the risky capital of emerging marketsinasimonovska.weebly.com/uploads/1/3/2/5/1325220/... · the...
TRANSCRIPT
The Risky Capital of Emerging Markets
Joel M. David Espen Henriksen Ina Simonovska
USC UCD UCD & NBER
February 24, 2014
1 / 27
Lucas Paradox Revisited
6 7 8 9 10 11 12
0.05
0.1
0.15
0.2
0.25
0.3
Ret
urn
to U
S In
vest
or
Log GDP per Worker
slope=−0.0178***
Lucas Paradox: Should capital flow to poor countries? Lit Review
Not necessarily... Riskier capital in poor countries ⇒ higher average returns.
But can differences in risk quantitatively account for differences in returns?
Yes!
Risk (measured properly) accounts for > 3/4 of difference in returns between
poorest countries and US.
2 / 27
What we do...
First pass: workhorse CCAPM... works qualitatively; not quantitatively
⇒ Lucas Paradox: just another asset pricing puzzle
Towards a resolution:
Key Building blocks
1 Aguiar and Gopinath (2007): shocks to trend growth key in
poor/emerging market BC’s +
2 Bansal and Yaron (2004): implications of these shocks for asset
prices/returns, i.e., compensation for “long-run risks”
Our Approach
1 Endowment economy in spirit of BY; multiple international “assets”
⇒ assets differ in shocks to trend of cash flows from capital investment
2 Use TS moments in cash flows (“dividends”) to infer key parameters
⇒ Follow portfolio approach—3, 5, 10 “bundles”
3 Use estimated parameters to assess implications for capital returns
3 / 27
What we find...
Investments in poor countries are more exposed to long-run risk
⇒ US investor demands higher average returns
• Risk accounts for > 3/4 of return differential across portfolios
• Results are robust to levels of disaggregation
• LRR key; SRR implies negative risk premia abroad
• Differences in returns driven by differences in volatilities of cash flows
Some evidence of return convergence in mid-late 1990s
• Recalibrating model to this sub-period closes gap as well...(though noisy)
• Deeper explanation?
4 / 27
Measuring Returns to Capital for US Investor
Environment: N countries; two sectors: K is freely traded, C is not.
Representative US agent consumes, buys capital in US, invests it in country j .
Return from investment in j for US investor (builds on Caselli and Feyrer, ’07):
Rjt = α
Pusy,tY
jt
Kjt︸ ︷︷ ︸
CF j in Cus
1
Pusi,t
︸ ︷︷ ︸dividend yield, D/P
+(1− δjt
) Pusi,t
Pusi,t−1︸ ︷︷ ︸
capital gains
• Pusi ,P
usy is US price of K, Y relative to US price of C
5 / 27
Returns Revisited: “Bundles” of countries
Assume common α = 0.3.
Data:
• PWT 8.0, 1950-2009: real K j ,Y j , δj
• BEA 1950-2009:• Pus
i—price index of equipment + structures
• Pusy —price index of output
• Pusc —price index of non-durables + services
Countries bundled according to mean output per worker over period. Portfolios
6 7 8 9 10 11 12
0.05
0.1
0.15
0.2
0.25
0.3
Ret
urn
to C
apita
l
Log GDP per Worker
1 2
3 US
6 / 27
Lucas Paradox: Just Another Asset Pricing Puzzle
With CRRA preferences:
E [r ejt ] ≈ γcov(rjt ,∆ct)
• r ejt ≡ rjt − rft is excess (net) return on portfolio j over 3-month t-bill at t
• ∆ct is US (ND+S) consumption growth during 1950-2009
r̂ e
Portfolio r e cov (r ,∆c) γ = 10 907
1 11.60 0.00015 0.15 13.79
2 10.00 0.00009 0.09 8.33
3 6.93 0.00005 0.05 4.28
US 4.80 0.00004 0.04 3.29
Spread: 1-US 6.80 0.00012 0.12 10.50
Derivation Alternative Aggregation
7 / 27
Toward a Resolution
Key Building blocks:
1 Aguiar and Gopinath (2007): shocks to trend growth key in
poor/emerging market BC’s +
2 Bansal and Yaron (2004): implications of these shocks for asset
prices/returns, i.e., compensation for “long-run risks”
1950 1960 1970 1980 1990 2000 2010
100
200
300
400
600
log
Cas
h F
low
s, 1
951=
100
United States
1950 1960 1970 1980 1990 2000 2010
100
200
300
400
600
log
Cas
h F
low
s, 1
951=
100
India
0.00 0.02 0.04 0.06
010
2030
4050
6070
10−Year Avg. Cash−Flow Growth Rates
Den
sity
United StatesIndia
8 / 27
Model
1 Endowment economy in spirit of BY; multiple international “assets”
⇒ assets differ in shocks to trend of cash flows from capital investment
2 Use TS moments in cash flows (“dividends”) to infer key parameters
⇒ Follow portfolio approach—3, 5, 10 “bundles”
3 Use estimated parameters to assess implications for capital returns
In other words:
Is US investor’s SDF w LRR consistent with cross-section of returns?
9 / 27
Model: Processes
Representative US-based investor endowed with c and dj , j indexes asset:
∆ct+1 = µc + xt + σηt+1
xt+1 = ρxt + ϕeσet+1
∆dj,t+1 = µj + φjxt + πjσηt+1 + ϕjσuj,t+1
ηt+1, et+1, uj,t+1 ∼ NIID (0, 1)
Portfolio-specific variables:
• µj : mean growth rate
• φj : exposure to trend shock in US consumption growth
• πj : exposure to transitory shock in US consumption growth
• ϕj : volatility of dividend shock
10 / 27
Model: US Investor’s Preferences
Representative US-based investor’s preferences
Vt =
[(1− β)C
ψ−1ψ
t + βνt (Vt+1)ψ−1ψ
] ψψ−1
νt (Vt+1) =[Et
(V
1−γ
t+1
)] 11−γ
• νt (Vt+1) is certainty equivalent function
• ψ is intertemporal elasticity of substitution
• γ ≥ 0 is risk aversion coefficient
• β < 1 is discount rate
• CRRA special case iff γ = 1/ψ
11 / 27
Risks in the Short and Long Run
Key asset-pricing equation for any asset j :
Et [Must+1R
jt+1],
Must+1 is US investor’s SDF. SDF
Risk premium for asset j :
E [Rejt ] ≈
(φj −
1
ψ
)(γ −
1
ψ
)κj
1− κjρ
κc
1− κcρϕ2
eσ2
︸ ︷︷ ︸long-run risk
+ γπjσ2
(1
1
)
︸ ︷︷ ︸short-run risk
where:
• κj = fd (θj , θc)
• κc = fc (θc)
Solving Model
12 / 27
Model: Calibration of Consumption Growth Parameters
Recall:
∆ct+1 = µc + xt + σηt+1
xt+1 = ρxt + ϕeσet+1
ρ = 0.93 (Ferson, 2013; Bansal, Kiku, and Yaron, 2012)
E(∆ct) = µc
cov (∆ct ,∆ct+1) = ρϕ2
eσ2
1− ρ2
var(∆ct) =
(ϕ2
e
1− ρ2+ 1
)σ2
13 / 27
Model: Calibration of Dividend Growth Parameters
Recall:
∆dj,t+1 = µj + φjxt + πjσηt+1 + ϕjσuj,t+1
xt+1 = ρxt + ϕeσet+1
E(∆dj,t) = µj
φj =
√cov (∆dj,t ,∆dj,t+1)
cov (∆ct ,∆ct+1)
cov (∆dj,t ,∆cj,t) = φjϕ2
eσ2
1− ρ2+ πjσ
2
var(∆dj,t) = φ2j
ϕ2eσ
2
1− ρ2+(π2j + ϕ2
j
)σ2
14 / 27
An Example: US VS India
Data: E(Rus) = 6.0; E(R in) = 9.0
acov(∆dt ,∆dt+1) =
acorr(∆dt ,∆dt+1) · var(∆dt)
−0.1 −0.05 0 0.05 0.1 0.15−0.1
−0.08
−0.06
−0.04
−0.02
0
0.02
0.04
0.06
0.08
0.1
∆ d t+
1
∆ dt
IndiaUnited States
corr=0.282**
corr=0.279**
1960 1970 1980 1990 2000 2010
−0.1
−0.05
0
0.05
0.1
∆ d t
IndiaUnited States
var(∆ dus
)=0.00067
var(∆ dind
)=0.00211
⇒ Model: E(Rus) = 6.0; E(R in) = 8.3
15 / 27
Data: Moments
Consumption E [∆ct ] cov (∆ct+1,∆ct) std (∆ct)
0.019 0.00024 0.022
Portfolio E [∆dt ]√
cov(∆dt+1,∆dt )
cov(∆ct+1,∆ct )cov (∆dt ,∆ct) std (∆dt)
1 −0.017 5.94 0.00013 0.085
2 −0.015 6.12 0.00011 0.075
3 −0.012 4.22 0.00016 0.062
US −0.006 2.19 0.00018 0.026
• US Consumption TS: 1929-2009
• Portfolio TS: vary by portfolio/country based on PWT availability
• Portfolio moment computed as mean of country-specific moments
Observe:
• cov(∆d ,∆c) ≈ across portfolios
• std(∆d) falling in income/worker
• acov(∆d) high for portfs 1/2 compared to 3/US
Rebalanced Portfolios16 / 27
Calibrated Model: Parameters
∆ct+1 = µc + xt + σηt+1
∆dj,t+1 = µj + φjxt + πjσηt+1 + ϕjσuj,t+1
xt+1 = ρxt + ϕeσet+1
Preferences: γ = 10 ψ = 1.5 β = 0.99
Consumption: ρ = 0.93 µc = 0.019 σ = 0.015 ϕe = 0.39
Portfolio µd φ ϕd π
1 −0.017 5.94 7.75 −1.92
2 −0.015 6.12 6.45 −1.96
3 −0.012 4.22 5.97 −0.34
US −0.006 2.19 2.12 0.98
Observe: φj high for portfs 1/2 compared to 3/US.
Question: What are the implications for returns to capital?
Rebalanced Portfolios 17 / 27
Predicted VS Actual Returns: 3 Bundles
Portfolio r̂ r
1 11.75 12.82
2 12.00 11.22
3 9.47 8.15
US 5.94 6.02
Average: 9.79 9.55
Spread: 1-US 5.81 6.80
Percent of actual 85
• Predicted returns ≈ actual returns
• Returns for portf 1 ≈ 2 > returns for portf 3, US
PDR VARR
18 / 27
Predicted VS Actual Returns: 3, 5, 10 Bundles
3 Portfolios 5 Portfolios 10 Portfolios
r̂ r r̂ r r̂ r
1 11.75 12.82 1 12.06 14.16 1 12.95 15.59
2 12.00 11.22 2 12.42 11.49 2 11.33 13.26
3 9.47 8.15 3 10.87 10.84 3 10.87 11.22
US 5.94 6.02 4 11.16 10.57 4 13.60 11.64
5 8.53 6.64 5 11.04 9.95
US 5.94 6.02 6 10.56 12.08
7 11.81 10.11
8 10.37 11.03
9 9.39 7.31
10 7.34 5.99
US 5.94 6.02
Average: 9.79 9.55 10.16 9.95 10.47 10.38
Spread: 1-US 5.81 6.80 6.13 8.14 7.01 9.57
Percent of actual 85 75 73
corr (̂r , r) 0.94 0.91 0.83
Corr btwn predicted and actual returns high at different levels of aggregation.
Rebalanced Portfolios
19 / 27
Decomposition: Long VS Short Run Risk
Actual Predicted
Portfolio r r r f r esr r elr
1 12.82 11.75 = 1.27 + −0.43 + 10.91
2 11.22 12.00 = 1.27 + −0.44 + 11.17
3 8.15 9.47 = 1.27 + −0.08 + 8.27
US 6.02 5.94 = 1.27 + 0.22 + 4.44
High returns mainly driven by long-run risk.
20 / 27
High Returns Driven by High Volatility of Cash Flows
Actual Predicted r
Portfolio r acorr(∆d) std(∆d) baseline acorr(∆d) = US std(∆d) = US
1 12.82 0.22 0.085 11.75 12.68 4.57
2 11.22 0.28 0.075 12.00 12.06 5.32
3 8.15 0.19 0.062 9.47 10.84 4.64
US 6.02 0.28 0.026 5.94 5.94 5.94
Recall that φj =
√cov(∆dj,t ,∆dj,t+1)cov(∆ct ,∆ct+1)
and has first-order impact on returns.
Counterfactuals:
• Column 6: If acorr(∆d j ) = acorr(∆dus), returns remain high.
• Column 7: If std(∆d j) = std(∆dus), returns fall.
⇒ High returns driven by high volatility of cash flows.
21 / 27
Predicted VS Actual Returns by Country
7.5 8 8.5 9 9.5 10 10.5 11
0.05
0.1
0.15
0.2
0.25
Log GDP per Worker
Act
ual R
etur
n slope=−0.023***
7.5 8 8.5 9 9.5 10 10.5 11
0.05
0.1
0.15
0.2
0.25
Log GDP per Worker
Pre
dict
ed R
etur
n
slope=−0.011**
0.05 0.1 0.15 0.2 0.25
0.05
0.1
0.15
0.2
0.25
Predicted Return
Act
ual R
etur
n
corr=0.26**
69 countries s.t.: (i) in PWT in or prior to 1961; (ii) autocov(∆d ,∆d+1) ≥ 0
Portfolios22 / 27
Predicted VS Actual Returns: 69 Countries Bundled Up
3 Portfolios 5 Portfolios 10 Portfolios
r̂ r r̂ r r̂ r
1 11.50 12.20 1 13.03 13.93 1 13.03 15.37
2 9.58 9.36 2 10.18 10.47 2 13.03 13.32
3 9.50 7.48 3 9.14 9.29 3 9.38 10.34
US 5.94 6.02 4 10.63 9.31 4 10.97 10.59
5 8.34 5.74 5 8.72 8.96
US 5.94 6.02 6 9.52 9.59
7 9.71 9.21
8 11.68 9.42
9 8.83 6.41
10 7.69 4.86
US 5.94 6.02
Average: 9.13 8.76 9.54 9.13 9.86 9.46
Spread: 1-US 5.56 6.18 7.09 7.91 7.09 9.35
Percent of actual 90 90 76
corr (̂r , r) 0.90 0.91 0.88
23 / 27
TS Returns on Capital: 69 Countries
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22R
etur
n on
cap
ital
Portfolio 1Portfolio 2Portfolio 3United States
Open, Fixed Distribution Open, New Distribution Very Open, Fixed Distribution
24 / 27
Calibration of Model to Post-1995 Data, 69 Countries
1950-2009
Portfolio√
acov(∆d)acov(∆c)
acorr(∆d) std(∆d) cov(∆d,∆c) corr(∆d,∆c)
1 5.94 0.22 0.085 0.00013 0.13
2 6.12 0.28 0.075 0.00011 0.12
3 4.22 0.19 0.062 0.00016 0.22
US 2.19 0.28 0.026 0.00018 0.60
1995-2009√acov(∆d)acov(∆c) acorr(∆d) std(∆d) cov(∆d,∆c) corr(∆d,∆c)
1 2.82 0.10 0.060 0.00004 0.06
2 3.21 0.18 0.051 −0.00003 −0.05
3 6.19 0.47 0.061 0.00026 0.41
US 1.61 0.48 0.016 0.00007 0.40
• acov(∆d) for portfolios 1, 2, and US falls...
• acov(∆d) for portfolio 3 rises (driven by Eurozone countries)
Question: What are the implications for returns to capital?
25 / 27
Predicted VS Actual Returns After 1995, 69 Countries
1950-2009 1995-2009
Portfolio r r̂ r r̂ r̂acorrd r̂stdd r̂corrcd
1 12.82 11.75 7.34 6.26 8.43 8.40 6.37
2 11.22 12.00 6.09 6.74 8.00 9.10 6.97
3 8.15 9.47 5.26 12.11 9.09 12.28 11.87
US 6.02 5.94 4.79 4.07 3.16 6.34 4.17
Answer:
Returns for portfolios 1, 2, US fall in model/data... But returns for 3 rise.
Counterfactuals: Focus on portfolio 3
• Column 6: If acorr(∆d j ) = acorr(∆dj51−08), returns < re-calibrated model.
• 7/8: If std(∆d j) = std(∆dj51−08) or corr(∆c,∆d j ) = corr(∆c,∆d
j51−08),
returns ≈ re-calibrated model.
⇒ High returns for portfolio 3 post 1995 driven by high autocorr of cash flows.
Any idea why?...
26 / 27
Conclusion
Can differences in risk quantitatively account for differences in returns?
Yes!
Risk (measured properly) accounts for > 3/4 of difference in returns between
poorest countries and US.
Key intuition: US investor does not invest in poor countries because shocks to
trend growth of dividends are larger there.
Ongoing work:
• Explore return volatility and co-movement in model/data
• Implications about K/Y
Future work:
• Time-varying volatility
• Country/portfolio-specific permanent components of cash flow growth
• Time-series analysis
27 / 27
Closely Related Literature
Lucas Paradox
Lucas (1990), Caselli and Feyrer (2007)
Asset-pricing puzzle
Mehra and Prescott (1985)
Solving AP puzzles with long-run risk
Bansal and Yaron (2004), Colacito and Croce (2011), Lustig and Verdelhan
(2007), Borri and Verdelhan (2011)
Long-run risk around the world
Aguiar and Gopinath (2007), Naoussi and Tripier (2013), Nakamura, Sergeyev,
and Steinsson (2012), Lewis and Liu (2012)
Intro
– / 27
Measuring Returns to Capital for US Investor
Environment: N countries; two sectors: K is freely traded, C is not.
Representative US consumer supplies unit of labor inelastically and maximizes
E0
∞∑
t=0
βtU (C us
t ) s.t.
Cust + P
usi,t
K us,ust+1 +
∑
j 6=us
Kus,jt+1
= Wust +
[αPus
y,tYust
K ust
+ Pusi,t (1− δust )
]K
us,ust +
∑
j 6=us
[αPus
y,tYjt
Kjt
+ Pusi,t
(1− δjt
)]
Kus,jt
Euler equation for capital invested in any country j (including US):
1 = Et
[
βU ′ (C us
t+1)
U ′ (C ust )
(
αPusy,t+1Y
ust+1
K ust+1P
usi,t
+ (1− δust )Pusi,t+1
Pusi,t
)]
= Et [Must+1R
ust+1]
= Et
[βU ′ (C us
t+1)
U ′ (C ust )
(αPusy,t+1Y
jt+1
Kjt+1P
usi,t
+(1− δjt
) Pusi,t+1
Pusi,t
)]= Et
[M
ust+1R
jt+1
]
Intro
– / 27
Derivation: Risk Measure under CRRA Preferences
• Rep. US investor can invest in domestic and international capital markets.
• CRRA preferences: U(c) = c1−γ−11−γ
, γ ≥ 0 is risk aversion coefficient.
• Stochastic Discount Factor (SDF) is Mt+1 = β(
ct+1
ct
)−γ
.
• Return on (risky) investment in market j is Rj ; risk-free rate is Rf .
No-arbitrage Euler equation for any asset j (incl. risk-free asset):
1 = Et [Mt+1Rjt+1] (1)
1 = Rft+1Et [Mt+1] (2)
(1) and (2) ⇒ Risk premium:
EtRjt+1 − Rft+1 = −covt (Mt+1,Rjt+1)
EtMt+1(3)
Intuition: Asset that pays off when MU is low demands higher risk premium.
Linearize SDF around uncond. mean and use (3) to derive testable prediction.
CRRA
–/ 27
Lucas Paradox: Just Another Asset Pricing Puzzle
5 10 15x 10
−5
0.06
0.08
0.1
0.12
cov(r,∆ c)
re
γ=907
0.5 1 1.5 2x 10
−4
0.06
0.08
0.1
0.12
0.14
cov(r,∆ c)
re
γ=873
0 1 2x 10
−4
0.06
0.08
0.1
0.12
0.14
cov(r,∆ c)
re
γ=868
−5 0 5x 10
−4
0
0.1
0.2
cov(r,∆ c)re
γ=493
CRRA
–/ 27
Lucas Paradox: Just Another Asset Pricing Puzzle
What if ∆c is iid and prefs are CRRA (simple version of our model)?
∆ct+1 = µc + σηt+1
∆dj,t+1 = µj + ϕjσuj,t+1 + πjσηt+1
⇒ E[Re
jt
]= γπjσ
2 and πjσ2 = cov (∆d ,∆c)
r̂ e
Portfolio r e cov (∆d ,∆c) γ = 10 535
1 11.60 0.00013 0.13 6.76
2 10.00 0.00011 0.11 5.63
3 6.93 0.00016 0.16 8.38
US 4.80 0.00018 0.18 9.54
Spread: 1-US 6.80 −0.00005 −0.05 −2.78
CRRA
–/ 27
Data: Moments—Annually Rebalanced Portfolios
Portfolio E [∆dt ]√
cov(∆dt+1,∆dt )
cov(∆ct+1,∆ct )cov (∆dt ,∆ct) std (∆dt)
1 −0.019 6.16 0.00010 0.082
2 −0.017 5.69 0.00012 0.072
3 −0.011 3.92 0.00015 0.058
US −0.006 2.19 0.00018 0.026
Observe:
• cov(∆d ,∆c) ≈ across portfolios
• std(∆d) falling in income/worker
• acov(∆d) falling in income/worker
Benchmark Moments
– / 27
Calibrated Model: Parameters—Annually Rebalanced Portfolios
Portfolio µd φ ϕd π
1 −0.019 6.16 7.12 −2.53
2 −0.017 5.69 6.36 −1.76
3 −0.011 3.92 5.59 −0.32
US −0.006 2.19 2.12 0.98
Observe: φj falling in income/worker.
Benchmark Parameters
– / 27
Predicted VS Actual Returns—Annually Rebalanced Portfolios
3 Portfolios 5 Portfolios 10 Portfolios
r̂ r r̂ r r̂ r
1 11.86 12.27 1 11.62 13.57 1 11.32 14.92
2 11.33 10.45 2 12.07 11.00 2 11.77 12.65
3 8.98 7.58 3 11.39 10.26 3 12.44 10.58
US 5.94 6.02 4 10.05 8.65 4 11.69 11.47
5 8.23 7.17 5 11.71 10.29
US 5.94 6.02 6 11.04 10.23
7 10.79 9.24
8 9.10 8.06
9 8.93 6.88
10 7.17 7.60
US 5.94 6.02
Average: 9.53 9.08 9.88 9.44 10.17 9.81
Spread: 1-US 5.92 6.25 5.68 7.54 5.38 8.90
Percent of actual 95 75 60
corr (̂r , r) 0.95 0.88 0.79
Results
– / 27
Mean p-d ratios
3 Portfolios 5 Portfolios 10 Portfolios
p − d p − d p − d
1 2.11 1 2.09 1 2.03
2 2.10 2 2.06 2 2.14
3 2.31 3 2.19 3 2.19
US 2.73 4 2.17 4 1.96
5 2.40 5 2.17
US 2.73 6 2.21
7 2.10
8 2.25
9 2.30
10 2.56
US 2.73
Spread: 1-US: −0.62 −0.64 −0.70
Results
– / 27
Solving the Model
Standard approach; approximations to make exercise computationally feasible
Note that zt = log(
Pt
Dt
)is enough to characterize returns
• For each asset j , and asset that pays off aggregate consumption
To solve:
• Conjecture: zjt = A0j + A1jxt
• Approximate log returns: rjt+1 = κ0j + κ1jzjt+1 +∆djt+1 − zjt
where κ’s depend on z j
• Restrictions from Euler equation gives excess return (as in text) and rf rate
• Solve numerically for z j = A0j (z j ) and for A1j (z j )
Results
– / 27
List of Countries by Portfolio
Portfolio 1:Angola 1971
Bangladesh 1960
Benin 1980
Bhutan 1980
Bolivia 1951
Burkina Faso 1960
Burundi 1980
Cambodia 1971
Cameroon 1961
Cape Verde 1980
Central African Rep. 1980
Chad 1980
China 1953
Comoros 1980
Congo - Brazzaville 1980
Congo - Kinshasa 1971
El Salvador 1975
Gambia 1980
Ghana 1960
Guinea 1980
Honduras 1970
India 1960
Indonesia 1961
Ivory Coast 1961
Kenya 1960
Kyrgyzstan 1991
Laos 1980
Lesotho 1980
Liberia 1980
Madagascar 1961
Mali 1961
Mauritania 1980
Moldova 1991
Mongolia 1980
Nepal 1980
Niger 1961
Nigeria 1960
Pakistan 1960
Sao Tome and Pr. 1980
Senegal 1961
Tajikistan 1991
Tanzania 1961
Thailand 1960
Togo 1980
Uganda 1960
Vietnam 1971
Yemen 1990
Zambia 1960
Portfolio 2:Albania 1971
Argentina 1951
Armenia 1991
Azerbaijan 1991
Belarus 1991
Belize 1980
Bosnia 1991
Botswana 1980
Brazil 1951
Bulgaria 1971
Chile 1952
Colombia 1951
Costa Rica 1951
Djibouti 1980
Dominican Rep. 1960
Ecuador 1952
Fiji 1980
Georgia 1991
Grenada 1988
Guatemala 1951
Iran 1960
Jamaica 1960
Jordan 1960
Kazakhstan 1991
Latvia 1991
Malaysia 1960
Mauritius 1980
Morocco 1960
Namibia 1980
Panama 1969
Paraguay 1970
Peru 1951
Philippines 1960
Poland 1971
Portugal 1951
Romania 1989
Russia 1991
Saint Lucia 1971
Saint Vincent 1980
Serbia 1991
South Africa 1960
South Korea 1960
Sri Lanka 1960
Swaziland 1980
Syria 1961
Tunisia 1961
Turkey 1951
Turkmenistan 1991
Ukraine 1991
Uzbekistan 1991
– / 27
List of Countries by Portfolio
Portfolio 3:Antigua 2001
Australia 1951
Austria 1951
Bahamas 1973
Bahrain 1971
Barbados 1961
Belgium 1951
Brunei 1980
Canada 1951
Croatia 1991
Cyprus 1960
Czech Republic 1991
Denmark 1951
Estonia 1991
Finland 1951
France 1951
Gabon 1980
Germany 1970
Greece 1952
Hong Kong 1961
Hungary 1971
Iceland 1956
Ireland 1951
Israel 1960
Italy 1951
Japan 1951
Lithuania 1991
Luxembourg 1951
Macao 1980
Macedonia 1991
Mexico 1951
Montenegro 2005
Netherlands 1951
New Zealand 1951
Norway 1951
Oman 1971
Qatar 1971
Singapore 1961
Slovakia 1991
Slovenia 1991
Spain 1951
Sweden 1951
Switzerland 1951
Taiwan 1960
Trinidad and Tobago 1960
United Kingdom 1951
Uruguay 1960
Venezuela 1951
Portfolio 4:United States 1951
Data
– / 27
List of 69 Countries by Portfolio
Portfolio 1:Bangladesh 1960
Bolivia 1951
Cameroon 1961
China 1953
Ghana 1960
India 1960
Indonesia 1961
Ivory Coast 1961
Kenya 1960
Madagascar 1961
Nigeria 1960
Pakistan 1960
Philippines 1960
Senegal 1961
Sri Lanka 1960
Tanzania 1961
Thailand 1960
Uganda 1960
Zambia 1960
Portfolio 2:Argentina 1951
Brazil 1951
Chile 1952
Colombia 1951
Costa Rica 1951
Dominican Rep. 1960
Ecuador 1952
Greece 1952
Guatemala 1951
Iran 1960
Jamaica 1960
Japan 1951
Jordan 1960
Malaysia 1960
Mexico 1951
Morocco 1960
Peru 1951
Portugal 1951
South Africa 1960
South Korea 1960
Taiwan 1960
Tunisia 1961
Turkey 1951
Uruguay 1960
Venezuela 1951
Portfolio 3:Australia 1951
Austria 1951
Barbados 1961
Belgium 1951
Canada 1951
Cyprus 1960
Denmark 1951
Finland 1951
France 1951
Hong Kong 1961
Iceland 1956
Ireland 1951
Israel 1960
Italy 1951
Luxembourg 1951
Netherlands 1951
New Zealand 1951
Norway 1951
Singapore 1961
Spain 1951
Sweden 1951
Switzerland 1951
Trinidad and Tobago 1960
United Kingdom 1951
Portfolio 4:United States 1951
Results
– / 27
TS Returns on Capital: 34 Open Countries, Fixed Income Distribution
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18R
etur
n on
cap
ital
Portfolio 1Portfolio 2Portfolio 3United States
TS Plot
– / 27
TS Returns on Capital: 34 Open Countries
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16R
etur
n on
cap
ital
Portfolio 1Portfolio 2Portfolio 3United States
TS Plot
– / 27
TS Returns on Capital: 17 Very Open Countries, Fixed Income Distribution
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2R
etur
n on
cap
ital
Portfolio 1Portfolio 2Portfolio 3United States
TS Plot
– / 27
Return variance
std (∆d) std (r) std (̂r)
1 0.085 0.054 0.125
2 0.075 0.047 0.120
3 0.062 0.039 0.094
US 0.026 0.025 0.046
• Predicted variance > actual variance of returns
• Variance of cash flow growth > variance of returns
Results
– / 27
Details: SDF
In logs,
must+1 = θ log β −
θ
ψ∆ct+1 + (θ − 1) rct+1
where θ = 1−γ
1− 1ψ
and rct+1 is return on asset that pays consumption as dividend.
Model
– / 27