1.
The Blind Side of Public Debt Spikes
Laura Jaramillo, Carlos Mulas-Granados
and Elijah Kimani
Fiscal Affairs Department
International Monetary Fund
ADEMU conference-Bonn, June 7th, 2016

2.
Presentation outline
2
I. Motivation for this project
II. Characterizing Debt Spikes
III. What drives debt spikes?
IV. What is behind SFAs?
V. What are the consequences of sizable debt
spikes?
VI. Can more realistic SFAs improve debt
forecasts?
VII.Concluding remarks

3.
3
I. Motivation for
this project

4.
4
I. Motivation for this project
Contrary to popular belief, large debt increases are not driven by high
primary deficits…WHAT IS BEHIND DEBT SPIKES?
-40.0
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
1973-1987 1988-2007 2008-2015
Change in debt to GDP
Cumulative primary deficits
Source: FAD Historical Public Debt Database, Mauro (2013), IMF Fiscal Monitor,
and authors’ estimates Note: Each sub-period shows the averages across
advanced and developing countries for which data is available

5.
5
II. Characterizing
Debt Spikes

6.
II. What is a debt spike?
6
Criteria for Episode Selection
• At least 1%GDP increase in debt/GDP per year
• At least 10 percent of GDP increase in debt during the episode [Abbas, 2011;
Weber, 2012]
Conditions
• No time limit on the maximum duration of episodes
• At least 2 years difference between the episodes (if <2 years, then
considered the same episode)
• There is data available to make the calculations for the debt
decomposition for the duration of the episode

7.
II. What is a debt spike?
7
We find a total of 179 episodes of multiyear debt accumulation greater than
10 percent of GDP, 80 among advanced economies and 99 among emerging
and low income countries.
Table 1. Duration of debt accumulation episodes
Max 75% 25% Min Median
ALL 179 15 7 3 1 5
ADV 80 14 8 4 1 6
EM & LIC 99 15 6 2 1 4
Number of
Episodes
Episode Duration

8.
II. What is a debt spike?
8
Total debt increase and duration of debt accumulation episodes
Median debt spike episode
Median debt spike episode

9.
II. What is a debt spike?
9
Debt spikes are not rare events: After 10 years, the probability of falling into a
debt spike is 50%; after 20 years the probability is 80%.
0.000.250.500.751.00
0 20 40 60 80
analysis time
ADV EME
LIC
1-Kaplan-Meier survival estimates
0.000.250.500.751.00
0 20 40 60 80
analysis time
1-Kaplan-Meier survival estimate
10yr= 50%
20yr= 80%

10.
10
III. What drives
debt spikes?

11.
III. What drives debt spikes?
11
𝑑 𝑇 − 𝑑0 =
𝑡=1
𝑇
𝑟𝑡 − 𝐺𝑡
1 + 𝐺𝑡
𝑑 𝑡−1 +
𝑡=1
𝑇
𝑝𝑡 +
𝑡=1
𝑇
𝑠𝑡
𝑑 𝑇 − 𝑑0 =
𝑡=1
𝑇
𝑟𝑡
1 + 𝐺𝑡
𝑑 𝑡−1 −
𝑡=1
𝑇
𝜋 𝑡
1 + 𝐺𝑡
𝑑 𝑡−1 −
𝑡=1
𝑇
𝑔𝑡
1 + 𝑔𝑡
𝑑 𝑡−1 +
𝑡=1
𝑇
𝑝𝑡 + +
𝑡=1
𝑇
𝑠𝑡
Where: d= debt to GDP; r = nominal effective interest rate; G = nominal GDP growth
rate; π = growth rate of GDP deflator; g = real GDP growth rate; p = primary deficit
to GDP; s=stockflow adjustment to GDP
Debt decomposition formula:
Interest rate Inflation Real GDP
growth
Primary
deficit
Stock flow
adjustment
Interest rate
growth
differential
Primary
deficit
Stock flow
adjustment
Total
change in
Debt
Past Debt

12.
III. What drives debt spikes?
12
Stock-flow adjustment (SFA) is the major driver of debt spikes
AEs EMEs & LICs
For the median episode in AEs, debt increases 25% of
GDP; and SFA increases 20% of GDP
For the median episode in EMEs & LICs, debt
increases 24 % of GDP; and SFA increases 30% of GDP
Data on foreign currency denominated debt is only available for 24 out of the 99 developing country episodes. For these 24 episodes, stock-flow
adjustments at the median amount to 16 percent of GDP, of which 6 percent of GDP can be attributed to the depreciation of the exchange rate.

13.
13
III. What drives debt spikes?
These findings confirm previous studies. For example:
• Campos et al. (2006) used data from 117 countries between 1972-2003 and showed that
budget deficits account for a relatively small fraction of debt growth and that stock-flow
reconciliation is one of the key determinants of debt dynamics.
• Abbas et al. (2011) looked at 60 episodes of debt increases between 1880–2007 and found
that key contributors to debt surges during nonrecessionary periods were both primary deficits
and stock-flow adjustments.
• Weber (2012), using data for 163 countries between 1980 and 2010, shows that stock-flow
adjustments were a significant source of debt increases, while they played only a minor role in
explaining debt decreases.

14.
14
IV. What is
behind SFAs?

15.
IV. What is behind SFAs?
15
In the EU (2002-2014), the main driver of SFA increases during debt spikes
was the net acquisition of financial assets (9% of GDP)

16.
IV. What is behind SFAs?
16
A large portion of the financial assets acquired were relatively illiquid, such
as loans, financial derivatives, equity and investment fund units.

17.
IV. What is behind SFAs?
17
In our global sample, the size of SFAs during debt spikes increases with
contingent liabilities, currency depreciation and with ‘weak’ politics.

18.
IV. What is behind SFAs?
18
Better fiscal institutions can help reduce average SFAs

19.
IV. What is behind SFAs?
19
The probability of suffering a debt spike is lower when there are stronger
budget institutions (both measured by PIMA and PIME indexes above the
median value)
Strong Institutions (PIMA>median)
Weak Institutions (PIMA<median)
0.000.250.500.751.00
0 20 40 60
analysis time
pimascore_high = 0 pimascore_high = 1
1-Kaplan-Meier survival estimates
Strong Institutions (PIME>median)
Weak Institutions (PIME<median)
0.000.250.500.751.00
0 20 40 60
analysis time
pimescore_high = 0 pimescore_high = 1
1-Kaplan-Meier survival estimates

20.
20
V. What are the
Consequences
of Sizeable Debt
Spikes?

21.
III. What drives debt spikes?
21
Large SFAs are directly associated with a higher probability of suffering flat-
debt paths in the aftermath of debt spikes
Probit
Ex-post
Flat-debt path
Ex-post
Flat-debt path
Ex-post
Flat-debt path
Ex-post
Flat-debt path
Debt level (end-of-episode) 1.181*** 1.017*** 1.011*** 1.016***
(0.0569) (0.0566) (0.0525) (0.0533)
Average SFA (during episode) 0.671*** 0.782*** 0.613***
(0.1) (0.0954) (0.183)
Average Primary Balance (during episode) -2.841*** -2.776***
(0.524) (0.537)
Average Episode Growth (during episode) -0.214
(0.576)
Average Episode Inflation (during episode) 0.312
(0.32)
Constant 25.82*** 24.15*** 20.45*** 20.71***
(3.281) (2.935) (2.805) (2.832)
Observations 178 177 177 177
R-squared 0.71 0.773 0.806 0.807
If the average drop in debt between years three to five after the end of the debt spike is less than 10 percent of GDP, we consider that
this country suffers from a non-declining debt path (taking a value of 1)

22.
22
VI. Can more
realistic SFAs
improve debt
forecasts?

23.
23
Many countries are optimistic in their debt forecasts.
VI. Can more realistic SFAs improve debt
forecasts?
Distribution of Actual and Forecast Annual Changes in Debt to GDP
Sources: World Economic Outlook (WEO), Eurostat, and authors’ estimates.
Note: Forecasts for t+2, t+3, and t+4. Columns for "EU countries" correspond to observations for 27
countries for annual forecast vintages between 1991 and 2014. Columns for "all countries"
correspond to observations for 85 countries for Spring WEO vintages between 1995 and 2014.
Debt Forecast Error:
1.5% GDP per year;
6% GDP per
forecast period

24.
VI. Can more realistic SFAs improve debt
forecasts?
24
Forecasts of SFAs are typically biased (they are always higher than expected).
Distribution of Annual Forecast Errors of SFAs Average Stock Flow Adjustments, Actual vs Forecast,
1991-2014
Note: Observations for 27 countries between 1991-2014. Forecasts for T+2, T+3, and T+4.

25.
VI. Can more realistic SFAs improve debt
forecasts?
25
WEO forecasts that by 2018 debt would be on a declining path in 17 out of 23
European countries. If we instead used historical SFAs for these countries, only 9
of these countries would have debt on a declining path.
WEO Forecasted Change in Debt to GDP, 2015-2018 SGP Forecasted Change in Debt to GDP, 2015-2018

26.
26
VII. Concluding
Remarks

27.
27
• We use the longest historical time series (including the global financial
crisis) and show that large debt spikes are driven by high SFAs, not primary
deficits
• This blind side of debt dynamics is linked to political and institutional
weaknesses.
• The consequences of large SFAs are long-lasting, because they lead to
higher debt accumulation in exchange of illiquid assets.
• SFAs should be properly forecasted. This would make debt projections
more realistic.
• Higher debt buffers would be needed, given the size of SFAs, their lack of
forecasting and their relative frequency,
VII. Concluding remarks

28.
28
Thank you!

29.
29
Background
slides

30.
Episodes of debt declines > 10% of GDP
(47 ADV & 90 EME/LIC episodes)
30
AEs EMEs & LICs
For the median episode in AEs, debt declines of 18% of
GDP; SFA increases 4% of GDP; primary surpluses of
18% of GDP
For the median episode in EMEs & LICs, debt declines
35% of GDP; SFA increases 3% of GDP; primary
surpluses of 5% of GDP

31.
Debt Spikes > 20% of GDP
(Total 107: 49 ADV & 58 EME/LIC episodes)
31

32.
Debt Spikes > 10% of GDP
Advanced vs. Developing
32

33.
Debt Spikes > 10% of GDP
Advanced vs. Developing
33

34.
Contingent Liabilities (from banking sector)
and SFA
34
Table 1.3.1. Regression Analysis: Financial Contingent Liabilities and Size of SFA
SIMPLE REGRESSIONS
Coefficients (simple regressions) Average Size Number
of SFA Obs
Size of Realized Contingent Liabilities 0.383*** N=51
[2.34]
Size of Realized Contingent Liabilities (financial) 0.639*** N=35
[3.30]
Coefficients (simple regressions) Average Size Number
of SFA Obs
Size of Realized Contingent Liabilities 0.393***
[2.06] N=39
Size of Realized Contingent Liabilities (financial) 0.618*** N=26
[2.60]
T-statistics in parentheses
*** p<0.01, ** p<0.05, * p<0.1
20% Threshold
10% Threshold