Debt, defaults and dogma

Debt, Defaults and Dogma
Politics and the Dynamics of Sovereign Debt Markets
Johnny Cotoc1
Alok Johri1
C´esar Sosa Padilla2
1McMaster University
2University of Notre Dame

Motivation
What accounts for high levels of debt, and high and volatile
interest rates in EMEs?
• interest rates have a significant effect on productivity and on
the amplification of shocks (eg. Mendoza-Yue 2012).
• the behavior of interest rates is an important factor
accounting for differences between the business cycles of
emerging and developed economies (eg. Uribe-Yue 2006,
Nuemeyer-Perri 2005).
• high debt levels are of particular relevance in emerging
economies because the high volatility of their borrowing cost
makes them vulnerable to crises.
1/44

Motivation
• In addition to econ. vars., political factors are often consider
to play a non-trivial role in debt markets and default decisions.
• Brazil, Ecuador, Argentina and Greece’s events are natural
examples.
• EME have low political stability (i.e. high turnover).
• Hence, political fluctuations are a natural candidate to explain
diff. in debt mkts and fiscal policy.
2/44

Motivation - what we do
1. Build a database covering 63 countries and 22 years:
• macro quantities and fiscal measures (WDI)
• country spreads (EMBI)
• political affiliations/leanings (IDB’s DPI)
2. Uncover (new) stylized facts regarding the influence of politics
on debt mkts and fiscal policy.
3. Propose a model of sov. default with endogenous political
fluctuations (turnover) to rationalize the facts.
3/44

Empirical evidence – What’s in (political) a label?
• Political data: party orientation wrt economic policy (based
on own parties’ descriptions).
• Left: parties deﬁned as communist, socialist, social
democratic, or left-wing.
• Right: parties deﬁned as conservative, Christian democratic, or
right-wing.
4/44

right-wing.
• We think of L and R as labels.
4/44

right-wing.
• We think of L and R as labels.
• Do these labels align w/ our typical understanding of L vs. R?
4/44

Table 1: Politics and Fiscal Policy
Left Right
E(T/Y ) 18.1% 15.4%
E(G/Y ) 15.3% 13.8%
5/44

Left Right
E(T/Y ) 18.1% 15.4%
E(G/Y ) 15.3% 13.8%
(i) L govs. collect more taxes than R govs.
5/44

Left Right
E(T/Y ) 18.1% 15.4%
E(G/Y ) 15.3% 13.8%
(i) L govs. collect more taxes than R govs.
(ii) L govs. have higher public spending than R govs.
5/44

Empirical evidence – Spreads and politics
Table 2: OLS estimation
Dep. variable: Spreads
(i) (ii)
constant 507.5∗∗∗
275.4
Political index 165.9∗∗
149.5∗
Debt/GDP 7.5∗∗∗
9.6∗∗∗
Y growth −28.4∗∗∗
−34.0∗∗∗
Y growth × Political index −46.2∗∗∗
−46.5∗∗∗
Year and region FE no yes
Adj. r2
.27 .28
Sample size 276 276
6/44

(i) (ii)
275.4
149.5∗
9.6∗∗∗
−34.0∗∗∗
−46.5∗∗∗
Adj. r2
.27 .28
Sample size 276 276
Fact 1: L govs. pay higher spreads than R govs.
6/44

(i) (ii)
275.4
149.5∗
9.6∗∗∗
−34.0∗∗∗
−46.5∗∗∗
Adj. r2
.27 .28
Sample size 276 276
Fact 1: L govs. pay higher spreads than R govs.
Fact 2: L govs. face more counter-cyclical spreads than R govs. 6/44

Table 3: Spread volatility
Left Right
σ(Spread) (in bps.) 594 481
σ(Spread)L/σ(Spread)R 1.24
7/44

Table 3: Spread volatility
Left Right
σ(Spread) (in bps.) 594 481
σ(Spread)L/σ(Spread)R 1.24
Fact 3: L govs. face more volatile spreads than R govs.
7/44

Model
• SOE w/ a continuum of households.
• Two political parties (L and R) which alternate in power.
• SOE trades bonds w/ competitive foreign lenders. Can’t
commit to repay.
• Time is discrete and goes on forever.
8/44

Model – Households
• Preferences: U(c, g) = αu(c) + (1 − α)u(g) (1)
u(x) =
x1−γ − 1
1 − γ
, for x = {c, g}.
• Endowment y follows Markov process w/ trans. fun. µ(y |y).
• Flow budget constraint:
c =
(1 − τ)y, if gov’t repays
(1 − τ)ya, if gov’t defaults
(2)
where ya ≤ y ∀y.
9/44

Model – Political turnover
• An election may occur in every period w/ prob π.
• Similar to Chatterjee and Eyigungor (2017) and Scholl (2017).
• If an election occurs, the incumbent may be replaced.
10/44

• An election may occur in every period w/ prob π.
• Similar to Chatterjee and Eyigungor (2017) and Scholl (2017).
• If an election occurs, the incumbent may be replaced.
• The re-election probability P depends on
Gov’t spending: ↑ g =⇒ ↑ P
Taxation: ↑ τ =⇒ ↓ P
10/44

Evidence on taxes:
• ↑ τ =⇒ ↓ P
Tillman and Park (2009), Beasley and Case (1995), Bosch and Sole-Olle (2004), Happy (1992),
Landon and Ryan (1997), Vermeir and Heyndels (2006).
11/44

Evidence on taxes:
• ↑ τ =⇒ ↓ P
• R parties are more strongly aﬀected by ↑ τ.
Tillman and Park (2009).
11/44

Evidence on taxes:
• ↑ τ =⇒ ↓ P
Evidence on gov’t spending:
• ↑ g =⇒ ↑ P
Shin (2016), Arulampalam et al. (2008), Cerda and Vergara (2008), Dahlberg and Johansson
(1999), De La O (2013), Johansson (2001), Landon and Ryan (1997).
11/44

Evidence on taxes:
• ↑ τ =⇒ ↓ P
Evidence on gov’t spending:
• ↑ g =⇒ ↑ P
Shin (2016), Arulampalam et al. (2008), Cerda and Vergara (2008), Dahlberg and Johansson
(1999), De La O (2013), Johansson (2001), Landon and Ryan (1997).
• L parties receive more political support from ↑ g
Shin (2016)
11/44

Based on the previous evidence, we guarantee that Pi (τ, g)
satisﬁes four properties:
P1. ↑ τ =⇒ ↓ P,
P2. R parties are more strongly aﬀected by ↑ τ,
P3. ↑ g =⇒ ↑ P, and
P4. L parties receive more political support from ↑ g
(more on P)
12/44

Model – Fiscal Policy
In case of repayment:
g + b = τy + b q(b , y) (3)
13/44

Model – Fiscal Policy
In case of repayment:
g + b = τy + b q(b , y) (3)
In case of default:
g = τya (4)
13/44

Model – Timing
• Incumbent enters period t in good credit standing and w/ b
debt
1. y is realized.
2. Default decision is made.
3. Consumption (c, g), taxation (τ) and new borrowing (b , if
not excluded) are chosen.
4. w/ prob π there is an election. w/ prob P(τ, g) incumbent
wins.
• end of period t
14/44

Model – Government’s Problem
Vi (b, y) = max V R
i (b, y), V D
i (y)
15/44

Vi (b, y) = max V R
i (b, y), V D
i (y)
V R
i (b, y) = max
g,τ,b
U(c, g) + β(1 − π)
y
Vi (b , y )µ(y , y)dy +
βπ Pi (τ, g)
y
Vi (b , y )µ(y , y)dy + (1 − Pi (τ, g))
y
¯Vi (b , y )µ(y , y)dy
15/44

Vi (b, y) = max V R
i (b, y), V D
i (y)
V R
i (b, y) = max
g,τ,b
U(c, g) + β(1 − π)
y
Vi (b , y )µ(y , y)dy +
βπ Pi (τ, g)
y
Vi (b , y )µ(y , y)dy + (1 − Pi (τ, g))
y
¯Vi (b , y )µ(y , y)dy
subject to
c = (1 − τ)y,
g = τy + qi (b , y)b − b.
15/44

V D
i (y) = max
g,τ
U(c, g)+
β(1 − π) θ
y
Vi (0, y )µ(y , y)dy + (1 − θ)
y
V D
i (y )µ(y , y)dy +
βπ Pi (τ, g) θ
y
Vi (0, y )µ(y , y)dy + (1 − θ)
y
V D
i (y )µ(y , y)dy +
(1 − Pi (τ, g)) θ
y
¯Vi (0, y )µ(y , y)dy + (1 − θ)
y
¯V D
i (y )µ(y , y)dy
16/44

V D
i (y) = max
g,τ
U(c, g)+
β(1 − π) θ
y
Vi (0, y )µ(y , y)dy + (1 − θ)
y
V D
i (y )µ(y , y)dy +
βπ Pi (τ, g) θ
y
Vi (0, y )µ(y , y)dy + (1 − θ)
y
V D
i (y )µ(y , y)dy +
(1 − Pi (τ, g)) θ
y
¯Vi (0, y )µ(y , y)dy + (1 − θ)
y
¯V D
i (y )µ(y , y)dy
subject to
c = (1 − τ)ya,
g = τya,
with
ya =



y if y ≤ ψ¯y,
ψ¯y otherwise,
16/44

Model – Default decision
The default policy of incumbent i is characterized by:
di (b, y) =



0 if V R
i (b, y) ≥ V D
i (y)
1 otherwise.
(5)
17/44

di (b, y) =



0 if V R
i (b, y) ≥ V D
i (y)
1 otherwise.
(5)
Default set:
Di (b) = {y ∈ Y : di (b, y) = 1}
17/44

di (b, y) =



0 if V R
i (b, y) ≥ V D
i (y)
1 otherwise.
(5)
Default set:
Di (b) = {y ∈ Y : di (b, y) = 1}
Default probability:
λi (b , y) = Di (b ) µ(y , y)dy
17/44

Model – party i not in power
• ¯Vi (b, y) depends on the opponent’s (−i) decision
• ¯V R
i (b, y) : value when the incumbent repays.
• ¯V D
i (y) : value when the incumbent defaults.
• The value of not being in power is just the discounted
expected probability of being back in power.
value functions
18/44

Foreign Lenders
• Risk neutral, deep-pocketed agents
• r∗ is the international risk-free interest rate
• Bonds are priced in a competitive market
19/44

Foreign Lenders
• Risk neutral, deep-pocketed agents
• r∗ is the international risk-free interest rate
• Bonds are priced in a competitive market
qi (b , y) = (1 − π)
1 − λi (b , y)
1 + r∗
+
π Pi (τ, g)
1 − λi (b , y)
1 + r∗
+ (1 − Pi (τ, g))
1 − λ−i (b , y)
1 + r∗
19/44

Calibration (i/ii)
• 1 model period ≡ 1 year.
• log(y ) = ρ log(y) + with E[ ] = 0 and E[ 2] = σ2.
• Pi (τ, g) = c(τ)
y − κi
φi
+ g
y
ωi
with i = {L, R}.
20/44

Calibration (i/ii)
• 1 model period ≡ 1 year.
• log(y ) = ρ log(y) + with E[ ] = 0 and E[ 2] = σ2.
• Pi (τ, g) = c(τ)
y − κi
φi
+ g
y
ωi
with i = {L, R}.
Table 4: Parameter values.
y autocorr. ρ 0.78 Estimation
SD of σ .034 Estimation
Borrower’s risk aversion γ 2 Prior literature
Risk-free rate r∗
0.04 Prior literature
Duration of defaults θ 0.154 Prior literature
Probability of elections π 0.25 Prior literature
20/44

Calibration (ii/ii)
Table 5: Parameter values set by SMM.
Discount factor β 0.67 Mean spread
Income losses ψ 0.888 Mean b/y
Weight on gov. spending α 0.025 Mean g/y
Political parameter φ 0.95 Mean T/Y
Political parameter κl 0.39 Mean TL/Y
Political parameter κR 0.42 Mean TR/Y
Political parameter ωL 0.75 Avg. prob. of reelection
21/44

Results – Road-map
1. Model ﬁt
2. Main results
3. Business Cycle statistics
4. Fiscal policy over the cycle
5. Quantitative relevance of endogenous reelections
22/44

Results – Model ﬁt
Table 6: Targeted moments.
Panel Data (1994-2015) Model
s (in bps.) 495 495
Mean debt-to-income 10% 10%
Re-election prob. 66% 66%
T 17% 17%
TL 18% 18%
TR 15% 15%
G 15% 16%
23/44

Results – Main results
Data Model Simulations
469.4∗∗∗
Political index 165.9∗∗∗
45.6∗∗∗
6.1∗∗∗
−23.5∗∗∗
−10.4∗∗∗
Year and region FE no no
Adj. r2
.27 .28
24/44

Data Model Simulations
469.4∗∗∗
Political index 165.9∗∗∗
45.6∗∗∗
6.1∗∗∗
−23.5∗∗∗
−10.4∗∗∗
Year and region FE no no
Adj. r2
.27 .28
Our model replicates Fact 1 and Fact 2.
24/44

Figure 1: Spread and debt menus.
sets
25/44

• Main takeaways:
1. L always faces worst spread-debt menus, and pays higher
spreads in eq. (Fact 1).
2. As income increases, L decreases spreads by more than R
(Fact 2).
• Mechanism:
(i) The default region is larger for L
(ii) The optimal mix of τ, g and b diﬀers across parties.
(i) − (ii) are determined simultaneously.
26/44

Figure 2: Default sets for L and R. 27/44

Figure 3: Taxes and Gov’t spending.
relative
28/44

Why is L’s default region larger?
1. As income decreases, both parties decrease debt.
29/44

2. But, R decreases debt more (i.e. it increases taxes less, as it
suﬀers more from doing so).
29/44

3. L relies more on debt → leaves it more vulnerable to future
bad shocks.
29/44

3. L relies more on debt → leaves it more vulnerable to future
bad shocks.
4. Hence, we get diﬀerent default regions =⇒ diﬀerent spread
menus.
29/44

Results – Business cycle statistics
Table 8: Non-targetted moments.
Panel Data (1993-2015) Model
sL (in bps.) 518 513
sR (in bps.) 463 472
σ(sL) (in bps.) 594.5 224.1
σ(sR ) (in bps.) 481.1 183.93
σ(sL)/σ(sR ) 1.24 1.22
σ(C)/σ(GDP) 1.53 1.46
σ(T)/σ(GDP) 3.66 3.45
σ(G)/σ(GDP) 2.34 2.51
ρ(GDP, G) 0.40 0.60
ρ(GDP, τ) -0.24 -0.32
ρ(GDP, s) -0.35 -0.44
ρ(GDP, TB/GDP) -0.12 -0.51
30/44

Results – Default incentives and bond prices
Figure 4: Bond price schedules. 31/44

• Country defaults in bad times → countercyclical spreads.
• q(b , y) increases w/ y, decreases with b .
32/44

• qL(b , y) ≤ qR(b , y) ∀(b , y).
32/44

• qL(b , y) ≤ qR(b , y) ∀(b , y).
Intuition:
1. L’s reelection relies more on g
2. g decreases as b increases.
3. As b , re-election of L and L becomes more
short-sighted.
4. Lenders understand this ⇒ charge higher premia.
32/44

Results – Fiscal policy over the cycle
Figure 5: Cyclical behavior of c and g. 33/44

Figure 6: Cyclical behavior of τ. 34/44

• corr(τ, y) < 0 ⇒ Procyclical ﬁscal policy
• good times: borrowing is cheap, so gov’t relies less on τ
• bad times: borrowing is expensive, so more reliant on τ
• during defaults: no borrowing, so even more procyclicality.
• Both c and g are procylical. Usual explanation.
35/44

Results – Equilibrium reelections
• P(τ, g). τ and g are endogenous
• We ﬁnd that P(τ, g) is increasing in income growth.
• Consistent w/ empirical evidence (Brender and Drazen, 2008).
• Intuition: as income grows, borrowing is cheaper, can aﬀord
both: τ ↓ and g ↑.
36/44

Results – Equilibrium reelections
Figure 7: Re-election probability and income growth. 37/44

Results – Endogenous vs. exogenous turnover
Figure 8: Endo. vs. exo. political turnover: P(τ, g) and b/y. 38/44

Figure 9: Endo. vs. exo. political turnover: borrowing policy functions. 39/44

• Endogenizing turnover has big implications for debt capacity.
• Incumbent’s re-election is decreasing in debt, conditional on
not defaulting.
• However, defaulting can increase re-election prob (frees up
resources to ↑ g and/or ↓ τ).
• Lenders anticipate this and restrict lending in the “endog.
turnover economy.”
40/44

Results – Exogenous turnover vs. No-turnover
Figure 10: Bond price schedule for “exogenous-turnover” and “no-turnover”
economies. 41/44

Results – Exogenous turnover vs. No-turnover
• “No-turnover economy” → Pi (τ, g) = ¯P = 1 ∀i
• Introducing (exo) turnover leads to a decrease in prices.
• Gov’t becomes de-facto more short-sighted.
• Results consistent w/ Cuadra and Sapriza (2008) and
Hatchondo et al. (2009).
42/44

Conclusion
• Politics, bond markets and ﬁscal policy interact in a
meaningful way.
• Established some (new) stylized facts.
• Propose a model which delivered the following features, all
consistent with the data:
• higher, more volatile and more counter-cyclical spreads for L
governments,
• endogeneous procyclical ﬁscal policy,
• political stability is (endogenously) increasing in output and
decreasing in debt,
43/44

Conclusion
• Linking back to the motivation.
• There are evidence and theories suggesting that:
“level and volatility of spreads matter for EME.”
• We’ve shown that political diﬀerences matter for both:
level and volatility of spreads.
44/44

Database of Political Institutions
Party orientation with respect to economic policy, coded based on
the description of the party in the sources:
• Right: for parties that are defined as conservative, Christian
democratic, or right-wing.
• Left: for parties that are defined as communist, socialist,
social democratic, or left-wing.
• 0: for all those cases which do not fit into the above
categories (i.e. party’s platform does not focus on economic
issues, or there are competing wings), or no information.
back

World Bank’s WDI
• Taxes: Tax revenue refers to compulsory transfers to the
central government for public purposes. Certain compulsory
transfers such as ﬁnes, penalties, and most social security
contributions are excluded. Refunds and corrections of
erroneously collected tax revenue are treated as negative
revenue.
• Government Spending: General government ﬁnal
consumption expenditure includes all government current
expenditures for purchases of goods and services (including
compensation of employees). It also includes most
expenditures on national defense and security, but excludes
government military expenditures that are capital formation.
back

P(τ, g) details
Pi (τ, g) =
c(τ)
y
− κ
φi
+
g
y
ωi
(6)
where i = {L, R}
P1. ↑ τ =⇒ ↓ P: ∂Pi
∂τ < 0 ∀ i
P2. R parties are more strongly aﬀected by ↑ τ: ∂PL
∂τ < ∂PR
∂τ
P3. ↑ g =⇒ ↑ P: ∂Pi
∂g > 0 ∀ i
P4. L parties receive more support from ↑ g: ∂PL
∂g > ∂PR
∂g
back

P(τ, g) details
Figure 11: Components of P(τ, g): taxes (←) and gov. spending (→).
back

Value functions while not in power (back)
¯V R
i (b, y) = β(1 − π)
y
¯Vi (b−i , y )µ(y , y)dy +
βπ (1 − P−i (τ−i , g−i ))
y
Vi (b−i , y )µ(y , y)dy +
P−i (τ−i , g−i ))
y
¯Vi (b−i , y )µ(y , y)dy
¯V D
i (y) = β(1−π) θ
y
¯Vi (0, y )µ(y , y)dy +(1−θ)
y
¯V D
i (y )µ(y , y)dy
+βπ (1−P−i (τ−i , g−i )) θ
y
Vi (0, y )µ(y , y)dy +(1−θ)
y
V D
i (y )µ(y , y)dy
+ P−i (τ−i , g−i ) θ
y
¯Vi (0, y )µ(y , y)dy + (1 − θ)
y
¯V D
i (y )µ(y , y)dy
¯Vi (b−i , y) =



¯V R
i (b−i , y) if d−i (b−i , y) = 0
¯V D
i (y) if d−i (b−i , y) = 1

Default sets
Figure 12: Default sets for L and R.
back

Taxes and gov’t spending (relative)
Figure 13: Relative movements in τ and g.
back

Debt, defaults and dogma

Recomendados

Recomendados

Más contenido relacionado

Similar a Debt, defaults and dogma

Similar a Debt, defaults and dogma (11)

Más de ADEMU_Project

Más de ADEMU_Project (20)

Último

Último (20)

Debt, defaults and dogma