Lall, Wang and da Mata find that reducing land use regulation in Brazil would increase the formal housing supply, but not necessarily reduce slum formation

I just finished a two day visit to Sao Paolo, where officials are concerned that constrained housing supply is exacerbating the size of favelas.  This lead me to read Lall, Wang and da Mata:

In this paper, we examine the effects of land use and zoning regulations on
housing supply and slum formation across Brazilian cities between 1980 and 2000. We
find very low price elasticities of housing supply in the Brazilian formal housing market,
which limits formal housing supply adjustments in response to demand increases, and is
linked with growth of informal settlements. The imputed Brazilian formal housing supply
elasticity is similar to those in Malaysia and South Korea, which have been regarded to
have restrictive regulatory environments.

We also find that land use regulations that manage densities – in particular,
minimum lot size regulations, have important effects in terms of housing supply and slum
formation. Contrary to conventional wisdom, lowering minimum lot size regulations do
not lead to a reduction in slum formation. If city population growth were exogenous and
households did not consider local regulations in their migration and residential location
decisions, then lowering minimum lot sizes would allow cities to accommodate more
residents into formal housing developments – and unambiguously reduce slum formation.
However, when we consider that regulations are a part of household migration
and residential choice decisions, the exact effects of lowering regulatory standards are not
obvious. In fact, our model suggests that the net effect of land regulations depends on the
extent to which new formal housing supply absorbs demand, both from current informal
sector residents and population growth induced by lowering regulations. Our estimation
strategy considers both effects, and we find that cities that lowered minimum lot size
regulations from the Federal stipulation of 125 m2 experienced higher growth in the
formal housing stock. However, this was also accompanied by higher population growth
from migration, and the resulting city population growth was higher than the formal
housing supply response, exacerbating the slum formation problem.

Local innovations that increase access to land for the poor – such as flexible land
sub divisions – are welfare enhancing as they allow houses with different specifications
to be available in the market, thereby allowing low income residents to benefit from
services that meet their preferences (and affordability). However, if some cities offer
improved access to land compared to their peers, these cities are likely to
disproportionately attract (poor) migrants. If the induced population growth is higher than
formal housing supply adjustment, informality is likely to grow. Cities that absorb
migrants increase welfare – and in this context the challenge is to identify strategies that
increase formal housing supply relative to population growth. The econometric results
should not be viewed as a failure of flexible zoning to reduce slum formation. Rather, the
focus should be on identifying pre existing distortions in the land and housing market that
reduce the formal housing supply response to additional demand. 

Two important issues for future research emerge from this analysis. First, to
understand why some cities deviated from the federally stipulated minimum lot
specification of 125 m2 to favor low income housing development. And second, to
identify sources of land and housing supply distortions that reduce the elasticity of formal
housing supply.

One fact about the Sao Paulo context: according to officials, the vacancy rate in the formal sector there is about 13 percent, suggesting excess housing supply.  But the construction cost of the minimum size house allowed under Brazilian law is simply too high for many Brazilians to afford.  Allowing smaller houses to be constructed would almost certainly improve housing welfare.