counterfactual—represented by the blue line, with a fixed mortgage rate—is an exercise in manipulating variables using a constant: The modified formula replaces this variable by a 3/100 constant value, which uses a modified formula of ((b*(3/100))/c)*100.

Suggested by Carlos Garriga.

More than half of the data in FRED can be displayed in choropleth maps, which is a type of data visualization where geographical areas are colored differently according to the range of their data values. The contours of the geographical areas in FRED maps represent political and statistical boundaries. Political boundaries shape the counties, states, and nations we’re familiar with. But what is a statistical boundary? Who draws them? How did those boundaries come to be?

Statistical boundaries are groupings of smaller geographies, such as counties, or states, that are drawn by data collection organizations like the US Census and the US Bureau of Economic Analysis (BEA) to facilitate the presentation and analysis of statistical and economic information.

The FRED map above shows data on the number of people residing in the US during 2022. The BEA organizes them into eight regions: New England, Mideast, Great Lakes, Plains, Southeast, Southwest, Rocky Mountain, and Far West.

Each region represents groupings of states with similar economic and social conditions. This classification was drawn after the 1950 Census to group states with similar economic and social indicators. You can learn more about the process of drawing BEA regions here.

Economic and social conditions change over time, so certain statistical boundaries are periodically revised. Research can be used to draw alternative boundaries for regional economic data geographies. You can read about an alternative definition of economic regions here.

Where can you go next using FRED maps? Read this short essay to learn more about data geographies in FRED.

How this map was created: In FRED, search for “Resident Population in the Plains BEA Region.” Click on “View Map.”

Suggested by Diego Mendez-Carbajo.