HLEG thematic workshop on Measurement of Well Being and Development in Africa, 12-14 November 2015, Durban, South Africa, More information at: www.oecd.org/statistics/measuring-economic-social-progress

1. SUTIRTHA BAGCHI† & JAN SVEJNAR‡
† Villanova University
‡ Columbia University
NOVEMBER 2015
Does Wealth Inequality Matter for
Growth? The Effect of Billionaire Wealth,
Income Distribution, and Poverty

2. Motivation for the paper
 “. . . the absence of data on the distribution of wealth
for a sufficient number of countries forces researchers
to use proxies in empirical studies. The most common
approach is to use data on income inequality as a proxy
for wealth inequality.” Aghion, Caroli, and Garcia-
Penalosa (1999)
 Bénabou (1996) echoes this point and notes that the
lack of almost any data on the distribution of wealth is a
general problem, given that in most theories it is this
distribution rather than that of income which is the
determinant of outcomes.
 Ravallion (2012) emphasizes that “wealth inequality is
arguably more relevant though this has been rarely
used due to data limitations.”

3. Research questions
1. How does wealth inequality affect economic
growth?
2. Does the relationship between growth and
inequality depend on the nature (source) of
this inequality?
e.g. Does inequality based on political connections differ
from one that is based on success as an entrepreneur?
3. What is the relative growth effect of wealth
inequality, income inequality, and poverty?

4. Theoretical literature provides arguments for why
inequality is good for growth
1) Marginal propensity to save of the rich is
higher than that of the poor
2) Investment indivisibilities:
 Low inequality  Low levels of innovation  Low
productivity growth  Low growth in real GDP
per capita
1) Trade-off between equity and efficiency

5. … but it also provides arguments for why
inequality is bad for growth
1) Credit market imperfections: You cannot
borrow against your human capital
2) Greater demand for redistribution leading to a
choice of economically inefficient policies; and
3) Greater social unrest, possibly also leading to
a higher degree of macroeconomic volatility

6. Existing empirical evidence: Mixed
Cross - country & cross-sectional regressions
suggest that income inequality is bad for growth:
 Alesina & Rodrik (QJE, 1994)
 Persson & Tabellini (AER, 1994)
Results do not always hold up under robustness
checks; do not answer the question of what
happens when inequality in a given country
changes
Distinctly different results when examined in a
panel set-up
 Forbes (AER, 2000)

7. Data source for the paper
 Forbes magazine’s list of billionaires:
 Published list of billionaires from around the world since
1987
 Estimate wealth based on the holdings of individuals in
public companies or estimated holdings in private
companies using standard price multiples
 We use the Forbes ’ billionaire data set to create two
variables:
 Proxy measure of wealth inequality =
Sum of wealth of all billionaires in a country/ Country GDP
 E.g. Country 1 has 3 billionaires with wealths equal to $5 billion,
$2 billion and $1 billion, and country’s GDP = $500 billion.
 Measure of wealth inequality = (5 + 2 + 1)/ 500 = 1.6%

8. Correlations between wealth distribution data
from UNU–WIDER & Forbes’ list of billionaires
Raw correlation coefficient and Spearman rank correlation
coefficient for the share of wealth going to the top decile and our
measure of wealth inequality for a sample of 18 countries are 0.54
(p-value = 0.0199) and 0.58 (p-value = 0.0122).
Cross-country correlation between the Gini coefficients of wealth
available for 22 countries for the year 2000 from the Davies et al.
(2008) data set and our measure of wealth inequality for 2002:
0.50 (p = 0.0188).
These are relatively high positive correlations

9. We split wealth inequality into two components
Wealth Inequality (or Billionaire wealth/GDP)
Classify billionaires as politically connected or not
(A billionaire can be in only one of the two categories)
Previous example: Suppose billionaire 2 gets
classified as politically connected
Politically connected billionaire wealth / GDP =
$2/$500 = 0.4%
Politically unconnected billionaire wealth / GDP =
$6/$500 = 1.2%
“Politically connected” “Politically unconnected”
billionaire wealth /GDP billionaire wealth /GDP

10. How do we classify someone as “politically
connected”?
 Extensive search on Factiva & Lexis-Nexis
 “Criteria”:
 Have political connections played a material role in the
success of the billionaire?
 Would they have been billionaires absent political
connections?
 Careful to distinguish between explicit government support
from a generally pro-business regulatory environment
 Classic examples: Oligarchs from Russia or the
cronies of Suharto (Indonesia)

11. Ranking of countries in terms of politically connected
matches priors
Countries that rank highest in terms of politically connected wealth inequality
1. Malaysia
2. Colombia
3. Indonesia
4. Thailand
5. Mexico
Other countries which just follow these include – Chile, South Korea,
Philippines, Argentina, and, India. Italy has the 11th
highest level of politically
connected wealth inequality in our sample – the highest of any European
country.
Median rank on TI’s Corruption Perceptions
Index: 32 /41 (1995) & 94/174 (2012)
Countries that rank lowest in terms of politically connected wealth inequality
1. Hong Kong
2. Netherlands
3. Singapore
4. Sweden
5. Switzerland and
6. United Kingdom
Median rank on TI’s Corruption Perceptions
Index: 9 /41 (1995) & 8/174 (2012)

12. What we include in our data set
20-year period from 1988 – 2007 divided into 4 periods
of 5 years duration each
All countries in the world subject to availability of data
on covariates. When a country does not have billionaires,
we set billionaire wealth = 0 (more on this later)
~ 60 countries (and 160 country-period combinations)
appear in the final estimation
Growthi,t = β0 + β1Wealth inequalityi,(t−1) + β2Income inequalityi,(t−1) +
β3Headcount povertyi,(t−1) + β4Incomei,(t−1)+ β5Schoolingi,(t−1) + β6PPPIi,(t−1)
+ β7Dummyi,(t−1)+ αi + ηt + νi,t

13. One may be concerned about reverse causality
Relationship runs not from inequality to
growth but from growth to inequality (Kuznets’
hypothesis, 1955)
 The early stages of development exacerbate inequality
while later stages of development improve equality.
 Empirically this lacks support. (See e.g. Fields, 2001)
Empirical strategy - use lags of the explanatory
variables, which are pre-determined as
regressors
Also we use IV & GMM estimation approaches

14. Number of countries & billionaires on the list
Year Countries Number of billionaires/
billionaire families
1987 23 201
1992 31 340
1996 38 543
2002 42 568

15. Impact of wealth inequality, income inequality,
and poverty on economic growth (Benchmark)
Including the controls: GDP/ capita; Schooling & Price level of investment S.e. in parentheses * p < 0.10, ** p < 0.05, *** p<0.01
(1) (2) (3) (4) (5) (6)
Dependent variable: Growth rate in real GDP per capita
Wealth -0.132*
-0.547 -50.07***
Inequality (0.0771) (0.351) (13.27)
Politically unconnected -0.0464 -0.154 -48.98
wealth inequality (0.0714) (0.301) (36.52)
Politically connected -0.331*** -1.625*** -51.01**
wealth inequality (0.0965) (0.536) (22.79)
Income Inequality 0.000564 0.000763*
0.000498 0.000530 0.000753 0.000498
(0.000422) (0.000455) (0.000417) (0.000426) (0.000456) (0.000418)
Headcount Poverty 0.000301 0.000252 0.000353 0.000298 0.000243 0.000352
(0.000296) (0.000307) (0.000286) (0.000298) (0.000310) (0.000297)
N 160 149 160 160 149 160
R2
0.59 0.59 0.61 0.60 0.60 0.61

16. Comparing our results with Forbes (2000) (1/2)
S.e.in parentheses * p < .10, ** p <.05, *** p <.01
(1) (2) (3) (4)
Panel A: Assuming income and wealth inequality to have the same effect during the entire sample period
Income Inequality 0.000751 0.000991 0.00102 0.000947
(0.000886) (0.000830) (0.000858) (0.000840)
Wealth Inequality -0.154***
(GDP used for normalization) (0.0484)
Wealth Inequality -0.578***
(Physical capital used for normalization) (0.179)
Wealth Inequality -6.255***
(Population used for normalization) (2.061)
Number of observations 162 162 152 162
R2
0.39 0.45 0.44 0.42
F 5.343 8.717 8.740 7.138

17. Comparing our results with Forbes (2000) (2/2)
S.e.in parentheses * p < .10, ** p <.05, *** p <.01
(1) (2) (3) (4)
Panel B: Introducing dummy variable for first half of the sample period & corresponding interactions
Income Inequality 0.000419 0.000757 0.000698 0.000630
(0.000894) (0.000858) (0.000896) (0.000847)
Wealth Inequality -0.131**
(GDP used for normalization) (0.0493)
Wealth Inequality -0.525**
(Physical capital used for normalization) (0.201)
Wealth Inequality -7.771***
(Population used for normalization) (2.690)
Income Inequality X First half of sample period 0.000750**
0.000492 0.000614*
0.000742**
(0.000327) (0.000333) (0.000327) (0.000317)
Wealth Inequality X First half of sample period
(GDP used for normalization) 0.0691
Wealth Inequality X First half of sample period (0.0797)
(Physical capital used for normalization) -0.0110
Wealth Inequality X First half of sample period (0.324) -6.665
(Population used for normalization) (5.169)
Number of observations 162 162 152 162
R2
0.41 0.46 0.46 0.46
F 4.720 9.321 9.280 6.751

18. Robustness checks
RC1: Robustness to Forbes magazine’s choice of countries for the
billionaires in the data set
RC2: Use of alternative econometric approaches:
i. Random effects instead of a fixed effects specification
ii. Instrumental variables
iii. Dynamic panel methods of estimation (Arellano & Bond
difference-GMM and Blundell & Bond system-GMM)
RC3: Robustness to inclusion of additional control variables:
i. Adding a measure of institutional quality
ii. Controlling for the exchange rate
RC4: Using $1.25 per day per person as the poverty line

19. Impact of wealth inequality, income inequality,
and poverty on economic growth (Using RE)
Including the controls: GDP/ capita; Schooling & Price level of investment S.e. in parentheses * p < 0.10, ** p < 0.05, *** p<0.01
(1) (2) (3) (4) (5) (6)
Dependent variable: Growth rate in real GDP per capita
Wealth -0.162*
-0.652 -59.05***
Inequality (0.0962) (0.431) (14.67)
Politically unconnected -0.0145 0.0261 -17.52
wealth inequality (0.0688) (0.284) (48.52)
Politically connected -0.458***
-2.332***
-90.14***
wealth inequality (0.0600) (0.409) (20.85)
Income Inequality -0.000143 -0.0000126 -0.000145 -0.000171 -
0.0000258
-0.000151
(0.000441) (0.000513) (0.000435) (0.000438) (0.000509) (0.000437)
Headcount Poverty 0.000386*
0.000364 0.000417**
0.000434**
0.000406*
0.000425**
(0.000214) (0.000223) (0.000205) (0.000209) (0.000218) (0.000205)
N 160 149 160 160 149 160

20. Impact of wealth inequality, income inequality, and poverty
on GDP per capita (Using Blundell-Bond system-GMM
estimator) (Taking wealth inequality as pre-determined)
Including the controls: GDP/ capita; Schooling & Price level of investment S.e. in parentheses * p < 0.10, ** p < 0.05, *** p<0.01
(1) (2) (3) (4) (5) (6)
Dependent variable: Log of GDP per capita
Wealth -0.833* -3.154* -258.2***
Inequality (0.470) (1.898) (90.84)
Politically unconnected -0.389 -0.731 -94.39
wealth inequality (0.370) (1.210) (311.0)
Politically connected -2.092*** -10.04*** -403.3**
wealth inequality (0.629) (2.866) (162.0)
Income Inequality -0.000634 -0.000545 -0.000931 -0.000348 0.000497 -0.00107
(0.00255) (0.00267) (0.00256) (0.00302) (0.00296) (0.00306)
Headcount Poverty 0.00310* 0.00333* 0.00334** 0.00320* 0.00297 0.00343*
(0.00174) (0.00189) (0.00166) (0.00187) (0.00192) (0.00184)
Lagged log GDP per 0.991*** 1.031*** 1.000*** 0.987*** 1.003*** 1.000***
capita (0.0345) (0.0429) (0.0323) (0.0232) (0.0310) (0.0291)
N 161 149 161 161 149 161

21. Why is politically connected wealth inequality
detrimental?
Example 1: Birla family of India:
“The nationalists who later became free India’s power elite
rewarded the Birla family with lucrative contracts. After
independence, the Birlas continued their lavish contributions
to the ruling Congress Party. So accomplished are they in
manipulating the bureaucracy, and so vast their network of
intelligence, that they frequently obtain preemptive licenses,
enabling them to lock up exclusive rights for businesses as yet
unborn.” (Forbes, 1987)

22. Why is politically connected wealth inequality
detrimental?
Example 2: Tobacco billionaires in Indonesia:
Indonesia is the only country in Asia to have not signed the
WHO Framework Convention on Tobacco Control, a treaty
that as of September 2013 had been signed by 177 parties.
This is in spite of the fact that in Indonesia, Muslims
constitute 86 percent of the population and “smoking is either
completely prohibited in Islam or abhorrent to such a degree
as to be prohibited.” (WHO Regional Office for the Eastern
Mediterranean).
Indonesia’s average tobacco tax of 37 percent is the lowest in
Southeast Asia and well below the global average of 70 per
cent of the sales price (South China Morning Post, 2008).

23. Conclusions
1. High levels of wealth inequality appear to have
negative consequences for economic growth;
income inequality and headcount poverty do not
2. Wealth inequality arising on account of political
connections reduces economic growth v. wealth
inequality arising otherwise
3. Growth-related policy debate should focus on
distribution of wealth

25. Work that we have done since the paper
Also distinguish between self-made and inherited
billionaires. We split billionaires into three groups:
1.Self-made & politically unconnected (e.g. Bill Gates)
2.Self-made & politically connected (e.g. Mikhail
Fridman)
3.Inherited (e.g. Alice Walton)

26. Impact of wealth inequality, income inequality,
and poverty on economic growth
Including the controls: GDP/ capita; Schooling & Price level of investment S.e. in parentheses * p < 0.10, ** p < 0.05, *** p<0.01
(1) (2) (3)
Dependent variable: Growth rate in real GDP per capita
Self-made Politically Unconnected 0.0333 0.325* -21.40
Wealth Inequality (0.0335) (0.166) (30.88)
Self-made Politically Connected -0.287*** -1.327** -42.76
Wealth Inequality (0.0960) (0.561) (26.36)
Inherited Wealth Inequality -0.356* -2.413 -96.02
(0.199) (1.567) (64.52)
Income Inequality
0.000525 0.000695 0.000489
(0.000415) (0.000437) (0.000411)
Headcount Poverty
0.000402 0.000378 0.000403
(0.000289) (0.000301) (0.000286)
N
160 149 160
R2
0.62 0.62 0.61

27. Idea behind the Instrumental Variables (IV)
strategy
Wealth Inequality = Billionaire wealth / GDP
= “Average” wealth held by billionaire / Per capita income
* Number of billionaires / Population
Average wealth held by billionaires across countries
within the same region are correlated.
We predict wealth inequality in a given country by
predicting the average level of billionaire wealth in a
country.
e.g. A weighted average of the billionaire wealth in Canada
and Mexico is used as an instrument for the wealth
held by the “average” U.S. billionaire.

28. 1. Correlation with Davies et al. (2008)
measures
2. General pattern of increasing
inequality
SANITY CHECK ON WEALTH
INEQUALITY MEASURE

29. Wealth distribution data from the UNU –
WIDER data set & Forbes’ list of billionaires
Table 3: Wealth distribution data from the UNU-WIDER data set & Forbes’ list of
billionaires
Country
Share of
wealth going
to the top
decile
Year for the
wealth stats
Closest
year(s) in the
billionaire list
Billionaire wealth /
GDP in that year (s)
Australia 45 2002 2002 1.36%
Canada 53 1999 1996 & 2002 4.38%
… …. … … …
United
Kingdom
56 2000 2002 2.01%
United States 69.8 2001 2002 8.28%

30. Large variation in wealth inequality over time
with a general trend of increasing inequality

31. 1. Which countries show up on the list?
2. Correlation with ICRG Corruption Scores
3. Ranking of countries on Transparency
International’s Corruption Perceptions Index
4. Correlation with Easterly (2007)’s measure of
structural inequality
SANITY CHECK ON THE MEASURE
OF POLITICALLY CONNECTED
WEALTH INEQUALITY

32. Ranking of countries in terms of politically connected
matches priors
Countries that rank highest in terms of politically connected wealth inequality
1. Malaysia
2. Colombia
3. Indonesia
4. Thailand
5. Mexico
Other countries which just follow these include – Chile, South Korea,
Philippines, Argentina, and India. Italy is 11th
– the first European country
to appear on the list.
Median rank on TI’s Corruption Perceptions
Index: 32 /41 (1995) & 94/174 (2012)
Countries that rank lowest in terms of politically connected wealth inequality
1. Hong Kong
2. Netherlands
3. Singapore
4. Sweden
5. Switzerland and
6. United Kingdom
Median rank on TI’s Corruption Perceptions
Index: 9 /41 (1995) & 8/174 (2012)

33. Checking measure of political connectedness
with existing proxies for corruption
 Use data from International Country Risk Guide (ICRG)
 Specification tested:
 Politically connected wealth inequalityi = γ0 + γ 1 * ICRG
Corruption scorei + υi (3a)
 Politically connected wealth inequalityi,t = δ0 + δ 1 * ICRG
Corruption scorei,t + ηt+ υi,t (3b)

34. Political connectedness is highly correlated
with ICRG’s corruption index
Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01
(1) (2) (3) (4) (5) (6)
Panel A: Dependent variable: Politically connected billionaire wealth, normalized by GDP
ICRG Corruption
score
0.0364** 0.0426*** 0.0410*** 0.0226** 0.0347*** 0.0231***
(0.0159) (0.0136) (0.0148) (0.00837) (0.00645) (0.00784)
Constant -0.000419 -0.00155 0.00164 -0.00461* 0.0000741 0.00438
(0.00170) (0.00199) (0.00538) (0.00272) (0.00380) (0.00370)
R2
0.17 0.28 0.060 0.11 0.12
F 5.212 9.786 7.665 7.309 7.972
Panel B: Dependent variable: Politically unconnected billionaire wealth, normalized by GDP
ICRG Corruption
score
-0.0413**
(0.0158)
-0.0228
(0.0241)
-0.0000978
(0.0544)
-0.0638
(0.0449)
-0.0342*
(0.0200)
-0.0322
(0.0202)
Constant 0.0397*** 0.0330*** 0.0624** 0.0825*** 0.0377*** 0.0313***
(0.0104) (0.0112) (0.0252) (0.0262) (0.00961) (0.00835)
R2
0.093 0.017 0.000000061 0.050 0.078
F 6.791 0.891 0.00000323 2.019 2.825
Panel C: Dependent variable: Billionaire wealth, normalized by GDP
ICRG Corruption
score
-0.00487
(0.0208)
0.0198
(0.0290)
0.0409
(0.0574)
-0.0411
(0.0455)
0.000435
(0.0214)
-0.0106
(0.0219)
Constant 0.0393*** 0.0315*** 0.0640** 0.0779*** 0.0378*** 0.0368***
(0.0103) (0.0112) (0.0254) (0.0261) (0.0101) (0.00929)
R2
0.0012 0.013 0.0090 0.020 0.066
F 0.0547 0.466 0.508 0.817 2.288
Year(s) Included 1987 1992 1996 2002 All All
Econometric
Technique
OLS OLS OLS OLS Pooled OLS RE
N 22 31 37 41 131 131

35. Another validation of our measure of political
connected wealth inequality
 Easterly – JDE, 2007 – distinguishes between “structural
inequality” and “market based inequality”
 Follows from the work by Engerman and Sokoloff:
“… land endowments of Latin America lent themselves to commodities
featuring economies of scale and the use of slave labor and thus were
historically associated with high inequality. In contrast, the endowments of
North America lent themselves to commodities grown on family farms and thus
promoted the growth of a large middle class.”
 Uses this to develop a natural instrument for inequality: the
exogenous suitability of land for wheat versus sugarcane
 Measure used: “wheat–sugar ratio,” defined as
log [(1+share of arable land suitable for wheat) / (1+share of
arable land suitable for sugarcane)]

36. Correlation between LWHEATSUGAR and
components of wealth inequality
Correlation between the wheat–sugar ratio and wealth
inequality or components thereof
Correlation coefficient
between LWHEATSUGAR
&
p-
value
Politically connected wealth
inequality
- 0.425*** 0.010
Politically unconnected
wealth inequality
0.118 0.486
Wealth inequality - 0.148 0.382

37. 1. Is there variation in these measures over time?
2. How correlated are the two measures of
politically connected and politically
unconnected wealth inequality?
HOW REASONABLE IS IT TO
INTRODUCE THESE VARIABLES IN
THE WAY WE DO?

38. Large variation in wealth inequality over time
with a general trend of increasing inequality

39. Large variation in political connectedness
over time

40. Country rankings on “pol. conn.” & “pol. unconn.”
wealth inequality suggests they measure different
constructs
Countries that rank highest as per different classifications of billionaire
wealth
Politically unconnected billionaire
wealth/ GDP
Politically connected billionaire
wealth/ GDP
1. Hong Kong
2. Philippines
3. Singapore
4. Kuwait
5. Switzerland
1. Malaysia
2. Colombia
3. Indonesia
4. Thailand
5. Mexico

41. Patterns of correlation between components of wealth
inequality for 1987, 1992, 1996, and 2002