Testing Hypotheses About Rent Control
(Executive Summary)
Commissioned by Canada Mortgage and Housing Corporation
written by Frank T. Denton, Christine H. Feaver, R. Andrew Muller, A. Leslie Robb, and Byron G. Spencer
October, 1993
The full 2cm-thick statistical report may be purchased from Canada Mortgage and Housing Corporation, Ottawa
This project was carried out under the terms of a research contract with Canada Mortgage and Housing Corporation.
The project involved the formal statistical testing of a range of hypotheses about the effects of rent controls on the
rental housing market. The original statement of the hypotheses, as prepared by CMHC, was as follows:
Rent
Over the long run, rent regulations have no impact on rents. Rent regulations act to restrict rents from sharply
increasing during periods of strong economic growth, but also inhibit sharp rent decreases in recessionary periods.
Housing Starts
There are no significant differences in the responsiveness of rental supply with respect to vacancy rates and rents,
and with respect to changes in vacancy rates and rents.
Vacany Rates
Rent regulations are associated with lower vacancy rates, other things being equal.
Property Values
Rent regulations decrease the relative attractiveness of investment in rental housing. This is reflected in the fact
that percentage changes in rental property values are smaller in regulated markets than in unregulated markets.
Tenure Preferences
Under rent regulations, the relative financial advantages of homeownership are lower. This increases
the preference for renting.
Conversions
Rent regulations encourage conversions from rental to owner-occupied housing (particularly condominiums)
Maintenance and Repairs
Rent regulations discourage maintenance and repairs, and reduce the services, (e.g., cleaning,)
landlords provide to tenants.
We undertook a careful search for data that would be suitable for testing these hypotheses. It is a
common problem in applied statistical work that the data that are available are not ideally suited to the
task at hand. The curent project was no exception. It was not possible to test any hypothesis about property
values, and some recasting of the hypotheses relating to tenure preferences and conversions was required.
In addition, it was necessary to substitute apartment unit starts and vacancy rates for rental unit starts and
vacancy rates. Aside from those necessary changes, appropriate tests of the hypotheses proposed by CMHC
were designed and executed, and the results interpreted. In spite of the limitations of the data we believe that
the results of the estimation provide informative tests of most of the hypotheses.
The tests were of two types. The first was based on standard "parametric" procedures, as they are
termed in the statistical and econometric literature. The second type was "nonparametric." The two
types of tests led essentially to the same conclusions, thus making it possible to view those conclusions
with a greater degree of confidence than would otherwise have been the case.
The data used in the study came from a variety of sources. They included metropolitan area time
series for use in some of the tests and provincial time series for use in others. An important element of
the study was a careful survey of the rent control regulations that have been in effect in the ten provinces
over the period 1971-93. Based on that survey, a taxonomy of control regulations was developed. The
taxonomy was used to classify each province/year to one of three categories: (A) no rent regulation;
(B) rent control with review (mandatory); and (C) rent arbitration (voluntary). (The taxonomy provided also
some additional detail, as described in the main report.) The subsequent analysis then sought to determine
the effects of regulations of types B and C on the rate of increase of rents, the responsiveness of starts of
apartment units to vacancy rates and rents, the level of vacancy rates in the rental market, the proportion
of households that rent, the rates of conversion of single housing units into multiple units, and the
proportion of rental units in need of major repairs.
The major conclusion of the study was that there appears to be no convincing evidence that
rent regulations, as they existed in various provinces in Canada from the early 1970s through to the
early 1990s, had significant effects on rents, on the construction of rental units, or on vacancy rates.
A more detailed statement of the more important findings follows:
The data that we have used and the tests that we have carried out provide no evidence to suggest
that rent controls reduce the rate of increase of rents in the long run.
There is some evidence to suggest that controls cause rents to rise more rapidly than they
would otherwise in periods when the rental market is "soft," especially when the control regime is type B.
If the long-run rate of increase is unaffected, that would imply that under controls rent increases
must be less rapid in periods of market "tightness." However, there is no significant evidence that that is
the case. We are inclined therefore to discount the evidence of "soft" period effects, and emphasize the
finding of no long-run effects. Given the practical difficulties of defining market "softeness" and "tightness"
with precision, the latter seems to us the more credible finding.
There is no convincing evidence that the responsiveness of rental unit starts to vacancy rates
or rents is reduced by the imposition of controls.
The formal statistical evidence suggests that type B controls tend to be associated with lower
vacancy rates. However, there is no evidence of that for type C controls, of when types B and C are
combined into a single rent control regime. Given the practical statistical difficulties in trying to
distinguish between the effects of a particular type of regime, one the one hand, and the characteristics
of the provinces in which the regime is in force, on the other, we are inclined to emphasize the lack of
effects when the distinction between B and C is ignored, and to conclude that overall there is no
strong evidence that controls are accompanied by reduced vacancy rates.
There is some evidence that controls are associated with a higher proportion of renter households.
However, that should not be taken simply as evidence that household preferences for renting are
increased, although that may be the case; the proportion of renters depends also on the availability of
rental units. The evidence is present for type B controls and for types B and C combined, but not for
type C alone. In light of our scepticism about the reliability of distinctions between the two types we
would again attach greater emphasis to the evidence obtained whe the two are combined.
There is no evidence that rent controls affect the rate of conversion of single housing units into multiple units.
There is no evidence that rent controls increase the proportion of occupied rental units
that are in need of major repairs.
|
