Forecasting Multifamily Housing Needs

By Walt A. Nelson |

The rental housing market is an attractive investment alternative. Apartments are essentially a standardized, mass-market product, and rents and asset values provide a useful, if imprecise, hedge against inflation.

While recent articles in national and trade publications have predicted a significant increase in apartment construction, you will want to assess market supply and demand before recommending rental housing to investor-clients. Completing even a preliminary market analysis can be a challenge, however, because of the overwhelming quantity and underwhelming quality of relevant data and because collecting new data is not cost-effective. Despite these problems, it is still possible to make useful estimates of demand for multifamily housing using available data.

Using Census Bureau data published in the 1993 Statistical Abstract of the United States, you can create a preliminary market analysis, including forecasts of new construction, through the year 2004. Then, using both the household cohort and apartment ratio methods, you can examine trends in the housing market based on economic conditions and related household data. When you're done, you will have a five-year forecast of multifamily housing needs in the United States.

A word about using Census Bureau data: reading the Statistical Abstract of the United States requires patience, a calculator, and some practical knowledge of the housing market. Just when it seems that data lead to a certain conclusion, it turns out that the Census Bureau compiled data related to a crucial variable in a slightly different manner from one table to the next. You will find that cross-tabulations of certain variables are unavailable in print form but may be obtained on CD-ROM versions of the Abstract. The most usesful data are found in the population and contruction sections. Throughout this article I refer to tables in the 1993 Abstract so that you may make quick reference to the appropriate statistics if necessary.

A Market Overview
Let's start by examining select statistics that describe the American housing market and are found in the 1993 Statistical Abstract of the United States. According to Table 65 of the Abstract, there were about 95.6 million households in the United States. The average size of each household was 2.62 persons. These 95.6 million households occupied about 104 million units of various types, with an overall vacancy rate between 6% and 8%. About 33.3 million dwellings were rental units and about one-third of all tenants lived in multifamily housing structures with five or more units. (Vacant and seasonal units do not appear on the table, so the totals do not add up to 104 million.)

The Household Cohort Method
The household cohort method allows you to estimate apartment demand by calculating the rate of household formation within the age groups that most often rent apartments. The following analysis assumes that new households form primarily within the group or cohort of individuals ages 18 to 24 and secondarily within the group or cohort of individuals ages 25 to 34. (The data used in the following analysis come from the 1993 Abstract, Tables 71 and 72.)

In 1992, there were 25.919 million individuals ages 18 to 24 and 42.463 million individuals ages 25 to 34. The 18 to 24 year olds comprise a seven-year cohort starting with the year 1992; this means that, on average, about 3.702 million persons each year will enter the housing market for the first time. The older group spans a 10-year time period beginning in 1992, which works out to 4.246 million persons entering the market each year. Approximately 56% of 18 to 24 year olds live with relatives rather than on their own, another 23% are married, and about 21% live alone or with non-relatives. In addition, about 45% of 25 to 34 year olds purchase homes, so the analysis assumes these households vacate apartments (instead of simply leaving their parents' home for the first time). These vacated units constitute a supply source and reduce the need for new construction.

Table 2 forecasts a need for 247,000 new apartment units per year for the period 1995 through 1999. Of course, this forecast is simplified, ignoring death rates, net migration, and the interaction between the other age groups and the rental housing market. It also assumes that rates of new household formation will remain the same as in the recent past.

Sophisticated advisers and investors may wish to go beyond the basic approach I have presented. You can build a more elaborate forecast, again using the 1993 Abstract, because it breaks down population by single-year cohorts from ages 1 to 100. When conducting such an analysis you'll find it necessary to build a spreadsheet that incorporates rates of marriage, first domicile, and home ownership for each single-year age group as it advances through a particular time period.

For example, using the household cohort method to forecast new apartment construction for the 10-year period from 1995 to 2004 requires an examination of housing needs for those ages 8 to 17, ages 18 to 24, and ages 25 to 34. By the year 2004 the 8 year olds will be age 17, the 18 year olds will be age 27, and the 25 year olds will be age 34. To complete this process, factor in the number of apartments lost to fires, natural disasters, or razing. If you assumed even a 1% depreciation rate each year, the amount would be an annual loss of 330,000 units.

The Apartment Ratio Method
Another way to forecast the need for new construction is to compare the annual ratio of new apartments to the number of new households formed during a recent period of time. This method assumes that the ratio remains fairly stable over the years. Rather than using housing starts, this method uses the Census Bureau estimate of completed structures having five or more units. (Not all units under construction are actually completed in the given year, while others are either never completed or become owner occupied upon completion.)

Table 3 forecasts that new construction of apartment structures having five or more units will range from 228,000 units completed in 1995 to slightly more than 255,000 units in 2004. This forecast is based on a combination of facts and assumptions:

• Fact: The annual growth rate of households in the United States has ranged from 1.25% to 1.5% during the 1980s and early 1990s. According to the data in Table 65 of the Abstract, there were 93.3 million households in the United States in 1990 and 95.6 million in 1992. Dividing 95.6 million by 93.3 million gives us 1.0247 million, indicating a two-year growth rate of about 2.5%, which works out to 1.25% per year. In order to make the forecast, the 95.6 million figure will be the base figure for the number of households and the annual growth rate of 1.25% will be restated as 1.0125 million.
• Assumption: The future household growth rate will remain the same. Next, multiply 95.6 million households by 1.0125, which gives us 96.795 million households for 1993. Multiply 96.795 by 1.0125 to get 98.004 million for 1994; multiply that by 1.0125, to get 99,230 million for 1995. (See the top of Table 3.)
• Fact: The ratio of new apartments completed in structures of five or more units to total households has been slightly less than one-quarter of 1% since 1980. The number of households in 1980 was 80.776 million, according to Table 70 of the Abstract. Table 1230 of the Abstract reports that 196,100 apartment structures (each having five or more units) were completed in 1980. The ratio of completions to households in 1980 is 196,100 divided by 80,776,000, which equals 0.0024. To convert this to a percentage, multiply it by 100, which gives us 0.24%. A similar process for the year 1990 gives us a ratio of 0.23%.
• Assumption: This ratio (0.23%) of new apartment completions compared to total U.S. households will remain the same. Since the 0.23% figure is the more conservative of the two, it is used throughout the calculations in Table 3.
• Fact: Projecting total apartments (two- to four-family and multifamily housing combined) is more difficult because completions data for two- to four-family dwellings are not available in the Abstract. However, Table 1221 of the Abstract does show that since 1985, two- to four-unit starts have ranged from 14% to 26% of multifamily (five or more unit) starts. Starts in the two- to four-family category were about 93,000 in 1985 but have declined to about 31,000 units started annually from 1990 to 1992.
• Assumption: Completions of two- to four-family dwellings will be 10% of multifamily completions each year. We will assume that the ratio of two- to four-family starts to multifamily completions is stable. We will also assume that all two- to four-family units started will be completed.

The forecast for total annual apartment units completed during the years 1995 through 2004 appears in the right-hand column of Table 3. For purposes of clarification, see the 1995 forecast calculation.

Testing the Methods
It is imperative to test any forecast against recent data to see how well the method "predicts" the past. Testing these two methods against apartment construction data for the 1980s confirms the measure of accuracy for each approach. The household cohort method estimates an average annual need for about 522,000 units (multifamily and two- to four-unit structures combined) for the 1980s. Table 1221 of the 1993 Abstract shows that the actual average total starts per year during the 1980s was 487,000. It appears that the household cohort method may overestimate actual starts by 5% or 6% on average.

The apartment ratio method predicts an average need for 209,000 multifamily units during the 1980s. Completions for two- to four-unit structures are not available in the Abstract. Table 1230 of the Abstract shows that actual average annual multifamily completions were 256,000 from 1980 to 1991. It appears the apartment ratio method may understate multifamily completions by about 20% on average.

Using both the household cohort method and the apartment ratio method allows the advisor or investor to make estimates that should bracket the actual outcome. Tables 2 and 3 of this article show that both methods estimate annual apartment construction to be around 250,000 units for the next five to nine years.

Uncovering Market Trends
The forecast of 250,000 new apartments per year is based on a statistical analysis of past data. From the standpoint of forecasting future demand, it provides a certain minimal level of satisfaction. How-ever, we can use related data from the Abstract to develop guidelines to direct future placement of investment funds.

Declines in Home Ownership- For example, home ownership rates have been declining within the cohort of 25 to 34 year olds. In 1980, 51.6% of this group owned homes; in 1990 this proportion dropped to about 45%, in spite of the fact that interest rates had declined markedly during that time.

Table 4 offers some telling statistics that, in combination with other information, may indicate a continued decline in the rates of home ownership. During the past two decades home prices have outpaced inflation. While median home prices rose by a factor of 3.2 (or 320%) from 1976 to 1992, the cost of other goods and services rose by a factor of about 2.5 (the cost of living index was 53 in 1975 and 136 in 1991, according to the U.S. Bureau of Labor Statistics). Meanwhile, down payments as a portion of home price have dropped and the average age of first-time home buyers has increased.

Preference for Large Apartments- Vacancy rates for large apartments and rental houses have remained low and stable over the past 10 years, as shown in Table 5. Recent reports in the Wall Street Journal indicate that consumers with one child or no children prefer much larger homes than in the past.

Moreover, in 1992 about 1.2 million women between ages 30 and 44 gave birth (Abstract, Tables 103 and 104). Almost 60% of these births occurred in households having annual incomes higher than $35,000. Table 1237 of the Abstract implies that perhaps one-third of these high-income households rent rather than own. The table show thatas of 1990, renters comprised 33.8% of all surveyed households ages 35 to 44. The 1980 figure for the same group was only 26.8%. The Abstract does not provide a further breakdown by income category.

An Aging Housing Stock- In spite of at least two market cycles where significant new apartment construction took place, the average age of the rental housing stock has been increasing since 1970. In 1990 multifamily units were an average of about 20 years old, whereas in 1970 they averaged 17.7 years. The average age of rural rental housing (one- to four-unit structures) has remained at about 43 years during the same time period.

Guidelines for Apartment Investors
What can we conclude from this study? The rate of home ownership among young people, especially in the 25- to 34-year-old group, has declined as a result of the combined effects of higher interest rates, tough loan qualification standards, higher home prices, and lack of job stability. As households within this age cohort have children, vacate smaller, older apartments, and defer the purchase of the traditional single-family detached house, they will need more and better living space in some type of multifamily arrangement. As more households move to smaller urban (formerly rural) areas, they are likely to find a shortage of suitable rental housing. And while federal tax policy is driving construction activity in the low-income housing market on the one hand, the 1986 passive activity rules still inhibit investment in the middle-class apartment market. Under these conditions, a forecast of new apartment demand in the range of 250,000 per year for the next five years may actually turn out to be an understatement.