Shelter from the Storm: Upgrading Housing Infrastructure in Latin American Slums

The program

The TECHO program provides basic, pre-fabricated, transitional houses to extremely poor families living in informal settlements (slums) in Latin America regardless of whether or not they own the land on which they live. The aim of this program is to increase the well-being of these families. The program was started 19 years ago in Chile and now works in 19 Latin American countries. This NGO has built almost 100,000 houses with the help of an army of volunteers. Every year, more than 30,000 youths throughout Latin America volunteer to work with TECHO. We evaluate the program in three Latin American countries: El Salvador, Uruguay, and Mexico. The locations of the settlements in El Salvador are somewhat different than the sites in the other two countries. In El Salvador, TECHO works in poor areas scattered throughout the country, but not in the country's main urban center of San Salvador. In contrast, the TECHO intervention sites are concentrated closer to the largest urban centers in the other two countries. In Mexico, this includes slums in Estado de Mexico located adjacent to Mexico City and, in Uruguay, slums located in and around Montevideo.

TECHO targets the poorest informal settlements and, within these settlements, households that are lodged in very substandard dwellings. TECHO serves "irregular settlements," which are de fined as communities in which a majority of the families are living on plots of land that they do not own. These settlements are plagued by a host of problems, including insufficient access to basic utilities (water, electricity and sanitation), signifi cant levels of soil and water contamination, and overcrowding. The typical housing units in these informal settlements are no better than the surrounding dwellings, as they are rudimentary units constructed from discarded materials such as
cardboard, tin and plastic, have dirt floors and lack connections to basic utilities such as water supply and sewerage systems. The

TECHO housing units are 18 square meters (6m by 3m) in size. The walls are made of pre-fabricated, insulated pinewood or aluminum panels, and the roofs are made of tin to keep occupants warm and protect them from humidity, insects, and rain. Floors are built on top of 15 stacks that raise them up to between 30 and 80 centimeters of the ground in order to reduce dampness and protect occupants from floods and infestations. Although these houses are a major improvement over the recipients' previous dwellings, the amenities that they offer are limited, as they do not include a bathroom or kitchen or plumbing, drinking water hook-ups or gas connections. The houses are designed to be low in cost and easy to construct; they can be placed on a plot of land next to an existing house or as a new unit
that replaces the existing one. Units are modular and portable, can be built with simple tools, and are set up by volunteers working in squads of from 4 to 8 members. The cost of a TECHO house is less than US$1,000 -- with the bulk of the cost being accounted for by the acquisition, storage and transportation of the building materials, since there are essentially no labor costs. The bene ficiary family contributes 10% of that amount (around US$100). In El Salvador, US$100 is approximately equivalent to 3.3 months' per capita baseline earnings, while in Mexico and Uruguay, it is roughly equivalent to 1.6 and 1.4 months, respectively. Importantly, added to the fact that the TECHO house is heavily subsidized, there are no exact substitutes of TECHO houses on the market that households could be investing in incrementally. TECHO do not offer the house in the market and only offer it to a group of selected slum dwellers that are in the poorest conditions within slums. Hence, even if households did not face credit constrains to get access to housing improvements, they could not get access to TECHO houses neither in the form nor at the price offered by TECHO. This consideration is relevant for interpreting the results of our study. Finally, the houses are also easy to disassemble and move to a new location. It is important for the houses to be movable because most of the families in these makeshift settlements do not have formal title to the land that they live on. TECHO managers were concerned that upgrading the value of the land by building permanent housing might induce both public and private owners to try to force residents to move in order to reclaim the improved land. However, making the housing mobile does away with that
incentive.

2007-09-01

2012-04-30

Number of Rooms

Share of Rooms with Good Quality Floors

Share of Rooms with Good Quality Walls

Share of Rooms with Good Quality Roofs

Share of Rooms with Windows

Sink on Room where food is prepared

On-site water supply

Electricity connection inside the house

Use gas or kerosene stove to cook

House with own toilet

Satisfaction with Floors Quality

Satisfaction with Walls Quality

Satisfaction with Roofs Quality

Satisfaction with House protection against rain

Satisfaction with Quality of Life

Safe inside the House during the last 12 months

Safe leaving the house alone during the last 12 months

Safe leaving the kids in the house alone during the last 12 months

The house has been robbed during the last 12 months

TV

Fan

Kitchen or Gas Stove

Refrigerator

Bicycle

Household Size

Newborns (<1)

Newborns (<2)

Adults (>18)

Monthly Income Per Capita

Hours worked last week by head of household

Hours worked last week by spouse of head of household

If the children below 5 years old had diarrhea episodes during the last 4 weeks.

If the children below 5 years old had respiratory diseases during the last 4 weeks.

Number of Rooms: Number of rooms in the terrain (observed by the enumerator).

Share of Rooms with Good Quality Floors: Proportion of rooms with floors made of good quality materials like cement, brick, or wood (observed by the enumerator).

Share of Rooms with Good Quality Walls: Proportion of rooms with walls made of good quality materials like wood, cement, brick or zinc metal (observed by the enumerator).

Share of Rooms with Good Quality Roofs: Proportion of rooms with roofs made of good quality materials like cement, brick, tile and zinc metal (observed by the enumerator).

Share of Rooms with Windows: Proportion of rooms with at least 1 window (observed by the enumerator)

Sink on Room where food is prepared: Indicator equal to one if there is a sink inside the room where food is prepared (observed by the enumerator).

On-site water supply: Indicator equal to one if there is access to drinkable or not drinkable water in the terrain where the house is located (observed by the enumerator).

Electricity connection inside the house: Indicator equal to one if there is a formal or informal connection to the electricity system inside the house (observed by the enumerator).

Use gas or kerosene stove to cook: Indicator equal to one if the household reports the use of gas stove or kerosene to cook.

House with own toilet: Indicator equal to one if there is a toilet inside or outside the house, but inside the terrain (observed by the enumerator).

Satisfaction with Floors Quality: Indicator equal to one if the respondent reports being "satisfi ed" or "very satisfi ed" with the quality of floors, measured by a Likert scale of 4 categories that goes from "unsatisfi ed" to "very satis fied".

Satisfaction with Walls Quality: Indicator equal to one if the respondent reports being "satisfi ed" or "very satisfi ed" with the quality of walls, measured by a Likert scale of 4 categories that goes from "unsatis fied" to "very satisfi ed".

Satisfaction with Roofs Quality: Indicator equal to one if the respondent reports being "satis fied" or "very satisfi ed" with the quality of roofs, measured by a Likert scale of 4 categories that goes from "unsatis fied" to "very satis fied".

Satisfaction with House protection against rain: Indicator equal to one if respondent reports being "satisfi ed" or "very satisfi ed" with the house's protection against water when it rains, measured by a Likert scale of 4 categories that goes from "unsatisfi ed" to "very satisfi ed".

Satisfaction with Quality of Life: Indicator equal to one if respondent reports being "satisfi ed" or "very satis fied" with the quality of life of her family in that house, measured by a Likert scale of 4 categories that goes from "unsatisfi ed" to "very satis fied".

Safe inside the House during the last 12 months: Indicator equal to one if respondent has never or rarely felt unsafe inside the house during the last 12 months, measured by a Likert scale of 5 categories that goes from "never unsafe" to "always unsafe".

Safe leaving the house alone during the last 12 months: Indicator equal to one if respondent has never or rarely felt unsafe leaving the house alone during the last 12 months.

Safe leaving the kids in the house alone during the last 12 months: Indicator equal to one if respondent feels safe or very safe leaving the kids alone in the house during the last 12 months, measured by a Likert scale of 5 categories that goes from "never unsafe" to "always
unsafe".

The house has been robbed during the last 12 months: Indicator equal to one if respondent reports the house has been robbed during the last 12 months.

TV: Indicator equal to one if the enumerator observes and the household reports having a television.

Fan: Indicator equal to one if the enumerator observes and the household reports having a fan

Kitchen or Gas Stove: Indicator equal to one if the enumerator observes and the household reports having kitchen or gas stove.

Refrigerator: Indicator equal to one if the enumerator observes and the household reports having a refrigerator.

Bicycle: Indicator equal to one if the enumerator observes and the household reports having a bicycle.

Household Size: number of members residing in the house.

Newborns (<1): number of member below 1 years old residing in the house.

Newborns (<2): number of members below 2 years old residing in the house.

Adults (>18): number of members above 18 years old residing in the house.

Monthly Income Per Capita: Monthly Income per capita in US dollars of July 2007. It is calculated as the sum of the monthly earnings of each household's member divided by the household size.

Hours worked last week by head of household: Number of hours worked by the head of household at main and secondary job during the last week, conditioned on having worked during the last week.

Hours worked last week by spouse of head of household: Number of hours worked by the spouse of head of household at main and secondary job during the last week, conditioned on having worked during the last week.

If the children below 5 years old had diarrhea episodes during the last 4 weeks: Indicator equal to one if the mother reports that a child below 5 years old had diarrhea in the last four weeks.

If the children below 5 years old had respiratory diseases during the last 4 weeks: Indicator equal to one if the mother reports that a child below 5 years old had respiratory diseases in the last four weeks.

Experimental design

The TECHO programs budget and personnel constraints limit the number of housing units that can be built at any one time. Under these constraints, TECHO opted to select benefi ciaries through a lottery system that gives all eligible households in a pre-determined geographical area an equal opportunity to receive the housing upgrade in a given year. We exploit this experimental variability to assess the impact of improved housing conditions. TECHO first selected a set of eligible settlements and then conducted a census to identify eligible households within each settlement (i.e., those poor enough to be given priority). The eligible households were surveyed (baseline
survey) and then randomly assigned to treatment and control groups, within each settlement. Note that this is a household-level randomized experiment, where treatment and control units are co-residents within each selected slum. Each slum is considered as one strata, so we first stratify the sample by slum, and then within each slum we randomly assign households to treatment and control groups.

In order to obtain truthful information from the households and to avoid creating any desirability bias in the treatment group, the data collection work was separated from the implementation of the intervention by contracting a highly respected survey fi rm in each country. The enumerators told the people whom they interviewed that they were collecting data for a study on living conditions and did not make any reference to the TECHO program either verbally or in written form. After randomization, treatment households were told about the program and its requirements by TECHO officials. Some of them accepted the program and some rejected it. Note that the control households were not told that they would receive the TECHO houses in the future, and so their behavior should not have been
affected by the expectation of being treated in the next round, although they could have felt frustrated when they realized that they had lost the lottery.

Baseline surveys were conducted approximately one month before the start of the construction work in each settlement. Since the TECHO program did not have the capacity to work in all settlements at once, the program was rolled out in each country in two phases, and the follow-up surveys were therefore conducted in all settlements at the same time, i.e, between 15 and 27 months after the construction work. The latter generated a variation in the time of exposure to the treatment across treatment households. All the surveys included modules on socioeconomic characteristics, the labor market, assets, security, health and self-reported measures of satisfaction.

Our sample includes 23 settlements in El Salvador, 39 settlements in Mexico and 12 in Uruguay. The total number of eligible households in these settlements was 2,373, with the total being split more or less evenly across the three countries. Treatment was offered to 60% of the households in El Salvador, 51% in Mexico and 61% in Uruguay. In all, over 85% of the households in the intention-to-treat groups complied with the treatment assignment (the remaining 15% were unable to afford the required 10% copayment and hence did not receive a house), while the compliance rates for the non-intention-to-treat groups were practically perfect. Finally, we attempted to track all of the households that migrated out of the study settlements, but could fi nd and interview only a fraction of them. Attrition rates from the sample were between 5.5% and 7% in the intention-to-treat group and between 6.3% and 8.7% in the non-intention-to-treat group. Though the attrition rates are about one percentage point higher in the non-intention-to-treat group in all three countries, the differences are not statistically signifi cant at conventional levels. Finally, note that both non-compliance and attrition rates are pretty much the same across country samples; thus, potential differences in the causal effects should not be attributed to treatment or sample selection issues, but instead to baseline differentials between sites.

Randomization done in office by a computer.

TECHO first selected a set of eligible settlements and then conducted a census to identify eligible households within each settlement (i.e., those poor enough to be given priority). The eligible households were surveyed (baseline
survey) and then randomly assigned to treatment and control groups, within each settlement. Note that this is a household-level randomized experiment, where treatment and control units are co-residents within each selected slum. Each slum is considered as one strata, so we first stratify the sample by slum, and then within each slum we randomly assign households to treatment and control groups.

No

74 slums

2,373 households (32 households per slum, on average)

1,356 treatments; 1,017 controls

Our outcome of interest is housing quality. Considering a sample size of 2,373 units in 74 clusters (i.e, with an average of 32 observations per cluster); a control/treatment ratio of 0.75; a correlation coefficient of 0.3 between baseline and follow up outcome; alpha=0.05; std.dev = 1 (normal standardized outcome); and intracluster correlation = 0.07229; our experimental design is able to identify an effect size of 0.2 standard deviations with an statistical power of 0.8.

This paper provides empirical evidence regarding the causal effects that upgrading slum dwellings has on the living conditions of the extremely poor. In particular, we study the impact of providing better houses in situ to slum dwellers in El Salvador, Mexico and Uruguay. We experimentally evaluate the impact of a housing project run by the NGO TECHO ("roof"), which provides basic pre-fabricated houses to members of extremely poor population groups in Latin America. The main objective of the program is to improve household well-being. Our fi ndings show that better houses have a positive effect on overall housing conditions and general well-being: the members of treated households are happier with their quality of life. In two countries, we also document improvements in children's health; in El Salvador, slum dwellers who have received the TECHO houses also feel that they are safer. We do not find this result, however, in the other two experimental samples. There are no other noticeable robust effects in relation to the possession of durable goods or labor outcomes. Our results are robust in terms of both their internal and external validity because they are derived from similar experiments in three different Latin American countries.

Galiani, Sebastian; Gertler, Paul J.; Undurraga, Raimundo; Cooper, Ryan; Martinez, Sebastian; Ross, Adam. "Shelter from the Storm: Upgrading Housing Infrastructure in Latin American Slums". Journal of Urban Economics. Volume 98, March 2017, Pages 187–213.

https://doi.org/10.1016/j.jue.2016.11.001