Session 4: Implications for Data Collection

Download Report

Transcript Session 4: Implications for Data Collection

SESSION 4:
IMPLICATIONS FOR
DATA COLLECTION
Carmen Diana Deere
University of Florida
Gender Asset Gap Project
World Bank Workshop on Gender & Assets, Gender and Development Group,
June 14, 2012
Issues in Collecting Individual-level Asset
Data
1. Minimum questions required
2. How many assets?
3. Who should be interviewed?
4. If interview two people, how resolve differences?
1. Minimum Questions Required
For Gender Asset Gap, only two questions required beyond
the normal household assets inventory:
• Who are the owners of the asset, with space for multiple owners to
capture joint owners
• GAG project had three spaces, with codes for “all HH members” or non-
HH members who might be joint owners
• Whose names are on the ownership document where appropriate
• For Gender Wealth Gap, besides above, need at least
one measure of valuation for each asset
• See Doss et al (2011), Lessons from the Field, available
at http://genderassetgap.iimb.ernet.in
2. How Many Assets?
Table 1: Share of Gross Physical Household Wealth represented by the four major assets:
Principal residence, agricultural land, other real estate and non-farm businesses
Country
% Four Major Assets
Ecuador
89.9
Ghana
81.6
Karnataka, India
92.5
How Many Assets?
Table 2. Composition of Gross Household Wealth
(preliminary)
Physical
assets
Ecuador
Ghana
Karnataka
(US$)
(PPP US$)
(PPP US$)
82. 2 b.
104.5 b.
423.6 b.
%
Financial
assets
97.6
2.0 b.
%
Total
Gross
Wealth
5.8 b.
2.4
84.2 b.
%
94.7
100.0
Note: Karnataka, India estimate excludes Bangalore
95.1
19.9 b.
5.3
110.3 b.
100.0
4.5
443.5 b.
100.0
Problems in estimating Gross Household
Wealth with the GAG data sets
1. By design, only solicited information on financial assets
in the individual questionnaire to ensure privacy, thus
not a household measure
2. Ecuador & India ended up with truncated samples due
to refusals among upper income groups
3. Many missing values on financial assets
• Usual problem of reluctance to disclose savings/cash (distrust)
4. Problems for comparative analysis:
• Insurance instruments different, not valued consistently
• Have not yet analyzed pensions
Potential information for imputation of
financial assets
Table 3. Share of Physical Wealth of Principal Couple in Couple-headed Households
Country
Couples’ Share of
Wealth (%)
Ecuador
88.1
(n=1981)
Ghana
95.8
(n=994)
India
(n=3066)
83.0
3. Who Should Be Interviewed?
Bardasi, Beegle, Dillon & Serneels (2010) – Tanzania
(n=1344)
• Experimental design to test labor force participation questions and respondent
selection
• Self-reporting provides more accurate information than proxy
Fisher, Reimer & Carr (2010) – Malawi (n=130)
• Husbands underestimate wives’ income
• Disagree on her income in 94% of couple households
Cloke (2009) – Nicaragua (n=359)
• Gendered reporting bias among couples in who is property titleholder and credit holder
• Husbands and wives both more likely to report themselves as the owner rather than
their partner
Fletschner and Mesbah (2011) – Paraguay (n=210)
• Spouses do not fully share information
• Wives less likely to have information on financial institutions
Differences in Survey Design
The three surveys aimed to interview most knowledgeable
person(s) regarding assets owned by HH members
• Ghana & India:
• Principal respondent – completed HH asset inventory & individual
questionnaire
• Secondary respondent – spouse, if present, or other adult of opposite
sex; completed only individual questionnaire and some of asset
ownership & valuation questions re-asked during individual interview.
• Ecuador:
• HH asset inventory completed by the principal couple together
whenever possible (dual-headed households) or if the case, the sole
male/female head
• Each then completed the individual questionnaire separately
• If spouse not available for HH inventory, then asset ownership &
valuation questions always asked again during individual interview
Benefit of Interviewing 2 people
separately: More assets?
Table 4a. Major Assets added by interviewing a second respondent
(% added to Inventory)
Asset
Ecuador
Ghana
Karnataka, India
Principal dwelling
Agricultural
parcels
Other real estate
Non-farm
businesses
0.40
1.36
na
0.35
na
0.15
3.91
0.52
0.41
1.04
0.35
0.12
Table 4b. Major Assets added by asking the Primary Respondent about
ownership twice (% added to Inventory)
Asset
Principal dwelling
Agricultural
parcels
Other real estate
Non-farm
businesses
Ecuador
na
0
Ghana
na
2.25
Karnataka, India
na
0.97
na
na
1.22
0.80
2.82
0.68
Benefit of interviewing 2 people/the couple
separately: Allows gender analysis
• Men and women may have different perceptions of
ownership related to:
• Differences in legal knowledge
• Different understandings of the bundle of property rights
• Gender socialization
• Men and women may have different perceptions of the
value of their assets related to:
• Whether markets exist and degree of integration to markets
• Gender differences in mobility and social networks
Table 5: Disagreements among Couples over who owns the asset
Asset
Country
N (assets) % who
disagree
Dwelling
Ecuador
Ghana
450
510
35.1
7.7
Ecuador
Ghana
94
873
30.9
3.3
Ecuador
Ghana
164
413
20.1
7.8
Ecuador
Ghana
534
641
22.3
1.6
Agricultural land
Other real estate
Non-farm
businesses
Table 6: Disagreement among couples over valuation of the same asset
Asset
Country
Disagree
on Market
value (%)
Disagree on
Replacement
value (%)
Disagree
on
Rental
Value
(%)
Ecuador
60.4
(n=358)
84.0
(n=1718)
65.6
(n=356)
-
57.0
(n=358)
-
70.0
(n=80)
80.0
(n=1528)
-
-
-
-
51.2
(n=123)
79.0
(n=495)
60.7
(n=61)
45.6
(n=103)
64.2
(n=212)
80.0
(n=459)
-
-
-
-
Dwelling
Karnataka
Ag Land
Ecuador
Karnataka
Other Real
Estate
Ecuador
Karnataka
Non-farm
Business
Ecuador
Karnataka
Table 7a: Estimates of market value of principal residence
by husbands and wives
KARNATAKA
ECUADOR
(in INR)
(in US$)
Table 7b: Estimates of market value of agricultural parcels
by husbands and wives
KARNATAKA
(in INR)
Husbands
Wives
Husbands
Wives
N
1667
1667
272
272
Mean
263,621
204,289***
30,771
(s.d.)
(480032)
(459869)
Median
100,000
100,000
Husbands
Wives
Husbands
Wives
N
1516
1516
56
56
28,964
Mean
386,577
299,950***
14,575
10,793**
(31854)
(28695)
(s.d.)
(800700)
(581056)
(18386)
(12979)
20,000
20,000
Median
170,000
120,000
6,500
5,000
Table 7c: Estimates of market value of other real estate
by husbands and wives
KARNATAKA
Table 7d: Estimates of market value of non-farm businesses
by husbands and wives
KARNATAKA
ECUADOR
(in INR)
(in US$)
ECUADOR
(in US$)
(in INR)
ECUADOR
(in US$)
Husbands
Wives
Husbands
Wives
N
457
457
136
136
25,014
Mean
136,927
55,169*
9,459
9,737
(33367)
(33995)
(s.d.)
(1166042)
(325317)
(22743)
(22982)
10,000
10,000
Median
8,000
5,000
1,000
1,075
Husbands
Wives
Husbands
Wives
N
494
494
63
63
Mean
173,784
176,855
25,244
(s.d.)
(556717)
(512959)
Median
50,000
50,000
4. Main problem of interviewing couple
separately: How to reconcile responses?
Ownership disagreements (Ecuador’s rules)
1.
Solved by title (if had document & names agreed with report of one of
respondents)
Solved by “combination” (if document didn’t agree with either report, or no
document, included all mentioned as joint owners)
a)
b)
•
Ghana (where few documents) followed latter rule
Potential problems:
• Ownership question worded differently in individual questionnaire and
HH Inventory
• Bias towards respondent who answered HH Inventory since title
information only asked in this section
• Introduces/exacerbates bias towards joint ownership
Alternative:
• Estimate ownership based on characteristics of individual and joint
owners in full sample?
Ownership disagreements: Impact of
Rules
Table 8a. Ecuador - Disagreements on Home Ownership and
How Resolved by Sex
Disagreement
Frequency
Table 8b. Ecuador - Home Ownership in Household Inventory
vs. After Resolution of Disagreements by Rules
Type of Ownership
%
Between Spouses
158
35.1
Solved by Title
89
19.8
Siding with Male
42
Siding with Female
Solved by Combination
Frequency
according to
Household
Inventory
%
Frequency
following
Final
Resolution
%
Sole Male
49
31.0
29
18.4
9.3
Sole Female
32
20.3
23
14.6
47
10.4
Principle Couple
43
27.2
69
43.7
56
12.4
With Title
20
4.4
21
13.3
25
15.8
Without Title
36
8.0
Principle Couple
and Other
Over Status of Home
13
2.9
Male and Other
5
3.2
5
3.2
Male says Owned
6
1.3
1.9
2
1.3
7
Female and Other
3
Female says Owned
1.6
No Disagreement
292
Other
5
3.2
5
3.2
64.9
Total Homes with Separate
Responses
450
158
100.0
158
100.0
100.0
Total
How to reconcile different responses?
2. Disagreements over value of asset (Ecuador’s rules)
a) If ownership resolved by title and only 1 owner, took
that person’s response; if couple was joint owner,
averaged their estimates
b) If ownership resolved by “combination rule”, averaged
the estimates of the couple
Potential problems:
• Assumes ‘real owner’ knows value of his/her asset best
• Did not remove outliers since planned to average, thus
may overestimate ‘true’ value
• Would be useful to have administrative data as
benchmark
5. Tentative conclusions:
• Interviewing both spouses, second person in general,
minimally increased total assets captured
• In household context probably harder to hide assets than
income
• Most similar situation to income are financial assets:
• If spouse doesn’t know precisely what you earn, easier to hide savings
than other assets
• For that & other reasons, financial assets most difficult to capture fully
even when guaranteeing privacy & confidentiality of responses
• Value of interviewing 2 people rests on capturing gender
differences in perceptions
Thank you!
For the country studies
& comparative report
see:
http://genderassetgap.iimb.
ernet.in