Demand for climate resilient housing – Experimental evidence from Vietnam

Kategori	Artikler
Underkategori(er)	n/a
År	2017
Forfatter(e)	Henrik Lindhjem Sofie Skjeflo Tran Huu Tuan Tran Van Giai Phong Tuan Anh Tran Haakon Vennemo Nina Bruvik Westberg
Last ned	file_download (1.1 MB)
Les i nettleser	find_in_page

Demand for climate-resilient housing
Experimental evidence from Vietnam	
June 9, 2017	
Henrik Lindhjem	1, Soﬁe Waage Skjeﬂo	2, Tuan Huu Tran	3, Phong Van Giai Tran	4, Tuan Anh	
Tran	5, Haakon Vennemo	6and Nina Bruvik Westberg	7	
Abstract	
We present preliminary results from a randomized ﬁeld experiment in Da Nang in Central	
Vietnam. We assess the take-up of a microcredit program that aims to increase the adoption
of climate resilient housing among near-poor urban households. Households were randomly
assigned o	↵ers of either a loan and technical assistance package or a cash transfer in addition	
to the loan and technical assistance. The aim of this study is to investigate the di	↵erence in	
take-up of the packages, and how this di	↵ers with respect to household characteristics. This	
again speaks to the broader question of what are the necessary incentives for households to
invest in climate resilient housing. We ﬁnd that households that were o	↵ered the cash transfer	
in addition to the loan and technical assistance, are almost twice as likely to accept. Household
heads with less education are less likely to accept the o	↵er of a loan and technical assistance	
only.	
1Menon Centre for Environmental and Resource Economics2Vista Analyse3College of Economics, Hue University4ISET-Vietnam5ISET-Vietnam6Vista Analyse7Corresponding author: Vista Analyse, nbwestberg@vista-analyse.no. This work is supported by the Nordic Devel-
opment Fund under the Nordic Climate Facility.	
1

1 Introduction
Climatic hazards such as coastal ﬂooding, tropical storms and typhoons	8pose a signiﬁcant threat	
to low income households in growing cities in Vietnam. With strong rural-urban migration by low-
income households (	Nguyen et al.	,2015	), many households cannot a	↵ord permanent, high-quality	
housing, and are at risk of severe structural damage to their homes (	UN	,2014	). In addition, since	
the 1986 reform (Doi Moi Policy), there has been a shift from traditional construction materials,
such as wood and clay tiles, towards the use of modern materials, such as cement blocks and
corrugated steel sheets, but without adjusting construction practices to incorporate resilience to
climate hazards (	Norton and Chantry	,2008	). According to o	�cial estimates, almost 1,5 million	
houses were damaged, and about 50 000 houses totally destroyed due to storms and ﬂoods between
1999 and 2013, with most of the damage occurring in Central Vietnam (	Tran	,2016	). Storms and	
ﬂoods disproportionally a	↵ect low-income households who often live in houses that are constructed	
with structurally weak materials and techniques. These households frequently reconstruct their
houses after storm damage using the same building principles, thereby reproducing their vulnera-
bility (	Tran et al.	,2013	). A number of programs have addressed these issues, and made e	↵orts to	
promote storm- and ﬂood resilient housing for low income households (see for instance	Norton and	
Chantry	,2008	;Tran et al.	,2013	).	
This paper presents results from an ongoing randomized control trial of a microcredit program	
aiming to increase the adoption of climate resilient housing in the coastal city of Da Nang. The
programme is implemented by Women’s Union, a mass organization operating throughout Vietnam
with the aim of enabling women to take part in national development. The underlying motivation
of the program is that enabling households to invest in resilience rather than relief and repairs,
provides long term beneﬁts to both the household and to society. Based on a cost-beneﬁt analysis
of climate resilient housing,	Tuan et al.	(2015	) ﬁnd positive returns to investment, even without	
taking into account non-monetary costs of typhoon damage and possibly increased typhoon inten-
sity due to climate change. A high marginal return to a technology (in this case, climate resilient
8A typhoon is a rotating system of clouds and thunderstorms with wind speeds exceeding 74 miles per hour, originating
in the Northwest Paciﬁc region (	http://www.nhc.noaa.gov/climo/	).	
2

housing), indicates that there may be market failures that constrain adoption (	Foster and Rosen-	
zweig	,2010	). In this case, likely barriers to adoption include information barriers (both lack of	
knowledge about the returns to investment, and lack of knowledge of how to manage the technol-
ogy) and credit constraints. The program we evaluate aims to mitigate the credit constraint of low
income households through providing access to credit for housing construction or retroﬁtting, and
providing information and technical assistance to implement climate resilient building principles.	
The purpose of the randomized control trial is twofold: First, to assess the impact of investing	
in climate resilient housing on a number of household outcomes, amongst them various indicators
of resilience, sub jective resilience and well-being, household consumption, and investment in non-
housing assets. Here we are interested in both the intention-to-treat e	↵ect and the treatment e	↵ect	
on the treated. Secondly, to inform programme design through investigating the necessary level of
incentives provided in order to enable households to invest in climate resilient housing. Here we are
interested in the take-up of two “incentive packages”, one with a subsidized loan and free technical
assistance for house retroﬁtting or reconstruction, and one with cash transfer in addition to the
loan and technical assistance. We are also interested in the characteristics of the households that
accept each package, since we want to know which households can be reached through scaling up
the programme, and which types of households may have to be targeted through other channels.
This paper addresses the second purpose of the RCT, by investigating the take-up of each of the
two incentive packages, and investigating the characteristics of the households that accept each
package.	
The rest of this paper is organized as follows. Section 2 presents a theoretical model of the	
decision problem of the household, and section 3 explains the experiment. The data and results
are presented in section 4 and section 5 concludes.
2 Theoretical background
Consider a household	ithat maximizes expected utility over consumption and housing consumption	
in two periods. In period 2, the household enters state 1 (no typhoon) or state 2 (typhoon). The
household may have access to two saving and investment technologies in period 1. One is bank	
3

savings (	S), which may be positive or negative (borrowing). Saving pays o	↵	with the interest	rin	
both states. The other is housing reinforcement (	H), which only pays o	↵if there is a typhoon.	
To simplify, we assume two states of the world in period 2. The state “no typhoon” occurs	
with probability	p. The state “typhoon” occurs with probability (1	�	p). Utility over ordinary	
consumption and housing is additive. We suppress utility of existing assets, such as existing houses.
Utility of housing in the state “no typhoon” is normalized to zero. Hence, utility of housing in the
state “typhoon” is negative, unless the household has invested in housing reinforcement in period
1.	
The utility function of the households is	U(C1)+	✓pU	(C2)+	✓(1	�	p)(U(C2)�	V(H)), where	✓	
is the sub jective time discount factor of the household. The unconstrained maximization problem
of the household is then:	
maxC1,C2,S,H	U(C1)+	✓pU	(C2)+	✓(1	�	p)(U(C2)�	V(H))	�	µ(C1+	S+	H	�	R)�	⇢(C2�	(1 +	r)S) (1)	
R	in the problem is certain discounted household income, which can be used for either consumption	
or saving. Income is certain for simplicity. In the interior solution of the unconstrained maximum
we obtain the following ﬁrst order conditions:
U0(C1)�	µ=0	
✓U0(C2)�	⇢=0	
�µ+	⇢(1 +	r)=0	
�✓(1	�	p)V0(H)�	µ=0	
This leads to rules of motion for consumption and housing:	
U0(C1)�	(1 +	r)✓U0(C2) = (2)	
�	✓(1	�	p)V0(H)�	U0(C1) = (3)	
The ﬁrst rule of motion is familiar in intertemporal savings problems and says that the marginal	
utility of consumption in period one should equal the marginal utility of consumption in period	
4

two while considering the interest on saving period one consumption and the sub jective displeasure
(discount factor) of deferring consumption.	
The other rule of motion says that the marginal utility of consumption in period one should	
equal the expected marginal beneﬁt of housing consumption in period two. (Since	V0(H) is negative	
(�V0(H)) is positive). The logic of this equation is that the beneﬁt of housing consumption in the	
“typhoon” state should be balanced against the cost in terms of foregone consumption in period
one.	
We cannot expect that households in our sample are in the interior of an unconstrained optimum.	
But we can use the Lagrange function to deﬁne the household marginal willingness to pay for
reinforced housing as follows:	
q=	�✓(1	�	p)V0H	�	µ=	q(r, p	;H, R,	✓) (4)	
As above, the beneﬁt of housing reinforcement is the increase in utility (	�V0(H)) should a	
typhoon occur (1	�p) and considering the sub jective displeasure of deferring consumption (	✓). The	
cost of housing reinforcement (	µ) by deﬁnition equals the marginal impact of income on overall	
utility. The interpretation is that by spending on housing the consumer has less income to spend
overall. The cost of housing reinforcement depends on the parameters of the system. The marginal
willingness to pay,	q, equals the di	↵erence between the beneﬁt of housing reinforcement and its	
cost, that is the	net	marginal utility of housing reinforcement.	
Each household	iis expected to have an individual willingness to pay	q(i). Here we are interested	
in their willingness to pay at	H(i) = 0. In other words, what is their willingness to pay for housing	
reinforcement before the do any reinforcement? Based in the work of	Tuan et al.	(2015	), an	
unconstrained household is expected to have a positive willingness to pay for reinforced housing.
However, many households do in fact express a negative willingness to pay. These are the households
that need a ﬁnancial incentive to reinforce their homes. Their negative willingness to pay equals
their demand for ﬁnancial support. We are interested in these households.	
5

Hence, we focus on the following set of negative willingness to pay/demand for ﬁnancial support:	
Q	=	
Z	i(q=0)	
i(q=�¯q)q(i;r(i),p	(i),R	(i),✓(i),H	(i) = 0)	di	(5)	
We order individuals from the lowest willingness to pay, up to the cut o	↵at zero.	�¯qis the lowest	
of the	q’s. Note that the reasons for a demand for ﬁnancial support that were emphasized as	
barriers above, show up as explanatory variables for	q(i): Demand for support may arise because	
the household faces a prohibitively high interest rate	r(i) on borrowing, or because the sub jective	
probability of a typhoon is too low (	p(i)), or simply because it has low income (	R(i)). Factors	
such as a high sub jective discount factor (	✓(i)) may also be at play, and there are other sub jective	
di↵erences and characteristics between households.
Incentive package 1 o	↵ers a low interest loan, and should inﬂuence	q(i) through a decrease in	
the loan interest rate	r(i). Both the income and substitution e	↵ect should contribute to a higher	
q(i). Incentive package 2 exchanges part of the loan for a grant, and should inﬂuence	q(i) through	
a decrease in	r(i) and an increase in	R(i): The income e	↵ect is then given a boost from the grant	
element. We expect a higher impact on	q(i) from package 2, and hence higher uptake. In addition	
to the impact on	r(i) and	R(i), both packages will help neutralize unrealistic perceptions about	
the probability of a typhoon (	p(i)). The value of package 2 is of course higher than the value of	
package 1, which should entice higher uptake.	
A ma jor goal of our research is to contribute understanding of what constitutes necessary	
ﬁnancial support. In particular we would like to know the shape of the	q(i)curvebetween	q1and	
q2,where	q1is incentive package 1 and	q2is incentive package 2:	
Qexperiment	1=	
Z	�q1	
i(�q2)q(i)di	(6)	
How ﬂat is the demand-for-support curve in this interval, illustrated in Figure	1? Will increasing	
the incentive from	q1to	q2enlist many additional households? If so, what are the characteristics	
of these households? These are questions we address in the next chapter.	
6

i	q2	q1	Difference	in	uptake	q1	–	q2	q		
Figure 1: Willingness to pay for housing improvement, q	
We are also interested in the group	
Qexperiment	2=	
Z	0
i(�q2)q(i)di	(7)	
How ﬂat is the demand-for-support curve in this interval? In other words, how many reinforce	
their house without incentives, compared to those who do so with the strongest incentive? What
are the characteristics of these households? This is another set of questions we address in the next
chapter. We also note that experiment 1 and experiment 2 in combination imply a shape of the
demand-for-support curve in the interval (	�q1,0).	
3Experiment
3.1 Study setting
Da Nang is located on the South Central Coast of Vietnam in the tropical storm belt. It is the third
largest city of Vietnam, and has a fast growing population estimated at 1.03 million in 2015, with
an average annual growth of about 2.9 percent per year since 2010 (	GSO	,2016	). Numerous climatic	
7

hazards a	↵ect Da Nang, such as coastline erosion, landslides, ﬂoods and typhoons, with on average	
three to ﬁve tropical storms a	↵ecting the city per year (	Tran et al.	,2013	). The peak typhoon	
season is from September until November. Past typhoons that have heavily a	↵ected Da Nang	
include Xansane in 2006, Ketsana in 2009 and Nari in 2013. Typhoon Xansane caused the collapse
of 24 000 houses and damaged 325 000 houses in Central Vietnam (	Tran	,2016	), and 26 people in	
Da Nang were reported killed, mostly due to collapsing houses.	9	Typhoon Nari caused estimated	
damage to housing of about 4.6 billion USD, with 4200 severely damaged houses, and 122 completely
collapsed houses (	Tran	,2012	). In 2011, the Women’s Union of Da Nang and the Institute of Social	
and Environmental Transition (ISET) initiated the pro ject	The Storm and Flood-Resistant Credit	
and Housing Scheme in Da Nang City	. With funding from the Rockefeller fundation, a microcredit	
and technical assistance program aimed at supporting storm resistant shelters in Da Nang City, was
set up. Between 2011 and 2016, 237 new houses were constructed and 177 houses were retroﬁtted
with support from the Women’s Union revolving loan fund, with free technical assistance from
local architects. When typhoon Nari hit in 2013, 244 households participating in the program had
completed construction, and none of the completed houses su	↵ered damages (	Tran	,2012	). The	
program was awarded the UNFCCC Lighthouse Activities Award in 2015 for its e	↵orts to increase	
urban poor’s resilience to climate change.	10	
Previous research indicates that investing in climate resilient features when retroﬁtting or re-	
constructing a house is proﬁtable from a cost-beneﬁt perspective.	Tuan et al.	(2015	) use recall	
data from 98 households in Da Nang that were a	↵ected by the Xansane and Ketsana typhoons.	
The beneﬁt of investing in resilient housing is calculated based on information about direct costs
(damage to housing and other assets) and indirect monetary losses (loss of work days, medical
costs). The avoided costs of typhoon damage is then compared to the additional cost of resilient
housing construction, based on a number of principles such as choice of building material, reinforce-
ment, anchoring and wall thickness. Assuming a return time of a category 1 storm (like Xangsane)
every 12.5 years, the internal rate of return to climate resilient housing is as high as 20 percent,
and assuming a return period of 25 years, the internal rate of return is 14 percent. Non mone-
9Associated Press,	https://usatoday30.usatoday.com/weather/storms/2006- 10- 03- asia- typhoon_x.htm	10http://blog.i- s- e- t.org/we- are- honored- to- accept- the- unfcccs- lighthouse- activities- award/	
8

tary beneﬁts, such as the value of feeling safe, are not included in the analysis. According to this
study, investing in climate resilient housing, as deﬁned by the principles outlined and assuming a
10 percent discount rate for homeowners, is on average proﬁtable.
3.2 Experimental design
Da Nang is organized into eight districts and 56 wards and communes (hereon called wards).	11	Our	
experiment takes place in 49 of these wards, covering seven districts. This excludes ﬁve wards in
which most households reside in apartment buildings, rather than individual houses.	
We ﬁrst made a list of potential beneﬁciaries from each of the 49 wards. The local Women’s	
Union units were asked to compile a list of six near-poor	12	households per ward that were eligible	
for participation in the program. The information provided to the wards is described in Appendix
A. The number six was chosen due to budgetary constraints. The ﬁnal number of households per
ward varies however, but totals 306 households. According to the targeting criteria, the households
had to have housing conditions vulnerable to climate risks, limited access to ﬁnancial resources
for strengthening climate resilience housing, have a stable job but low income, have some savings
and be able to mobilize labor force for their housing improvement, have capacity for repayment,
and have legal or any related documents to prove land ownership. The near-poor households were
approached, and asked whether they had a need for house retroﬁtting or reconstruction to ensure
storm resilience, and with the wish to carry out such retroﬁtting or reconstruction starting from
March 2017.	
The 306 households were surveyed between late December 2016 and early February 2017. The	
survey instrument included modules on household composition as well as di	↵erent conditions and	
issues that may a	↵ect or be a	↵ected by the housing investment. This includes modules on income,	
11Wards and communes are the same administrative unit, but the name di	↵ers between rural areas (communes) and	
urban areas (wards).12The Da Nang City poverty lines were updated in 2016 (	City People’s Council, Da Nang City	,2015	): Urban poor	
households: Income per capita per month (ICM)	<1300000 VND, Urban near-poor households: 1300000 VND		
ICM	1690000 VND, Rural poor households: ICM	<1100000 VND, Rural near-poor households: 1100000 VND	
	ICM		1430000 VND. This corresponds to a near-poor poverty line of about 75 USD per capita per month in	
urban wards, and 63 USD per capita per month in rural areas of Da Nang. By comparison, the national poverty
line for urban areas was set at 760 000 VND per capita per month for urban areas, and 615 000 VND per capita
per month for rural areas, according to the General Statistical O	�ce of Vietnam (	General Statistics O	�ce of Viet	
Nam	,2016	).	
9

consumption, assets as well as past storm exposure and housing repairs.	
Randomization of wards into treatment and control groups took place after the baseline survey.	
We randomized using a random number generator. We chose to randomize at the ward/commune
level so as to avoid spillover e	↵ects of information. Randomization was stratiﬁed at the district	
level, in line with the wishes of the Women’s Union.	
We have two treatment arms: package 1 and package 2. Households in package 1 are provided	
with a subsidized loan and free technical assistance for house retroﬁtting or reconstruction. House-
holds in package 2 are given a cash transfer, in addition to the loan and free technical assistance.
The loan is 30 million VND (close to one third of the value of an average house), while package 2
also consists of a grant of 10 million VND. The maximum loan repayment period is 40 months and
the monthly interest on the loan is 0.75 percent, or about 9 per cent per year. The loan and grant
is disbursed to the household when their house is dismantled or they have started retroﬁtting. The
households start paying capital and interest monthly, starting from the ﬁrst month that the loan
is provided. The technical assistance includes technical designs of the new house/retroﬁtted house,
as well as technical guidance during the building period. Figure	2shows a map of Da Nang City,	
with the wards assigned to the two treatment arms and the control group.	
10

Figure 2: Map of Da Nang City with wards assigned to package 1, package 2 and control
Following the randomization, the local Women’s Union units at the ward/commune level were	
informed which category their households belonged to (package 1, package 2 or control), and the
households were then contacted directly by the local Women’s Union unit. The households were
provided with the information both verbally and by means of an information brochure that detailed
the speciﬁcs of the package, such as interest rate and repayment conditions of the loan. They were
given a consideration period of roughly two weeks. Our measure of take up reﬂects their ﬁnal
decision. Actual reconstruction and retroﬁtting is currently taking place, and will for most of the
households continue until September 2017. A few households have asked for permission to postpone
reconstruction/retroﬁtting until after the next Vietnamese New Year, i.e. in 2018.
3.3 Baseline means and balance checks
Table	??	reports baseline means in columns 1-3 for the control group and the two treatment arms.	
Around half of the households are female-headed, in line with the Women’s Unions’ targeting of	
11

women. The household heads are on average 54 years old, and have on average 6.5-8.5 years of
education. The households consist on average four to ﬁve people. Average monthly income per
person is around 1 340 - 1 660 thousand VND per month, whereas average monthly expenditures
per person vary between 630 - 970 thousand VND, depending on the treatment arm. Around 46
percent of the households in each of the treatment arms have a loan from beforehand, whereas
the same applies for just over 60 percent of the households in the control group. These loans are
primarily for business activities, house construction/retroﬁtting/repairs and education expenses.
A large share of the households own motorbikes, an important asset in the Vietnamese context,
whereas between 52-68 percent of the households own a TV.	
We have randomized at the ward level. Although all wards have identiﬁed households that are	
near-poor and satisfy the above mentioned criteria, there may be systematic di	↵erences between	
the wards. In column 4, Table	??	we report the p-value from a joint orthogonality test (F-test) of	
treatment arms. There seem to be some systematic di	↵erences between the groups, for instance,	
there is a statistically signiﬁcant di	↵erence in the level of education, household size, expenditure and	
income, asset ownership and previous loan experience between the three groups. We acknowledge
this by including the baseline variables as controls in the analysis.	
In total, 112 households among the 214 households that were o	↵ered accepted the package,	
equivalent to 52 percent take-up rate. We have registered reasons for not accepting among 78 of
the remaining 102 households. The most common reasons for declining are that the household
cannot a	↵ord to repay or that they are not interested in a loan, either because they are already	
indebted or because they are simply not interested.	
12

Table 1: Baseline means and balance checks	
(1) (2) (3) (4)	
Control Treatment 1 Treatment 2 p-value from	
joint	
orthogonality	
test of	
treatment arms	
Female household head 0.554 0.538 0.481 0.551	
(0.052) (0.049) (0.048)	
Age of household head 56.185 53.575 54.620 0.393	
(1.440) (1.247) (1.313)	
Education level of household head 6.511 8.500 6.824 0.002	
(0.497) (0.400) (0.398)	
Household size 4.793 4.028 4.444 0.009	
(0.213) (0.137) (0.169)	
Monthly income (1000 VND) per pers 1339.926 1658.638 1611.220 0.033	
(64.339) (97.651) (97.716)	
Monthly expenditures (1000 VND) per pers 898.882 630.789 974.610 0.000	
(65.024) (53.710) (67.581)	
Owns motorbike 0.837 0.934 0.861 0.088	
(0.039) (0.024) (0.033)	
Owns TV 0.522 0.679 0.620 0.074	
(0.052) (0.046) (0.047)	
Household member has a loan 0.620 0.462 0.463 0.042	
(0.051) (0.049) (0.048)	
Years lived in house 23.451 19.858 22.046 0.247	
(2.058) (1.146) (1.336)	
N	92 106 108	
13

4Estimationstrategy
We are interested in the take-up of the packages, in particular the di	↵erence in take-up between the	
two packages. We estimate the intention-to-treat (ITT) on the households in both the treatment
and control groups, regardless of whether they were o	↵ered and whether they accepted.	
We estimate the following linear probability model:	
Yhwd	=	↵0+	↵1T1+	↵2T2+	X0hwd	�+	⌘hwd	(8)	
where the dependent variable	Yhwd	indicates whether the household accepted a package or not;	h	
indicates household in ward	w	in district	d.T1and	T2are dummies indicating whether a household	
was o	↵ered package 1 or package 2, respectively. Since the treatment and control groups di	↵ered	
on a number of aspects, we control for these di	↵erences. The vector	X0hwd	indicates a number	
of baseline characteristics. We include these to control for di	↵erences in baseline characteristics	
and to improve the precision of our estimates.	⌘hwd	is a mean zero error term, and we cluster the	
standard errors at the ward level.	
In addition we investigate whether take-up di	↵ers systematically according to baseline charac-	
teristics. Speciﬁcally, we are interested in gender, education and income.
5 Results
5.1 Take-up
The main results on take-up are reported in Table	??	. Not surprisingly, take-up is higher in the	
treatment groups than in the control group. By deﬁnition none of the households in the control
group are registered as accepting a package, simply because they were not o	↵ered one. This does	
however not mean that none of the households in the control group may decide to reconstruct or
retroﬁt their house to improve climate resilience in the same period as the treated households. This
is something we will investigate in the follow-up survey, to be conducted in the spring of 2018.	
We are more interested in the di	↵erence in take-up between the two packages. Households in	
14

treatment 1 are 36.1 percentage points more likely to accept the package than in the control group,
whereas households in treatment 2 are 63.2 percentage points more likely to accept. There is in
other words a di	↵erence in take-up by close to 30 percentage points, which is statistically signiﬁcant.	
Households in treatment 2 are roughly twice as likely to accept as the households in treatment 1.	
The di	↵erence between the two packages, is that package 2 provides 10 mill. VND as a grant	
rather than as a loan. Package 2 is in other words more costly to provide. The grant will not be
repaid, and thus the money paid out will not reenter into the revolving fund. Half of those that
accepted package 2 are likely to have accepted package 1, if they were o	↵ered that package instead.	
On the other hand, if households are only o	↵ered package 1, then this results in lower take-up. The	
question then is: who are we reaching by o	↵ering a more attractive package?	
15

Table 2: Take-up of packages	
Package 1 0.361	⇤⇤⇤	
(0.096)	
Package 2 0.632	⇤⇤⇤	
(0.080)	
Female household head 0.069	
(0.051)	
Age of household head -0.000	
(0.002)	
Education level of household head 0.003	
(0.007)	
Household size 0.004	
(0.015)	
Monthly income (mill. VND) per person 0.067	
(0.042)	
Monthly expenditures (mill. VND) per person 0.034	
(0.062)	
Owns motorbike -0.026	
(0.058)	
Owns TV -0.003	
(0.059)	
Household member has a loan -0.059	
(0.048)	
Years lived in house -0.001	
(0.002)	
Constant -0.095	
(0.159)	
Observations 305
Prob	>	F (Package 1 = Package 2 ) 0.000	
Notes:	Linear probability model. Dep.var.: take-up is set equal to 1 if	
household accepted package 1 or 2, and set equal to if the household
declined or was not o	↵ered a package. Standard errors in parentheses,	
clustered at the ward level.	⇤p<	0.05,	⇤⇤	p<	0.01,	⇤⇤⇤	p<	0.001	
16

5.2 Heterogeneity in take-up
We investigate whether take-up di	↵ers according to baseline characteristics, speciﬁcally gender,	
education and income. Results are reported in Table	??	. We ﬁnd that the interactions between	
education and package 1 is positive and statistically signiﬁcant from zero. The probability of
accepting package 1 relative to package 2 increases with education, in line with our expectation
that less educated may be less aware of the beneﬁts of investing in climate-resilient housing. We
ﬁnd no statistically signiﬁcant di	↵erence in take-up with regards to gender and annual income per	
person.	
We also investigate whether there is any di	↵erence in whether the household chose to retroﬁt	
or rebuild their house depending on what package they received. We ﬁnd that there is no statis-
tically signiﬁcant di	↵erence (results available upon request), although this may be because we are	
underpowered.	
17

Table 3: Heterogeneity in take-up	
1234	
Package 1 0.421	⇤⇤⇤	0.031 0.186 0.015	
(0.119) (0.109) (0.098) (0.144)	
Package 2 0.634	⇤⇤⇤	0.634	⇤⇤⇤	0.643	⇤⇤⇤	0.641	⇤⇤⇤	
(0.079) (0.079) (0.080) (0.080)	
Female household head 0.115 0.049 0.067 0.078	
(0.063) (0.050) (0.051) (0.060)	
Education level of household head 0.005 -0.009 0.004 -0.007	
(0.007) (0.008) (0.007) (0.008)	
Annual income (mill. VND) per person 0.006 0.006 0.002 0.003	
(0.003) (0.003) (0.003) (0.003)	
Package 1 * Female household head -0.136 -0.072	
(0.095) (0.091)	
Package 1 * Education level of household head 0.041	⇤⇤	0.035	⇤	
(0.015) (0.014)	
Package 1 * Annual income (mill. VND) per pers 0.009 0.005	
(0.005) (0.005)	
Observations 305 305 305 305
Prob	>	F (Package 1 = Package 2 ) 0.000 0.000 0.000 0.000	
Notes:	Linear probability model. Dep.var.: take-up is set equal to 1 if household accepted package 1 or 2, and	
set equal to if the household declined or was not o	↵ered a package. Variables not reported:	household size	,	
monthly expenditures (mil l. VND) per person	,owns motorbike	,owns TV, household member has a loan	and	years	
lived in house	. Standard errors in parentheses, clustered at the ward level.	⇤p<	0.05,	⇤⇤	p<	0.01,	⇤⇤⇤	p<	0.001	
18

6 Conclusion
We present preliminary results from a randomized ﬁeld experiment in Da Nang, Vietnam. House-
holds were randomly provided with a loan and technical assistance package or a cash transfer in
addition to a loan and technical assistance. The aim of this study is to investigate the di	↵erence	
take-up of the packages, and how this di	↵ers according to household characteristics. This again	
speaks to the broader question of what are the necessary incentives for households to invest in
climate resilient housing.	
We ﬁnd that households that were o	↵ered a cash transfer in addition to the loan and technical	
assistance are almost twice as likely to accept, than households not o	↵ered a cash transfer. In partic-	
ular less educated household heads are less likely to accept package 1 as opposed to package 2. This
suggest that there may be informational and/or ﬁnancial barriers that result in underinvestment
in climate resilient housing that persist despite the loan and technical assistance package.
References
City People’s Council, Da Nang City (2015). The Resolution to adopt the Poverty Reduction Plan	
for 2016-2020 in Da Nang.	
Foster, A. D. and Rosenzweig, M. R. (2010). Microeconomics of technology adoption.	Annu. Rev.	
Econ.	, 2(1):395–424.	
General Statistics O	�ce of Viet Nam (2016). General poverty rate by residence and by region and	
year.	
GSO (2016). Statistical yearbook of vietnam 2015. Technical report, General Statistical O	�ce of	
Vietnam.	
Nguyen, L. D., Raabe, K., and Grote, U. (2015). Rural–urban migration, household vulnerability,	
and welfare in Vietnam.	World Development	, 71:79–93.	
Norton, J. and Chantry, G. (2008). Vaccinate your home against the storm – reducing vulnerability	
in Vietnam.	Open House International	, 33(2).	
19

Tran, T. A. (2016).	Developing Disaster Resilient Housing in Vietnam: Chal lenges and Solutions	.	
Springer.	
Tran, T. A., Tran, V. G. P., Tran, T. H., and Mulenga, M. (2013). Community consultation for	
long-term climate-resilient housing in Vietnamese cities: a comparative case study between Hue
and Da Nang.	Asian Cities Climate Resilience, IIED	.	
Tran, V. G. P. (2012). Statistical yearbook of vietnam 2015. Technical report, Institute for for	
Social and Environmental Transition International (ISET).	
Tuan, T. H., Tran, P., Hawley, K., Khan, F., and Moench, M. (2015). Quantitative cost-beneﬁt	
analysis for typhoon resilient housing in Danang city, Vietnam.	Urban Climate	, 12:85–103.	
UN (2014). Migration, resettlement and climate change in Viet Nam. reducing exposure and	
vulnerabilities to climatic extremes and stresses through spontaneous and guided migration.
Technical report, United Nations Viet Nam.	
Appendix A
Information to ward leaders/WU at ward level on the selection of eligible households
The Women’s Union of Da Nang is carrying out a survey to investigate the need for storm resilient
housing in Da Nang City. The pro ject is carried out with funding from the Nordic Development
Fund, and will support the Da Nang City’s Resilience Strategy. In this ﬁrst round, the Women’s
Union is carrying out a mapping of the need for housing retroﬁtting and reconstructions for house-
holds with houses that are vulnerable to storms. The six households will be visited for a more
detailed survey at a later point, tentatively in December 2016.	
At this point, the Women’s Union would like to identify households categorized as	near poor,	
with a need for house retroﬁtting or reconstruction to ensure storm resilience, and with
the wish to carry out such retroﬁtting or reconstruction starting from March 2017.
We are therefore asking each ward to provide a list of six households that ﬁt these criteria. If there	
20

are more than six households in the ward that ﬁt the criteria, we would also like a speciﬁcation of
how the six households were nominated among the larger group of households.	
21

Parse pdf-doc by clicking the button below. Parsed pdf-documents will be searchable.

Metadata Found
Producer
Mac OS X 10.12.4 Quartz PDFContext
CreationDate
2017-06-14T08:56:57+00:00
ModDate
2017-06-14T08:56:57+00:00
Pages
21