Where do the homeless come from?

Where Do The Homeless Come From?

THE EMERGENCE of a permanently homeless population in America has been viewed as one of the prime domestic failures of the Reagan Administration. While the stock market piles up "paper profits' and Wall Street waits for the next scandal, more and more people find themselves living in welfare hotels or on park benches. The free market--so the argument goes--obviously doesn't work for everyone.

At the same time, the effort to solve the problem promises a return to the budget-busting practices of the Sixties and Seventies. New York City already spends $300 million a year on the homeless, with little visible effect. Washington, D.C., is herding homeless families into welfare motels, where the city pays thousands of dollars a month in rents. On July 24 the President signed a two-year, $1-billion appropriation for aid to the homeless. The Homeless Persons Survival Act, currently before Congress, calls for spending up to $4 billion.

What is causing this destitution in the midst of plenty? There have been many explanations. Testifying before Congress last winter, Governor Mario Cuomo summed up the conventional wisdom. "The nature of our homeless population is largely misunderstood,' said Cuomo. "They are the people we all know in our families and our communities, homeless only because they are unemployed, because of chronic poverty, or simply because of the nationwide shortage of affordable housing.' Reductions in welfare and Social Security payments are sometimes touted as a contributing factor. Cutbacks in the federal public-housing budget are often blamed. The deinstitutionalization of mental patients is also frequently mentioned.

Self-evident as these explanations might seem, each has its problems. Unemployment reached a postwar high in 1982, just about the time homeless populations started to burgeon. Yet it has now dropped to its lowest level in 15 years with no visible reduction in homelessness. Poverty rates also rose in the early 1980s, but have since drifted downward. Still, the number of homeless rises.

Perhaps the most powerful argument against the Reagan Administration has been that the severe cutbacks in federally subsidized public-housing construction are causing homelessness. At first blush the figures seem compelling. From 1977 through 1981 the Federal Government authorized more than forty thousand units of new public housing per year. From 1982 through 1986, the average dropped to barely five thousand. Since homelessness appeared to rise dramatically in the early Eighties, these figures seem to make a prima-facie case against the austerity of the Reagan years.

The reality, however, is much more complicated. Public-housing construction is an on-going process. Units authorized in one year may not be completed for another six years or more. Here are the figures for the number of federally constructed housing units actually completed in each year since 1977:

Year Units Completed

1977 5,192

1978 10,134

1979 10,763

1980 15,109

1981 32,485

1982 27,611

1983 27,041

1984 24,058

1985 19,093

1986 14,897

In short, public-housing construction peaked during the very same years that homelessness was rising.

Thus, each of the standard arguments--unemployment, poverty, public-housing cutbacks--has its flaws. An argument that has always appealed to conservaties--the deinstitutionalization of mental patients--does not hold up very well either. Across the country, cities uniformly report that about 25 to 33 per cent of their homeless people are mentally ill. Thus deinstitutionalization may explain, at best, a third of the problem.

IN SEARCH OF AN EXPLANATION, I have done a study of homeless populations based on the Department of Housing and Urban Development's 1984 Report to the Secretary on the Homeless and Emergency Shelters.

The report contained estimates of the homeless population in forty large and medium-sized cities. Using 35 of these cities, and adding 15 more as a result of my own investigation, I calculated a homeless rate per thousand for fifty major American cities. Then, through the technique of regression analysis, I compared the homeless rates per thousand to seven conceivably relevant factors: poverty rates, unemployment rates, the availability of public housing, the size of the city itself, weather conditions, rental vacancy rates, and the presence or absence of rent control.

Regression analysis measures the coincidence of each independent variable (unemployment, etc.) with the dependent variable (homelessness). If any factor correlates with homelessness--if unemployment is high where homelessness is high, for example--we can infer some kind of relationship. Regression analysis--it should be emphasized--cannot prove cause and effect. It can only measure coincidence. Once some correlations have been discovered, however, we can theorize about what the causal connections might be.

Looking at the fifty cities (see table), ranked in order of homelessness per thousand, right away we see some interesting patterns. Miami has the highest rate in the country --a figure that may be fed by illegal Latin American immigration. St. Louis is also high, although this figure may be somewhat exaggerated by the city's narrow municipal boundaries. (See sidebar on methodology, "How Many Homeless?')

Beyond that, homelessness seems to be concentrated on the East and West Coasts. The highest concentrations are around San Francisco and Los Angeles (including Santa Monica) and in major cities of the Northeast. On the other hand, the "Rust Belt' cities of the Midwest and the "Oil Bust' cites of the South and Southwest do not have particularly high levels of homelessness.

Within this pattern there are some notable exceptions. While most Northeastern cities are high, Philadelphia and Baltimore are relatively low. On the West Coast, San Diego --the eighth-largest city in the nation, with a major influx of illegal Mexican immigrants--has less than one-third the homelessness of Los Angeles and San Francisco.

The median rate of homelessness for these fifty cities is 3.1 homeless per thousand--a rate shared by both San Diego and Salt Lake City. Nearly 70 per cent of the cities fall between 0.5 and 5.0 homeless per thousand, with only a handful of cities above this range.

Let us now begin by analyzing the standard explanations of homelessness to see how they hold up. The table on page 35 shows to what extent factors often considered possible explanations for homelessness apply to each city. Let's look first at one of the most common explanations --the poverty rate.

Obviously, most homeless people are poor people. But is poverty itself the deciding factor in driving people from their homes? If it is, then a regress analysis should show that differences in poverty rates among cities correlate with differences in rates of homelessness.

We are working with 1984 homelessness figures, but the latest figures identifying the percentage of poor people in each city are from the 1979 census. This is disappointing, but not a severe problem. Poverty rates generally to not fluctuate as rapidly as some of the other factors we will be looking at, such as unemployment.

The regression analysis proceeds by placing our fifty cities on a graph that plots homelessness against poverty. The distribution of cities (the Y's and N's on Graph 1) yields a "correlation coefficient,' the slope of which is represented by the line displayed on the graph. The maximum range of correlation is from 1.0 to -1.0. A correlation coefficient of 1.0 would mean homelessness occurs whenever poverty occurs; -1.0 would mean homelessness occurs only when poverty does not occur, both powerful relationships. A correlation coefficient of 0.0 would indicate that the relationship of poverty and homelessness is completely random. Numbers close to zero are "statistically insignificant.' That is, given the imprecision of the data, they are so close to complete randomness that they cannot be considered to have any explanatory power.

The correlaton coefficient of poverty and homelessness, according to our analysis, is .231. Squaring this figure, we get the percentage of the variation in homelessness that can be explained by poverty, about 5 per cent.

This figure is quite low, but might just be considered significant except for one other factor, which statisticians call the "P value.' Note the distribution of N's and Y's on the graph. In the case of a significant result, these points should cluster around the line, not necessarily closely, but noticeably. In this case they aren't clustering very noticeably. The P value is .107; it should be lower for a significant result. Given that the correlation coefficient is pretty low anyway, we can consider the result here to be not significant. This does not mean--it should be reiterated --that poverty does not ever cause homelessness, or that homeless people are not poor. It does mean that differences in homelessness among cities cannot be explained by the differences in their poverty rates.

How about another common explanation--unemployment? In this case we have 1984 figures. To save space I haven't displayed the graph this time, but plotting homelessness against unemployment rates yields a correlation coefficient of .121. Thus less than 2 per cent of the variation in homelessness can be explained by variations in unemployment, clearly a statistically insignificant result.

Another common explanation for homelessness has been the cuts in the federal budget for public housing. As already noted, this argument does not really hold up, since the completion of public-housing units has proceeded apace. Nonetheless, we can measure the impact that existing public housing is having on homelessness rates. But when we do the regression (the data on public housing are shown in the table on page 35; once again I will omit the graph for space), we find that the correlation coefficient is but .156, suggesting that there is a 2.4 per cent correlation between differences in the availability of public housing and differences in the homelessness rate, clearly insignificant. (And even that 2.4 per cent runs the wrong way--the more public housing, the more homelessness.)

Or how about the "big-city effect'? Shelter groups often report that, as a last measure of desperation, people who have lost their jobs or homes pick up and head for the nearest large city. Yet plotting homelessness against city size yields a cofficient of only .108, for a correlation of 1 per cent, again insignificant.

So far, out comparisons have revealed nothing positive. This could mean that the conventional explanations do not work. It could also mean that data on homelessness are so amorphous that nothing useful can be extracted from them. Until we are able to find some kind of correlation at some point, it will be difficult to tell one way or the other.

So let us try a much more prosaic explanation--the weather. After all, a number of Sun Belt politicians have argued that there has been a migration of the homeless to gentler climes. It thus makes sense to ask whether the wide variance in mean annual temperature among our fifty cities correlates with the variation in homelessness. Doing the regression we come up with a coefficient of .194, suggesting that the difference in mean annual temperature accounts for about 4 per cent of the variations in homelessness. This is a marginally significant result, but the P value is too high (.177). Though the possibilities were intriguing, in itself temperature looks insignificant.

One more factor that common sense would tell us might be having an effect on the number of homeless is the availability of private housing. (See sidebar, "Squeezing the Poor.') Since lower-income people almost always rent their homes, we will use the rental vacancy rate as the indicator. (See Graph 2.) Here we have a highly significant correlation. The coefficient is .387, suggesting that the vacancy rate accounts for 15 per cent of the variation--cities with lower vacancies having more homelessness. Put another way, a 1 per cent decline in the rental vacancy rate is roughly associated with a 10 per cent increase in homelessness in any city.

The vacancy effect suggests we should try one final factor: the presence or absence of rent control. Nine cities in the study--New York, Boston, Hartford, Washington, Newark, Yonkers, San Francisco, Los Angeles, and Santa Monica --practice some form of rent control. In Graph 3 (page 43), non-rent-controlled cities are aligned on the left and rent-controlled cities on the right.

Remarkably, rent control, with a correlation coefficient of .521, turns out to be the single most important factor for predicting homelessness. All by itself it accounts for 27 per cent of the variation among cities. The presence of rent control is associated with an increase in homelessness of 250 per cent.

It's also possible to do the regressions by combining variables--say, unemployment and temperature--to see if they amplify or detract from each other's significance. Sometimes a variable that is actually significant will, when run alone, appear insignificant; running it with another variable can filter out some statistical interference and show a truer picture. I reran every possible combination of the seven factors we've gone through as well as some others, but discovered only two noteworthy results. First, when temperature and rent control are run together, temperature becomes significant, explaining an additional 4 per cent of the variation in homelessness.

When we run rent control and vacancy together, an even more remarkable thing happens. The vacancy factor essentially disappears. Once rent control is factored into the equation, vacancy rates and nothing more.

A quick look at the vacancy rates given on the table on page 35 indicates why this should be true. The nine rent-controlled cities have the nine lowest vacancy rates. No city with rent control has a vacancy rate over 3.0 per cent, and no city without rent control, except Worcester, has a vacancy rate under 4.0 per cent. Here are the 11 cities with the lowest vacancy rates:

City Vacancy Rate

San Francisco 1.6

Santa Monica 1.8

Washington 2.0

Yonkers 2.1

Los Angeles 2.2

New York  2.2

Newark 2.3

Boston 2.6

Hartford 2.6

Worcester 3.0

Philadelphia 4.0

Thus the rent-control factor is the vacancy factor. The wide variations in vacancy rates among cities without rent control--from 4 per cent in Philadelphia to 18 per cent in New Orleans--do not affect homelessness. These variations apparently represent only normal fluctuations of the housing market. (See sidebar, "The Chicken or the Egg?' for an explanation of Worcester's low vacancies.) Only where rent control has been imposed do vacancies become pathologically low, with a consequent escalation in homelessness.

That rent control should be the most reliable indicator in predicting high levels of homelessness would come as no surprise to most economists. In a 1984 survey of more than two hundred economists, published in the American Economic Review, only 1.9 per cent disagreed with the statement: "A ceiling on rents reduces the quantity and quality of housing available.' This level of disagreement was the lowest among 27 issues surveyed.

Most laymen, however, still have trouble grasping the situation. When presented with New York's 300,000 abandoned apartments--the result of forty years of rent control --they have no trouble understanding that city's housing shortage. But what about San Francisco, Los Angeles, and Santa Monica? They are not the South Bronx, littered with burnt-out buildings. Why should they have a high level of homelessness? (Santa Monica, with probably the strictest rent-control ordinance in the country, is often called "The Homeless Capital of the West Coast.')

There are several reasons. First, builders unquestionably stop building once they confront rent control. After Los Angeles adopted rent control in 1979, for example, the city's newspapers were filled with stories of builders moving their operations to Arizona and Texas. Thus, new housing supplies diminish.

Cities often think they can avoid this consequence by leaving new construction exempt from the regulations. But this produces its own effects, which only exacerbate the situation.

To begin with, savvy builders are rarely fooled by "new-construction exemptions.' They know they will only be temporary. New York exempted all new construction when wartime rent control was made permanent in 1947. The exemption lasted until 1969, when tenant pressures led the city government to "recapture' (the city's own word) all postwar construction and bring it under "rent stabilization.' Once again, new buildings were left unregulated. This second exemption lasted only five years, until the city once again "recaptured' new construction. It is still possible to build new rental apartments outside rent control in New York, but few builders rise to the bait. By the same token, while Dallas, with a 16 per cent vacancy rate, added 11,000 new housing units last year, San Francisco, with one of the nation's worst housing shortages, added only two thousand.

But the loss of new construction, plus the deterioration of old housing, is still only one side of the equation. The other involves consumers' reactions to the regulations.

Typically, a city will leave some part of the market unregulated, subtly trying to shift the burdens to newer tenants. Thus, Boston and Los Angeles both practice "vacancy decontrol,' which permits vacated apartments to take a one-time jump to market values.

These gimmicks only channel pent-up demand into the unregulated sectors. As a result, rents in these sectors soon rise above their normal market levels. Anyone trying to move from one apartment to another will find himself giving up a below-market rent for an above-market rent. The difference can often amount to $600 a month for virtually identical apartments.

As a result, nobody moves. "Housing gridlock' develops, with people desperately hanging on to too-large or too-small apartments because they are cheap. Even when people leave town they hang on to the old apartment if they can, keeping it as an "office' or a pied-a-terre. (Paris gave us the term, after adopting rent control "temporarily' during World War I.)

Perhaps most important, housing gridlock blocks the process of "filtering,' whereby once-expensive housing eventually becomes available to the poor. Harlem, after all, was not built for the poor, or for blacks--it was built as luxury housing. With rent control, however, this filtering process grinds to a half. As Seymour Durst, Manhattan's 73-year-old developer-philosopher, puts it: "We've got plenty of low-income housing in New York. We've just got upper-income people living in it.'

All these behavioral changes add up to a "housing shortage.' But the effects of this shortage are not equally distributed. People with longstanding tenancy in regulated apartments will live year after year at remarkably low rents, proclaiming to one and all that rent control is a "great deal.' Others, however, will find almost nothing available.

It is these other people "at the margin'--those who have just arrived in the city or have been forced to abandon their old lodgings--who will bear the full brunt of rent control. From these marginal ranks--not surprisingly --has come the nation's homeless population.

THE MODEL for homelessness suggested by these figures, then, goes like this. In almost every American city there is a hard core of homeless people--about 3.1 people per 1,000. Probably one-third of these people would have been in mental institutions twenty years ago. The rest are victims of poverty, unemployment, alcoholism, drug addiction, and the vagaries of life. Among these people, there has been some migration to warmer climates. A few may be living in the street "by choice,' but the vast majority are not. These people need social services.

Truly widespread homelessness does not occur, however, until a city imposes rent control. At this point, all the people living in marginal circumstances suddenly find themselves confronting a "housing shortage.' Because of the way the benefits and adversities of rent control are distributed, the poorest and least capable tenants usually find themselves bearing the brunt of its ill effects. This pushes homelessness to pathological levels--about two and a half times what it would be without rent control.

Unfortunately, rent control has now been adopted by a number of major cities on the East and West Coasts. Although large numbers of tenants in these cities--perhaps even the majority--may reap some benefits, the adverse effects are being concentrated on a small but highly vulnerable minority. Unless these cities can be persuaded to give up rent control, the ranks of this minority--the homeless --will continue to grow.

Table: Homelessness and Some Factors Commonly Cited as Explanations

Table: Graph 1

Table: Graph 2

NB: The higher the vacancy rate, the lower the homelessness rate.

Table: Graph 3

Y's here and throughout indicate rent control; N's, no rent control. Here the Y's and N's are distributed on the extremes of the horizontal axis because Rent Control = 1 and No Rent Control = 0. The important point to note is the relative vertical distributions of the two groups.