A partnership with academia

Policy ReseaRch WoRking PaPeR 4287

A Cross-Country Analysis of Public Debt
Management Strategies

Martin Melecky

The World Bank
Banking and Debt Management Department
July 2007

WPS4287
Pu

bl
ic

Di
sc

lo
su

re
A

ut
ho

riz
ed

Pu
bl

ic
Di

sc
lo

su
re

A
ut

ho
riz

ed
Pu

bl
ic

Di
sc

lo
su

re
A

ut
ho

riz
ed

Pu
bl

ic
Di

sc
lo

su
re

A
ut

ho
riz

ed

Produced by the Research Support Team

Abstract

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

Policy ReseaRch WoRking PaPeR 4287

This paper analyzes results of a survey on debt
management strategies conducted by the Banking and
Debt Management Department of the World Bank.
The analysis focuses on (1) whether a public debt
management strategy exists in a given country, (2)
whether it is made public, and (3) in which form it is

This paper—a product of the Banking and Debt Management Department—is part of a larger effort in the department to
understand the development of public debt management strategies. Copies of the paper are available free from the World
Bank, 1818 H Street NW, Washington, DC 20433. Please contact Martha Rosenquist, room MC7-126C, telephone
202-458-2602, fax 202-522-2101, email address Mrosenquist@worldbank.org. Policy Research Working Papers are also
posted on the Web at http://econ.worldbank.org. The author may be contacted at mmelecky@worldbank.org. July 2007.
(37 pages)

imparted. The paper analyzes the distribution of the latter
characteristics over different regions, income groups, and
levels of indebtedness using graphical analysis. Using
regression analysis, it investigates the extent to which
basic economic factors can explain the characteristics of
public debt management strategies across countries.

A Cross-Country Analysis of Public Debt

Management Strategies*

Martin Melecky#

Banking and Debt Management Department
World Bank

Keywords: Public Debt Management; Strategy; Cost-Risk Trade-off; Strategic
Guidelines; Strategic Benchmark; Economic Indicators;
JEL Classification: H63, H74, O50

*
I am grateful to Marianne Sarkis for an excellent research assistance. I thank Phillip Anderson, Elizabeth

Currie, Lars Jessen, Tomas Magnusson, Eriko Togo, and Antonio Velandia for their comments and
suggestions. All remaining errors are mine.

#
Banking and Debt Management Department, World Bank, 1818H Street NW, DC 20433, USA, email:

mmelecky@worldbank.org

1 Introduction

Governments have to often borrow in order to …nance expenditures on public goods

and services that promote growth and increase nations’welfare. The decision of how

much to borrow is that of …scal policy which determines the targeted level of debt

based on a sustainability analysis of government debt. One concept of sustainability

relates to solvency, the ability of the government to service its debt obligations in

perpetuity without explicit default, Burnside (2004). Another concept put forth

by Burnside (2004) renders …scal sustainability a broader scope by relating it to

the government’s ability to maintain its current policies while remaining solvent.

Within the latter concept, one can discuss the types and consequences of …scal and

monetary policy adjustments needed to avoid future insolvency. Even more broadly,

this concept can encompass discussions on the optimality of …scal policy rather than

its mere feasibility.

Once the government decides on how much funding needs to be raised, it has to

further determine the form in which the funds will be delivered.1 In other words, the

government has to decide which debt instruments are going to be used to raise the

intended funding. Similar to any other private borrower, the government will seek the

best terms for its borrowing. However, given the size of government borrowing, the

analogy to a private investor might be misleading as none of the government’s choices

or policy actions is considered to be irrelevant for the equilibrium outcome,2 see e.g.

Missale (2000). A government’s seeking of the best borrowing terms refers to the aim

of minimizing the cost of borrowing within existing constraints while respecting the

government’s risk preferences (aversion). In other words, the government not only

aims to raise funding at low cost but also to structure the composition of its debt

portfolio in such a way as to minimize the impact of relevant shocks on its budget

1The variety of options that is available to the government certainly varies across countries
mainly with regard to their stage of development.

2One most common example being the possibility of crowding out e¤ect of government borrow-
ing, see e.g. Briotti (2005) or Elmendorf and Mankiw (1998), but also a crowding-in e¤ect of public
…nance can be expected, see e.g. Alani (2006) or Friedman B. (1978). Also, given the size of gov-
ernment debt portfolio its …nancial characteristics may constitute a systemic risk for the domestic
…nancial sector.

1

or long-term expenditure plan. The debt instruments for …nancing of government

debt are determined by the public debt managers based on the delegated authority

from the government.3 The debt portfolio composition is thus the policy instrument

of public debt managers.

The fundamental document that guides debt managers in their decisions and

operations is the public debt management strategy. The strategy is built upon foun-

dations (goals) stated in government’s debt management objectives. The debt man-

agement objectives are usually expressed along the following lines, see IMF and WB

(2001):

The main objective of public debt management is to ensure that the gov-

ernment’s …nancing needs and its payment obligations are met at the low-

est possible cost over the medium to long run, consistent with a prudent

degree of risk.

The debt management objectives also typically contain sections addressing the

government’s involvement in domestic bond market development and coordination

of its actions with …scal and monetary policies. The latter relates to the fact that the

objectives of …scal policy, monetary policy, and public debt management di¤er but

there are various interdependencies among their policy instruments, see e.g. Wheeler

(2004) or Togo (2007). Missale (2000) argues that the objectives of minimizing the

expected cost of debt servicing and minimizing risk are of little help operationally.

According to Missale the objectives are also unuseful as principles on which one

can construct benchmark portfolios against which the performance of debt managers

could be evaluated.4 He bases his arguments on the fact that meeting government

3The process underlying delegation of authority to the debt management o¢ ce to borrow and
execute related transactions in …nancial markets on behalf of the state is described in more detail
in IMF and WB (2001) and Wheeler (2004).

4Simple as it seems, it might be a di¢ cult task to evaluate performance of debt managers
against a benchmark portfolio as not achieving the benchmark may be desirable on some occasions.
One can use the analogy of the role of an in‡ation forecast in in‡ation targeting. Although, policy
instruments are used to anchor in‡ation expectations at the targeted level the actual future in‡ation
can end up away from the target due to the e¤ect of unexpected shocks or shocks that the monetary
policy does not want to counteract.

2

objectives is not an easy task to accomplish, especially in the absence of any theory

of the appropriate degree of risk-aversion that a government should exhibit, or more

generally elicit the preferences of society on this matter. Ideally, the debt manage-

ment objectives of the government including its risk preference (aversion) guide the

debt managers in the design of debt management strategy, and are re‡ected in the

chosen cost-risk trade-o¤.

Now consider the process by which the strategy comes about in practice. The

debt management strategy is proposed by the debt management authority or more

speci…cally its middle o¢ ce5 to the minister of …nance. The …nance minister then

reviews and approves the proposed strategy, often based on the input of the advisory

board for debt management. The advisory board can comprise representatives of

…scal policy, government administration, monetary policy, and other regulatory and

supervising bodies a¤ected by the course of the debt management strategy. The

review and approval of a debt management strategy at the level of a minister of

…nance is aimed at ensuring that the proposed strategy is consistent with the debt

management objectives of the government, including its preference for risk.

In general, the formalized debt management strategy can take two basic forms.

Either be presented in terms of guidelines or constitute a benchmark for the optimal

government debt portfolio. The former relates to a document which guides the debt

managers on types of risks that should be considered as relatively more important,

and thus indirectly points to the desired structure of a debt portfolio. Therefore,

the guidelines provide directions for future debt management operations rather than

quantitative targets. On the other hand, strategic benchmarks state explicitly what

are the desired risk characteristics of the optimal debt portfolio in a quantitative

manner. The strategic benchmarks can quantify the targeted risk characteristics of

the optimal debt portfolio either in terms of speci…c magnitudes or more often speci…c

ranges. The basic types of risks that debt managers should consider when designing

their strategy are discussed in detail in e.g. Wheeler (2004, pp. 17) while various

pitfalls arising in debt management and constituting hidden risks are discussed in

5See Wheeler (2004) and IMF and WB (2003) for detailed description of the organizational
structure of debt management authorities.

3

IMF and WB (2000, box 2). For the purpose of this paper we consider three basic

types of risks: (i) foreign currency (FX) risk, (ii) re…nancing (roll-over) risk, and (iii)

interest rate risk. Risk type (i) addresses the desired currency composition of the

debt portfolio, i.e. the relative weight on domestic currency versus foreign currency

debt. Further, the currency composition of the foreign currency debt itself can be also

addressed. Risk type (ii) addresses the desired maturity structure and redemption

pro…le of the debt. Risk type (iii) deals with the desired proportion of ‡oating as

opposed to …xed interest rate debt or, in some cases, the price-indexed debt.6

Another attribute of a debt management strategy that will be considered in this

paper is whether such a formal document is made public. The debt management

strategy is considered as public if it is published either in the annual report of the

debt management body, or made available on the respective website.

This paper aims to summarize and analyze the results of a survey on debt man-

agement strategies conducted by the Banking and Debt Management Department

of the World Bank. We will focus on three main …ndings related to the debt man-

agement strategies across countries. Namely, (i) whether a public debt management

strategy exists in a given country, (ii) whether it is made public, and (iii) in which

form it is presented, i.e. either in the form of guidelines or a strategic benchmark.

We will analyze the distribution of the latter characteristics over di¤erent regions,

income groups and levels of indebtedness using graphical analysis. Moreover, we

will use regression analysis to investigate to which extent selected economic indi-

cators can explain the characteristics of public debt management strategies across

countries. In expectation of the results we would like to set forth the following hy-

potheses. Namely, that increasing income levels increase the incidence of strategies;

that increasing levels of indebtedness show a positive but likely non-linear (hump-

shaped) relationship with the incidence of debt strategies; and that countries facing

larger shocks show lower incidence of debt strategies. To our knowledge this paper

6Regarding the ful…llment of a strategy one can also think of specifying the pace at which the
strategic benchmark should be reached, as the latter represents another level of the cost-risk trade-
o¤. More speci…cally, in order to move the current debt portfolio structure faster towards the
benchmark’s structure the debt managers would have to proportionally relax their cost considera-
tions as both restructuring of the debt portfolio or hedging can be costly.

4

is a …rst attempt to analyze di¤erences in public debt management strategies across

countries and contribute to better understanding of the development economics of

public debt management.

The remainder of the paper is organized as follows. Section 2 describes the an-

alyzed survey data and its collection. Section 3 contains graphical analysis of the

survey data across income groups, regions and levels of indebtedness. Section 4 car-

ries our regression analysis of the survey data using economic indicators as candidate

explanatory variables. Section 5 provides a summary of …ndings and conclusions.

2 The Survey Data

Progressing in the e¤orts to better understand the development economics of public

debt management strategies across di¤erent country groups and individual coun-

tries, the Banking and Debt Management Department of the World Bank conducted

a survey on public debt management strategies. The survey was carried out during

the period from August 2006 to February 2007 and covers OECD, IBRD and Blend

countries.7 The questionnaire was sent out to and completed by national author-

ities responsible for public debt management, or if not feasible the questionnaire

was completed by the relevant country economist based on a dialog with the na-

tional governments. The information from the questionnaire was supplemented by

a search through websites of institutions responsible for central government’s debt

management. The questionnaire asked the following questions8

(i) Has the government established a debt management strategy for the

total central government debt portfolio?

(ii) Is the debt management strategy document published?

7The applied classi…cation into country groups is that of the World Bank and is available at
http://web.worldbank.org/WBSITE/EXTERNAL/DATASTATISTICS/
0,,contentMDK:20421402~pagePK:64133150~piPK:64133175~theSitePK:239419,00.html

8The survey was made con…dential regarding the aswers of individual countries so that no country
examples appear in the paper.

5

(iii) Have you established a strategic target/benchmark for the total debt

portfolio?

The questions were answered in a Yes/No manner and converted to 1=0 entries

for each country, respectively. Regarding point (i), due to the formulation of the

question the positive answers may include implicit strategies. After acquiring all

observation the data were reviewed and some adjustments made to ensure their

consistency across countries.9 The latter pertains to ensuring that the unobserved

quality of debt management strategies which are not made public meets certain

criteria. Namely, the emphasis was placed on the fact that a debt management

strategy has to address the cost-risk trade-o¤, not only the cost of …scal …nancing.

This requirement thus excludes references to purely …scal expenditure frameworks or

frameworks addressing …scal sustainability. Concerning point (ii) the questionnaire

was supplemented by website search to obtain the strategy documents. In point (iii)

all countries that appeared to have at least one benchmark target or targeted range

for one of the three risks below quali…ed for a positive answer.

If countries have established a strategic target/benchmark for their public debt

portfolio they were asked which types of risks the strategic target/benchmark ad-

dresses. Namely, they were asked

(iii.a) Have you established a strategic target/benchmark for currency risk

(% domestic vs. % foreign)?

(iii.b) Have you established a strategic target/benchmark for interest rate

risk (% …xed vs. % ‡oating; average time to re…xing (months); or modi…ed

or Macaulay duration (years))?

(iii.c) Have you established a strategic target/benchmark for re…nancing

risk (ceiling on debt maturing within one year (% of total outstanding);

or average time to maturity (years))?

The Yes/No answers to the latter questions were also converted into 1=0 entries.

9I am grateful to Lars Jessen and Antonio Velandia for their help in this process and Phillip
Anderson, Elizabeth Currie and Tomas Magnusson for their expert inputs.

6

The entire data set covers 105 countries where the analysis of question (i) is based

on all 105 observations, and analyses of questions (ii) and (iii) on 66 observations

on strategies. To broadly characterize our sample, we …nd that out of the total

of 105 countries 66 countries have a public debt management strategy, of those 38

communicate their strategies in terms of guidelines, and 28 in terms of benchmarks.

Regarding the source of information that provided the basis for our classi…cation

40% of countries responded to the questionnaire either by themselves or via the

WB’s country o¢ ce. In case of 9% of the countries the information from needs

assessments conducted by the Banking and Debt Management Department was used

and updated by means of a website search. Finally, 51% of countries were classi…ed

based on information from the relevant websites, 46% of those are OECD countries

and the remainder are countries for which a response either to the questionnaire

sent out for the PDM forum, the questionnaire sent out directly to the relevant debt

management authorities, or to WB’s country o¢ ces10 was not recovered. If none of

the applied …ve information channels worked out the country was assigned a response

of "No" to question (i), which excluded it from the analysis of questions (ii) and (iii).

There are 21 non-OECD countries that were assigned a response of "No" in such a

manner.

What concerns the regression analysis presented in section 4, the data sample

employed is reduced due to unavailable data for some of the economic indicators used

to explain the variation in strategies’characteristics across countries. We discuss the

countries included in the regression analysis and the availability of data for estimation

in section 4.2.

3 Graphical Analysis

In this section we analyze how the probability of having a public debt management

strategy varies across di¤erent income groups and regions. We use the World Bank’s

income classi…cation to divide countries into groups of high income, upper-middle

10The WB’s country o¢ ces were asked to respond after a dialog with the relevant country’s
authorities or after a thorough assessment of the subject matter.

7

income, and lower-middle income.11 Similarly, we use WB’s regional classi…cation

to divide countries into regional groups of East Asia and Paci…c (EAP), Europe and

Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and

North Africa (MNA), South Asia (SAR), and Sub-Saharan Africa (AFR).12 We will

discuss the division of countries into groups according to their levels of indebtedness

directly in section 3.3.

3.1 Comparison across Income Groups

Figure (1) shows the percentage of countries in our sample that have a public debt

management strategy when looking across di¤erent income groups. Further, panel B

of Figure (1) shows what percentage of strategies is made public, and what percentage

of strategies is expressed in terms of a strategic benchmark as opposed to strategic

guidelines when looking across di¤erent income groups.

Consider …rst panel A of Figure (1). As expected middle-income countries (MICs)

in general fall behind the high-income countries regarding the percentage of coun-

tries which have a debt management strategy. This would support the hypothesis

that more comprehensive management of public …nances comes with a higher stage

of economic development. However, it is interesting to observe that across the two

subgroups of MICs the pattern does not hold with the same signi…cance, i.e. that

countries in the lower MIC group show similar probability of having a debt manage-

ment strategy as countries in the higher MIC group. A tentative explanation might

be that while MICs pursue implementation of more stable (robust) macroeconomic

policies, a relatively higher improvement in this respect can be observed in the lower

MICs. Nevertheless, the pattern observed in Figure (1) can be to some extent an

artifact of the selected conventional income ranges to group the countries for the

purpose of constructing a histogram. We explore the relationship between income

levels (economic development) and the probability of a country having a strategy

using regression analysis later on in this paper.

11Recall that low income countries are not included in this analysis.
12Footnote 8 provides a link to WB’s website containing the list of countries in each income group

and region.

8

The percentage of strategies that are made public across income groups shown

in panel B of Figure (1) is positively linked to income levels of individual country

groups. This observation implies that at higher stages of development there is more

demand for transparency and accountability of public debt management, and also

higher capacity to meet such demand. The next characteristic of debt management

strategies plotted in panel B of Figure (1) is the percentage of strategies expressed

in terms of benchmarks as opposed to guidelines. Before examining the data one

can assume two alternative hypotheses. First, one would expect that expressing a

debt management strategy in terms of a benchmark requires higher capacity and

analytical rigor. This is due to the fact that setting numerical bounds for di¤erent

types of risks consistent with prudent debt management requires comprehensive risk

analysis. Also, MICs may wish to retain greater ‡exibility a¤orded by guidelines,

as they are more vulnerable to shocks and changing economic environment. This

is especially true in the case of developing economies that face various constraints.

Second, the MICs could opt for benchmarks more than high income countries since

the capacity to e¤ectively manage public debt is rather concentrated at more senior

levels so that the strategic benchmark appears to be a more e¢ cient way of guiding

the debt management sta¤ in its daily operations. In addition, the range of risks that

MICs face is broader than that of high income countries, and this relatively higher

complexity of risk management proves to be better handled via benchmarks for all

the risks the government wants to address. We can observe in the histogram that the

probability of country using a benchmark to express its strategy is positively related

to the levels of income, a …nding consistent with our …rst hypothesis.

3.2 Comparison across Regions

We now proceed to look at the distribution of the characteristics of interest across

di¤erent regions. Panel A of Figure (2) plots the percentage of strategies out of

the total number of observations in di¤erent regions. Panel B of Figure (2) shows

the percentage of strategies made public in each region, as well as the percentage of

strategies in each region expressed as a strategic benchmark.

9

We can observe in panel A that the probability of an OECD country having

a debt management strategy is signi…cantly higher than the probability of a non-

OECD country having a debt management strategy. This observation can again be

seen as a call for higher accountability, at least operational13, as a country reaches

higher stages of development. A strategy is linked to operational accountability

through a requirement to report and explain deviations of actions undertaken by debt

management from those consistent with the strategy. Even though, the percentage

of strategies in place reported for the high income countries does not reach 100%

the implicit reference point is in fact 100% as some OECD countries without a debt

management strategy show extremely low levels of indebtedness, so that having a

debt management strategy is not a priority itself. In addition, some OECD countries

may not have a traditional debt management strategy in face of extremely deep and

liquid …nancial markets and stable macroeconomic policies. Consider now the regions

in accord with WB’s classi…cation. We exclude SAR from our interpretations due

only to 1 available observation. Out of the considered regions, ECA appears to be the

leading region with the highest incidence of strategies, followed by MNA. As for the

remaining regions the percentage of countries having a debt management strategy

is below 50% and it is interesting to see that LAC and EAP are falling behind the

MICs of Africa.

Consider now panel B of Figure (2). It is interesting to observe that not all of

the strategies are made available to the public even in OECD countries, and that

certain debt managers prefer a lower degree of transparency in order to preserve more

room for their maneuvers.14 For some countries the non-transparency might have

di¤erent origins, though. Non-OECD countries again lag behind the OECD countries

in the percentage of public strategies. However, ECA not only leads the group of

WB’s regions regarding the percentage of public strategies, but is as a region more

13In a nutshell, operational accountability amounts to reporting and explaining actions with
respect to objectives, here those of debt management. See e.g. Buiter (2006) for more elaborate
description of operational versus substantive accountabilility.

14This preference could be especially justi…able if countries face higher uncertainty and operate
within a stringent accountability framework. Even public guidelines could thus be constraining in
fact.

10

transparent that the group of OECD countries. In this respect, EAP and AFR follow

after ECA while only MNA fails to cross the threshold of 50% of public strategies.

When we turn to the percentage of strategies expressed in terms of benchmarks

the OECD countries dominate in this respect the non-OECD countries, see also

section 3.1. From the WB’s regions, ECA, followed by LAC shows the highest

percentage of strategic benchmarks, a slightly lower one than the corresponding

percentage for OECD countries. One could be curious whether the higher incidence

of benchmarks in ECA and LAC could be attributed to a relatively higher analytical

capacity for risk management in ECA and LAC in comparison with AFR, EAP and

MNA.

3.3 Comparison across Levels of Indebtedness

The …rst row of panels in Figure (3) shows the distributions of the percentage of

strategies out of total observations across country groups with di¤erent levels of

indebtedness. The second row of panels in Figure (3) shows distributions of the per-

centage of public strategies and benchmarks out of strategies across country groups

with di¤erent levels of indebtedness. The distributions are constructed using two

di¤erent bases. In both cases countries were …rst ordered in ascending order in terms

of their levels of indebtedness. The …rst column of panels in Figure (3) uses the

ranges of indebtedness to classify countries into groups. The second column of pan-

els in Figure (3) divides the countries into groups of an equal size. The two slightly

di¤erent approaches will help us get a better picture about the distribution of the

characteristics of interest across levels of indebtedness.

Consider now the …rst row of Figure (3) which shows the distribution of the

percentage of countries with strategies. Starting from the left, we can observe that

the relationship between the probability of a country having a strategy and its level

of indebtedness is hump-shaped where countries with debt levels as a percentage of

GDP higher than 100% have the smallest probability of having a strategy, i.e. smaller

than countries with levels of indebtedness between 0100%. However, countries with
smallest levels of indebtedness between 0 50% have lower probability of having a

11

strategy than countries with a medium level of indebtedness ranging from 50 to 100

percent. The second plot in the …rst row of panels supports the …nding of a hump-

shaped relationship between indebtedness and an existence of a strategy. However,

the story coming from the tails of the distribution appears opposite to that from the

…rst plot. Namely, that countries with high levels of indebtedness have still higher

probability of having a strategy than countries with small levels of indebtedness.

The distribution of the percentage of public strategies across levels of indebtedness

is shown in the second row of panels in Figure (3). From the …rst plot it appears

that the relationship is rather non-linear and close to the U-shape. However, the

second plot contradicts this pattern and shows a strong linear relationship between

the percentage of public strategies and levels of indebtedness. This implies that

the higher the indebtedness of a country the more likely is the country to be non-

transparent about its debt management. We will investigate the latter relationship

further using regression analysis later on in this paper to get more de…nite answers.

Finally, the second row of panels in Figure (3) also shows the distribution of

benchmark strategies across levels of indebtedness. The left-hand side histogram

suggests a hump-shaped relationship between benchmark strategies and the levels of

indebtedness. The hump-shaped relationship implies that as countries are becoming

more indebted they use benchmarks more often to express their strategies. However,

as countries become highly indebted the use of benchmarks in debt management

decreases. Although a non-linearity is also suggested by the plot on the right it

follows a U-shape as opposed to a hump-shape thus completely contradicting the

implications from the …rst plot. Again, we hope to …nd more conclusive insights into

this relationship using regression analysis carried out in section 4.

3.4 Further Inspection of Strategic Benchmarks

In this section we further decompose the strategic benchmarks into three basic types

of risk that the benchmarks can be addressing. These are namely foreign exchange

(FX) risk, interest rate risk and re…nancing risk. Using graphical analysis, we exam-

ine the relative weight of the three types of risk in the existing benchmarks across

12

di¤erent income groups, regions, and levels of indebtedness. Figure (4) plots the

frequency (in percentages) with which a given type of risk is addressed in the bench-

marks. Panel A of Figure (4) does so while considering di¤erent income groups,

panel B di¤erent regions, and panels C and D varying levels of indebtedness. In

all panels we show an average proportion of all the three di¤erent risks addressed

in strategic benchmarks, denoted by "average". The latter is computed as the un-

weighted average of incidence with which each of the three risks is addressed in

strategic benchmarks.

It appears that the average number of di¤erent risks the benchmarks in MICs

address is higher than the average number for high income countries. This is mainly

attributable to the fact that high-income countries are not concerned much with

FX risk and re…nancing risk as they have access to large and liquid debt markets

usually in local currencies. It is interesting to observe that the average number

of risks addressed in benchmarks by lower MICs is signi…cantly higher than the

average number of risks addressed by upper MICs. This may be again due to the

fact that lower MICs are more concerned about FX and re…nancing risk than upper

MICs. This seems to imply that the consideration of re…nancing risks in strategic

benchmarks is negatively correlated with countries’income levels. This could apply

to FX risk as well, however here, the relationship is possibly non-linear as upper

MICs show slightly lower percentage of FX risk targets (targeted ranges) in their

benchmarks than high income countries. On the other hand, interest rate risk shows

rather the opposite tendency, i.e. its presence in strategic benchmarks appears to be

more positively correlated with countries’income levels. For high-income countries

the interest rate risk is probably the only concern given their …nancing opportunities,

and from the point of view of MICs it is the risk that is supposedly the easiest one

to manage.

Panel B looks at the distribution of the three risks addressed in strategic bench-

marks using cross-region comparison. The average number of di¤erent risks addressed

in benchmarks of non-OECD countries seems to be signi…cantly larger than in the

case of OECD countries. Again, the evidence shows that this is primarily due to the

higher concern of the non-OECD countries about FX and re…nancing risks. Neverthe-

13

less, the proportion of OECD countries concerned about interest rate risk dominates

that of non-OECD countries. It is hard to draw any conclusions for some of the

WB’s regions, as AFR, MNA or even EAP, since there is only one observation avail-

able in each case. Our interpretations thus shrink to comparison of ECA and LAC.

LAC seems to be leading in the average number of risks addressed in benchmarks.

Moreover, once a country in the LAC region employs a strategic benchmark for debt

management this benchmark is very likely to include targets (targeted ranges) for

all re…nancing, interest rate and FX risks. ECA seems to be most concerned about

interest rate risk and the risk pro…le of this region thus more resembles that of the

OECD countries.

Panels C and D of Figure (4) show the distribution of benchmarks addressing the

three basic types of risk across levels of indebtedness. Panel C uses as the base ranges

of indebtedness whereas panel D puts the countries, in ascending order according to

their levels of indebtedness, into equally populated groups. We can observe that

the average number of di¤erent risk addressed in the strategic benchmarks is rather

negatively related to the levels of indebtedness, although panel C shows possible

presence of non-linearity. Namely, the countries with levels of indebtedness between

50 100% of GDP appear to use a larger number of di¤erent benchmarks than low
and highly indebted countries. The incidence of re…nancing risk being addressed in

strategic benchmarks is strongly declining with levels of indebtedness. On the other

hand, the incidence of interest rate risk being addressed in the benchmarks appears

to be ambiguously related to levels of indebtedness as the story from panels C and D

is very di¤erent. The relationship between levels of indebtedness and the incidence of

exchange rate risk in strategic benchmarks appears to be possibly non-linear based

on the evidence across panels C and D. This is since the countries with levels of

indebtedness between 50 100% of GDP appear to use targets (targeted ranges) for
exchange rate risk most often and the countries in the range of 100%+ the least.

14

4 Regression Analysis

In this section we investigate to what extent the existence of a public debt man-

agement strategy in a country can be explained by economic indicators. We will

attribute the unexplained part of the event, i.e. a country having a strategy, to

political, institutional and idiosyncratic (country speci…c) factors. The economic in-

dicators employed are of a general character and pertain to, for instance, the stage

of economic development, macroeconomic management, indebtedness of a country,

‡exibility of the applied exchange rate regime, and are described in detail in section

4.1.

Ideally we would be interested in explaining the variation of the quality of debt

management strategies, yi ; across countries, i.e.

yi = Xi + i (1)

using economic indicators Xi: However we do not observe yi instead we observe

yi which takes values of 0 or 1 according to the following rule

yi = f 1 if y

i > y

0 otherwise
(2)

where the threshold yis set so that the strategy has to at least consider the

cost-risk trade-o¤when meeting government …nancing needs. It is also assumed that

i N (0; 2) : We thus have a vector of yi with 0 and 1 entries corresponding to
a country having or not having a public debt management strategy, and an index

i = 1:::81 denoting the countries in our sample. It can be shown, see e.g. Johnston

and Dinardo (2001), that the latent regression in (1) and the rule in (2) generate a

PROBIT model.

We are interested in modelling the probability that yi takes the value of 1 condi-

tional on selected economic indicators Xi, and thus transforming Xi into a proba-

bility, i.e.

prob (yi = 1) = F (Xi) (3)

15

where is a vector of parameters and F () is assumed to be a cumulative standard
Normal distribution. There are two most common alternatives to consider when

choosing the functional form of F . These correspond to models of linear probability,

and LOGIT. Since it appears that in vast majority of empirical cases the three models

seem to produce similar answers (see Johnston and Dinardo, 2001, chapter 13), we

choose to focus on the PROBIT model out of convenience15. This is due to the fact

that unlike the linear probability model the PROBIT model restricts the …tted values

to lie between 0 and 1, and we …nd its functional form more intuitive for our case

than that of LOGIT.

4.1 Selected Economic Indicators

We now discuss the economic indicators employed as explanatory variables Xi in

the PROBIT model in (3). The selection of those indicators was based on data

availability to maximize the coverage of the survey data, and an agnostic approach

to collecting basic economic indicators related to public debt management. The se-

lected economic indicators include measures of economic development, the level of

indebtedness and regional location in order to extend the graphical analysis of sec-

tion 3. In addition, we focus on some characteristics of government borrowing such

as the proportion of concessional debt, growth of government revenues approximated

by GDP growth, ‡exibility of applied exchange rate regimes, and volatility of do-

mestic and external shocks that may a¤ect cash‡ows related to the debt portfolio or

government primary balance. The volatility of shocks is mainly considered due to

the aim of debt management to minimize the shocks’impact on government budget

by optimizing the composition of the government debt portfolio. We now discuss the

employed economic indicators in detail.

GDP per capita - this variable is used to approximate the stage of development

of a country. One may expect that the higher the stage of development the higher

the probability that a country has a debt management strategy. A higher stage of

15We still compare the estimation results from the PROBIT model to those from the LOGIT and
LP models to check for possible misspeci…cation problems.

16

development is thus assumed to be associated with a better institutional framework

including a debt management strategy and its public availability. We used also a

quadratic of this variable in the model to capture possible non-linearities, however,

it appeared to be insigni…cant and is not reported in the estimation results. The

measure of GDP per capita is the PPP converted gross domestic product from the

Penn World Tables (Heston et al, 2006).

Indebtedness of government - increases in this variable, de…ned as the ratio of

total government debt to GDP, should result in an e¤ort to consolidate government

…nances and adopt a debt management strategy. One can also expect that if this

indicator reaches high levels the government may give up on debt management and

focus on debt renegotiations. Although debt renegotiations could be seen as a part of

the debt management strategy we do not include them in our indicator yi: Therefore,

inclusion of a quadratic of government indebtedness into our PROBIT model can be

justi…ed. As in the case of the GDP per capita the quadratic term appeared to be

insigni…cant and is not reported in the presented estimation results. The total debt-

to-GDP ratio was obtained from the GDF & WDI Central database of the World

Bank and the EIU database.

Government share of GDP - this variable is used to approximate the importance

of public sector (government) in economic performance of a country. One may expect

that a larger share of government on real GDP would result in a greater e¤ort to

stabilize government …nances in the sake of greater macroeconomic stability. Sim-

ilarly, if government actions are important for an economy the public will require

higher transparency and accountability from the government. Existence of a public

debt management strategy is thus deemed to represent increased e¤orts of the gov-

ernment to stabilize its …nances and meet the requirement of the public for higher

transparency and accountability. The measure used here is the government’s share

on real GDP from the Penn World Tables (Heston et al, 2006).

Degree of government debt concessionality - this variables is used to capture the

percentage of government debt …nanced by means of concessional resources, e.g. from

multilateral and bilateral donors. We assume that the higher the concessional share

of government debt the lower the incentive for the government to adopt a strategy

17

addressing cost-risk trade-o¤s in …nancing decisions, most importantly decisions on

debt composition. This indicator may appear to be perfectly correlated with GDP

per capita, in fact the correlation is estimated to be 0:55. Although the correla-
tion can be regarded as high an exclusion of the degree of concessionality from the

regression for strategies was rejected. We use the ratio of concessional debt to total

external debt to approximate this indicator. This measure was obtained from the

GDF & WDI Central database of the World Bank.

Internal macroeconomic management - the standard deviations of CPI in‡ation

and GDP growth are used to capture quality of internal macroeconomic management.

Since price stability is the basic objective and goal of monetary policy we …nd the

standard deviation of in‡ation indicative of the quality of internal macroeconomic

management. Also, the monetary and …scal policies aim at smoothing ‡uctuations

in economic performance, the economic growth cycle. We again draw the link from

sound and successful macroeconomic policy of a government to its likely engagement

in sound practice regarding public debt management. However, also in this case the

argument can be posed di¤erently. Namely, that the success of a sound macroeco-

nomic policy will depend on the institutional set up of the economy, such as e.g.

wage negotiation mechanisms or capital adequacy requirements for …rms, that deter-

mines the pass-through and size of domestic shocks. Although, the government can

in‡uence the institutional set up in the long run, facing larger domestic shocks can

lead to adoption of more advanced instruments for public debt management. The

CPI in‡ation and GDP growth series were obtained from the GDF & WDI Central

database of the World Bank.

Flexibility of exchange rate regimes - we use the standard deviation of the change

in the exchange rate to approximate this indicator. The lower the standard deviation

the lower the ‡exibility of an exchange rate regime. However, varying volatility of

exchange rates across countries is also attributable to varying impacts or sizes of

external shocks. This would be certainly the case if one dealt only with ‡oating

exchange rate regimes. In order to condition on the external shocks we employ other

variables such as volatility of current account or the terms of trade, see below. The

standard deviation of exchange rate is computed using the exchange rate series from

18

the Penn World Tables (Heston et al, 2006).

External macroeconomic management - we use the standard deviation of the cur-

rent account-to-GDP ratio (CA/GDP) to approximate this indicator. It may appear

that the actual exchange rate deviates from the equilibrium exchange rate that brings

the economy to external balance. This is especially true in the case of less ‡exible

exchange rate regimes present in our sample. If the external macroeconomic man-

agement (policies) are poor, i.e. there are frequent or large deviations of the actual

exchange rate from its equilibrium, this will result in higher variability of external

balances, measured here by CA/GDP. We use the quality of external macroeconomic

policies as one of the indicators of the overall quality of general government policies

and draw a link to the quality of public debt management - existence of a debt man-

agement strategy. To set an alternative hypothesis, one may argue that whatever

the external policy its success, as measured by the standard deviation of CA/GDP

here, depends on the magnitude and frequency of external shocks such as those to

the terms of trade and capital ‡ows. Further, one may extend this argument and

assume that the higher the importance (impact) of external shocks the more likely is

a country to adopt better instruments for public debt management, such as a debt

management strategy. The series of CA/GDP was retrieved from the GDF & WDI

Central database of the World Bank.

Management of foreign reserves - we measure the quality of management of for-

eign reserves using the coe¢ cient of variation in the stock of FX reserves-to-imports

ratio. Since management of foreign reserves is part of the …nancial management of

the consolidated government balance sheet we assume that its quality can be posi-

tively linked to the quality of public debt management. This is especially true if there

exist a high degree of coordination between monetary policy and debt management.

The series of FX reserves as a percentage of imports was taken from the GDF &

WDI Central database of the World Bank.

Volatility of o¢ cial transfers - more speci…cally we use the coe¢ cient of variation

for net o¢ cial current transfers. This indicator is employed to approximate the

exogenous volatility (risk) in foreign aid that developing countries may face. This

volatility may force countries to take some precautionary actions which may include

19

creation of a bu¤er stock of …nances to smooth out the volatility. Increases in the

required o¤setting …nancing could make the country acknowledge the need for a debt

management strategy. On the other hand, if the volatility in foreign aid is disrupting

the …nancing plans of the government at some point it could undermine the e¢ cient

continuation of a debt management strategy. The series of net o¢ cial transfers was

taken from the GDF & WDI Central database of the World Bank.

Regional dummy variables - we include also regional dummies into the regression

to explore the possibility that the existence of a strategy is dependent on the region

a country belongs to. The regional classi…cation corresponds to that used through-

out the graphical analysis in section 2.16 When constructing the regional dummy

variables we take as a base the LAC region due to the highest number of available

observations.

All indicators were calculated using available annual data covering the period

from 1990 to 2006.

4.2 Estimation Results

This section reports and discusses the results of PROBIT model estimation. Recall

that by using the PROBIT model we try to explain the probability of a country hav-

ing a public debt management strategy using selected economic indicators discussed

in section 4.1. The maximization of the log-likelihood of the PROBIT model17 is

carried out using the Berndt-Hall-Hall-Hausman algorithm. The inference is based

on the quasi-maximum likelihood (QML) standard errors due to Huber and White18

16We have tried to insert a regional dummy taking 1 if the country was an OECD country and
0 otherwise, but this dummy appeared to be insigni…cant and was dropped from the estimation.
This is most likely due to its colinearity of 0:87 with the income levels of the countries.

17The log-likelihood function is of the form:

l

= ln (L) =

X
i

yi ln

Xi

+ (1 yi) ln

Xi

where is standard Normal cumulative distribution and is the standard deviation of the unob-
served shock in the regression underlying the PROBIT model, see e.g. Johnston and Dinardo for
further details.

18The QML variance covariance matrix is computed as

20

which are robust to general misspeci…cation of the conditional distribution of yi. As a

check for possible misspeci…cation problems we estimate the LOGIT and LP models

for yi using the same set of explanatory variables. The sample used for estimation

includes 81 countries, shown in Table (3) for which data on the employed economic

indicators were available. 29 of the countries included in the regression are OECD

countries, 7 of them from AFR, 6 from EAP, 19 from ECA, 19 from LAC, and 8

from MNA.19 The results are reported in Table (1).

The estimation results are broadly consistent across the PROBIT, LOGIT and

LP models so that there seems to be no obvious signs of misspeci…cation prob-

lems. PROBIT and LOGIT show signi…cantly better …t than the LP model and

the estimated coe¢ cients from those models are generally more signi…cant than the

estimated coe¢ cients from the LP model.20 The selected economic indicators can

explain about 40% of the incidence when a country has a strategy, and appear to be

jointly highly signi…cant. We attribute the unexplained part to political, institutional

and country-speci…c factors. We now discuss the e¤ects of individual variables.

We …nd signi…cant evidence that as the GDP per capita in the country grows,

there is a higher probability that the government will have a public debt manage-

ment strategy. Also, the level of indebtedness shows signi…cantly positive relation-

ship with the probability of an existing strategy. This implies that as countries get

more indebted they put more weight on e¤ective debt management where a strategy

document is the basic building block.

The coe¢ cient attached to the share of government on GDP is negative but

statistically indi¤erent from zero. So that increasing importance of government in

economic performance of a country does not seem to increase the probability of a

varQML

b = bH1bgbg0 bH1
where bH and bg are the gradients (scores) and Hessian of the log likelihood evaluated at the ML

estimates.
19None of the countries from SAR happend to be included in the regression.
20Note that the coe¢ cients from the PROBIT and LOGIT model cannot be interpreted as mar-

ginal e¤ects of an explanatory variable on the dependent variable as in the case of the LP model.
In the case of PROBIT and LOGIT the marginal e¤ect varies with the level of the explanatory
variable.

21

present debt management strategy.

The increasing degree of governments debt concessionality appears to signi…cantly

decrease the probability of a country having a debt management strategy. Therefore

reliance of a country on multilateral and bilateral donors may act as an disincentive

for adopting sound public debt management practice.

On the other hand, higher ‡exibility of an exchange rate regime seems to increase

the probability of a country adopting a debt management strategy. This is due to the

fact that under an exchange rate ‡oat the country has to deal with FX risk explicitly

and cannot rely on the central bank to defend the …xed exchange rate parity as its

intermediate policy target. In fact, with increasing ‡exibility of an exchange rate

regime the opportunity for government to contract out some part of the FX risk to

the central bank decreases.

The signi…cantly negative coe¢ cient attached to the standard deviation of CA/GDP

o¤ers the interpretation that the higher the external macroeconomic vulnerability the

lower the probability of sound public debt management policy. In other words, as

the volatility of external balances increases the probability of a country having a

debt management strategy decreases.

The e¤ectiveness of FX reserves management seems to increase the probability of

a country having a debt strategy. Successful management of FX reserves can thus be

seen as a positive externality for public debt management, especially if coordinated

with the public debt management. Increasing variations in o¢ cial transfers, yet

another type of an external shock that developing countries can face, appear to have

a signi…cant negative e¤ect on the probability of a country having a debt management

strategy.

Finally, it is interesting to observe that if a country is located in the ECA region

its probability of having a debt management strategy increases signi…cantly. This is

not true of the other regions considered.

22

4.3 Extension to Public and Benchmark Strategies

In this section we extend the regression analysis to the binary variables distinguishing

between public debt management strategies made available to the public and those

not available to the public, and further between strategies formulated in terms of

benchmarks and those in terms of guidelines. We thus try to explain two additional

binary variables yPi and y
B
i de…ned as

yPi = f
1 if strategy made public

0 if strategy not public
yBi = f

1 if strategic benchmark

0 if guidelines
(4)

Again we opt for the PROBIT model21 when investigating to what extend can

yPi and y
B
i be explained by selected economic indicators Xi. Since the number of

observations in our sample available for estimation changes noticeably due to a dif-

fering number of observations available for each explanatory variable we resort to

the speci…c-to-general approach to build up the …nal models estimated for yPi and

yBi . We start with GDP per capita and add on other relevant variables according to

their signi…cance while maximizing the coverage of the survey data. We …rst discuss

some additional variables that appear in the PROBIT model for yPi and y
B
i , and

which were not signi…cant when used to explain existence of a strategy, i.e. yi.

GDP growth - when economy performs well and is experiencing higher growth

rates of GDP the government may be more willing to become transparent about

its actions and decisions. This argument implies that with higher GDP growth

government’s capacity in meeting public’s demand for higher transparency in public

debt management grows as well. The series of GDP growth was obtained from the

Penn World Tables (Heston et al, 2006).

Terms of trade volatility - this variable captures the intensity of real external

shocks that hit the economy. Higher risk of real external shocks, as measured by the

standard deviation, creates a genuine dilemma for country authorities of whether to

21We have carried out the estimation using LOGIT and LP models as well to check on any
mispeci…cation problems. We did not detect any. The estimation results are available from the
author.

23

engage in relatively more accountable frameworks. This is due to the fact that more

intense real external shocks make even operational accountability more burdensome,

and therefore does not necessarily imply a reluctance of country authorities to be

accountable. Following this argument one can expect that higher terms of trade

volatility can result in inclination towards debt management guidelines rather than

strategic benchmarks for debt management. The series was acquired from the GDF

& WDI Central database of the World Bank.

The estimation results for the PROBIT models of yPi and y
B
i are reported in

Table (2). The respective log-likelihoods were again maximized using the BHHH

algorithm and the inference is based on Huber-White QML standard errors.

The estimation results in Table (2) indicate that the probability of a strategy

being available to the public can be from about 21 percent explained by selected

economic indicators. Similarly, the probability of a strategy being expressed in terms

of a strategic benchmark, rather than strategic guidelines, can be from about 47

percent explained by selected economic indicators, a percentage signi…cantly higher

than in the case of public strategies. The unexplained part of the probability that

yPi = 1 or y
B
i = 1 is attributed to institutional, political and idiosyncratic factors.

We now proceed to a more detailed discussion of our results.

Consider …rst the estimation results for yPi in the …rst column of Table (2). We

…nd that the level of GDP per capita, the level of indebtedness and the average

growth rate do not seem to be important in explaining a country’s decision to make

its debt management strategy public. Further, variations in economic growth seem to

negatively impact on the probability of a strategy being made public, however, this

impact is not signi…cant at common levels. On the other hand, increasing volatility of

domestic prices, as measured by CPI, signi…cantly a¤ects the probability of a strat-

egy being public. This result could be related to the e¤ect of volatility of in‡ation

on the uncertainty pertaining to government revenues. The government in defence

of its strategy and for the sake of accountability prefers to make its debt manage-

ment strategy available to the public so that the e¤ect of an unexpected shortfall

in government revenues and the e¤ect of volatile prices on …nancing premiums are

24

apparent. Furthermore, increasing standard deviation of the exchange rate seems to

be positively in‡uencing the decision to make a debt management strategy public.

More ‡exible exchange rate regimes are often associated with more advanced macro-

economic policy, such as e.g. in‡ation targeting, the peer pressure within government

institutions can result in more transparent public debt management. Volatility of

the terms of trade seems to have a signi…cant negative e¤ect on the probability of

a strategy being public. If a government is relatively more dependent on tax rev-

enues from tradeable goods and/or its revenues are directly linked to the country’s

exports, such as commodities, larger shocks to government revenues could make gov-

ernment reluctant to publish its strategy and later be forced to publicly modify it.

Variations in CA/GDP seem to be positively related to the probability of making

strategy public, though not at common signi…cance levels. When looking over the

regional dummies we …nd that if a country belongs to the ECA region it has signi…-

cantly higher probability of having made its strategy public. This is not true of the

remaining regions.

Consider now the estimation results for yBi in the second column of Table (2).

The e¤ect of GDP per capita on the probability that yBi = 1, i.e. a strategy is

expressed as a strategic benchmark rather than guidelines for debt management,

appears to be negative. This would imply that developed countries do not favour

strategic benchmarks. This may be due to the fact that they face only a certain type

of risk, most commonly interest rate risk, and even operational sta¤ shows relatively

high capacity for managing this risk, so that the relatively strict and more explicit

guidance of a strategic benchmark is not necessary. The level of indebtedness appears

to be positively related to having a strategic benchmark, however, this e¤ect is not

statistically signi…cant at common levels. The results suggest a negative e¤ect of

average GDP growth on the probability of a benchmark-type strategy. Tentatively

and in relation to developing countries, higher average GDP growth over the period

1990-2006 may indicate less disruptions of macroeconomic performance due to crises

episodes and thus the need of addressing basic risks explicitly is not seen as so

bene…cial. On the other hand, developed countries experience relatively lower average

growth rates compared to developing countries which might not have the necessary

25

analytical capacity to derive benchmark targets. Increasing variation in in‡ation does

not seem to a¤ect the probability of using strategic benchmarks. Higher ‡exibility

of applied exchange rate regimes seems to imply lower probability of a benchmark

debt strategy. This result is somewhat puzzling unless one wants to acknowledge the

in‡uence of more developed economies on this …nding, as most of these apply some

type of a ‡oating exchange rate regime and often use guidelines for debt management.

Also, the terms of trade volatility impact negatively on the probability of using a

strategic benchmark. If developing countries are often hit by large external shocks

it may be hard for them to set a conventional benchmark with rather well-de…ned

ranges for selected types of risks. Volatility of CA/GDP appears to be insigni…cant in

explaining the use of strategic benchmarks. Further, governments appear to be less

in favor of using strategic benchmarks for debt management once facing increasing

variation in o¢ cial transfers (foreign aid). Increasing variation in o¢ cial transfers

can be seen as a speci…c kind of an external shock that again results in the lower use

of benchmark targets (ranges) for management of the basic types of risk. Finally,

when looking across the coe¢ cient estimates attached to regional dummies we …nd

that if a country belongs to the MNA region it has signi…cantly lower probability of

using a benchmark strategy.

5 Conclusion

This paper analyzed survey data on public debt management strategies across income

groups, regions and levels of indebtedness using graphical tools. Further, regression

analysis was carried out to extend the graphical analysis and condition on more

economic indicators possibly relevant for public debt management. More speci…cally,

the graphical and regression analyses were focused on explaining how the incidence of

(i) public debt management strategies, (ii) the published strategies and (iii) strategic

benchmarks varies across income groups, regions, levels of indebtedness and other

economic characteristics.

We found that a higher level of income in a country appears to increase its prob-

ability of having a debt management strategy. The level of indebtedness seems to

26

be also positively correlated with the incidence of a strategy where the graphical

analysis indicated that the relationship could be non-linear. The latter would im-

ply that as a country becomes more indebted it aims at increasing the quality of

debt management, however, after reaching high levels of indebtedness it gives up on

debt management and possibly engages in debt renegotiations and focuses on debt

sustainability issues. Across the World Bank’s regions, Europe and Central Asia ap-

pears to stand out in regards to the incidence of strategies. Concerning other factors,

the degree of debt concessionality appears to signi…cantly decrease the probability

of having a strategy, and so does the volatility of external shocks.

In regards to making strategies public, it appears from the graphical analysis

that their incidence is slightly positively related to the income levels, however, the

regression analysis …nds this e¤ect insigni…cant. The public strategies seem to be also

unrelated to the level of indebtedness. From the regional perspective, it is Europe

and Central Asia, followed by East Asia and Paci…c, and Sub-Saharan Africa, that

leads in terms of transparency (incidence of public strategies) and outperforms in this

respect even OECD countries as a group. Concerning other factors, the volatility

of domestic and external shocks seems to signi…cantly a¤ect the probability of a

present public strategy. In general, the degree to which we can explain the incidence

of public strategies is rather low compared to the incidence of strategies and strategic

benchmarks.

Incidence of strategic benchmarks seems to be slightly positively correlated with

income levels, however, when conditioning on other economic indicators we found a

signi…cant negative e¤ect of GDP per capita on a strategy being expressed in terms

of a benchmark. The relationship between benchmark strategies and levels of in-

debtedness appears slightly positive but not signi…cant at common levels. Europe

and Central Asia, and Latin America and the Caribbean appear to have the high-

est incidence of benchmarks among the World Bank’s regions. When conditioning

on other economic factors we …nd that countries from the Middle East and North

Africa have signi…cantly lower probability of using benchmark strategies compared

to other regions. Further economic factors which signi…cantly a¤ect the incidence of

a benchmark strategy include average GDP growth and volatility of external shocks,

27

with a negative e¤ect on benchmarks incidence, and volatility of domestic shocks,

with a positive e¤ect on benchmarks incidence.

As mentioned in the introduction, we see this paper as a …rst attempt to charac-

terize the variations in the survey data on public debt management strategies across

countries where establishing a regularly repeated survey would be incredibly bene…-

cial. The follow-up surveys on public debt management strategies could extend the

coverage to all low income countries, and focus on distinguishing between implicit

and formal strategies by structuring the applied questionnaire accordingly. Although

the present work is intended to provide the opportunity for public debt managers

to compare themselves to their peers or countries at a similar stage of development,

some policy implications could be derived from a similar analysis in the future. In

this respect inclusion of some institutional variables would be desirable. Examples

could include the number of institutions responsible for public debt management in

a given country, the degree of central bank independence, existence of a medium-

term expenditure framework, or the degree of transparency of measures concerning

domestic macroeconomic policies.

28

References

[1] Alani, E.M.A.A, (2006). Crowding-out and Crowding-in e¤ect of Government

Bonds Market on Private Sector Investment. Institute of Developing Economies

Discussion Paper 74. Chiba.

[2] Briotti, M.G. (2005). Economic Reactions to Public Finance Consolidation: A

Survey of the Literature. ECB Working Paper Series 38 (October). Frankfurt

am Main.

[3] Buiter, W.H. (2006). How Robust is the New Conventional Wisdom in Monetary

Policy? The 2006 Central Bank’s Symposium "Challenges to Monetary Theory".

Bank of England. London.

[4] Burnside, C. (2004). Assessing New Approaches to Fiscal Sustainability Analy-

sis. Working Paper, Duke University. Durham.

[5] Friedman, B.M. (1978). Crowding Out or Crowding In? The Economic Conse-

quences of Financing Government De…cits. NBER Working Paper Series 284.

Cambridge.

[6] Gill, I., Pinto, B. (2005). Public Debt in Developing Countries: Has the Market-

Based Model Worked? World Bank Research Working Paper 3674 (August).

Washington, D.C.

[7] Heston, A., Summers, R., Aten, B. (2006). PennWorld Table Version 6.2. Center

for International Comparisons of Production, Income and Prices at the Univer-

sity of Pennsylvania. September 2006.

[8] Johnston, J, Dinardo, J. (2001). Econometric Methods. McGraw-Hill, Fouth

Edition.

[9] Missale, A. (1999). Public debt management. Oxford Press. University of Oxford.

[10] Wheeler, G. (2004). Sound Practice in Government Deb Management. World

Bank. Washington, D.C.

29

[11] IMF, WB (2004). Guidelines for Public Debt Management: Accompanying Doc-

ument and Selected Case Studies. The International Monetary Fund and the

World Bank. Washington, D.C.

[12] IMF, WB (2001). Guidelines for Public Debt Management. The International

Monetary Fund and the World Bank. Washington, D.C.

[13] Elmendorf, D.W., Mankiw, G.N. (1999). Government Debt. In Handbook of

Macroeconomics, vol. 1, Taylor, J.B., Woodford, M. (eds.), Elsevier Science,

B.V., 1999, pp. 1615-1699.

[14] Togo, E. (2007). Coordinating Public Debt Management with Fiscal and Mon-

etary Policies: An Analytical Framework. manuscript, Banking and Debt Man-

agement Department, World Bank. Washington, D.C.

[15] UNITAR (2004). A General Overview of Legal Issues in Debt Management.

UNITAR’s Occasional Series 1. Geneva.

30

Table 1: Estimation Results - PROBIT Model for Existing Strategies
PROBIT LOGIT LP

Variable Coe¢ cient p-value Coe¢ cient p-value Coe¢ cient p-value
GDPpci 6:7E-5 0:0403 0:0001 0:0342 1:6E-6 0:8470
Indebti 0:0139 0:0421 0:0232 0:0536 0:0019 0:1902
GovSharei 0:0269 0:2931 0:0414 0:3557 0:0065 0:3558
(ConsDebt/TotDebt)i 0:0216 0:0962 0:0329 0:1485 0:0085 0:0480
stdev (inf)i 0:0004 0:4686 0:0007 0:5427 9:7E-5 0:6654
stdev (growth)i 0:0649 0:2863 0:1134 0:2721 0:0182 0:3182
stdev (ER)i 2:2E-6 0:1137 4:4E-6 0:1261 2:5E-7 0:3143
stdev (CA/GDP)i 0:1694 0:0241 0:3008 0:0264 0:0222 0:0672
stdev (FXRes/IM)i 3:8911 0:0262 7:0172 0:0451 0:5549 0:0467
coefvar(O¢ cTrans)i 0:2400 0:0003 0:4093 0:0007 0:0133 0:2825
constant 0:0231 0:9789 0:3039 0:8330 0:7743 0:0035
dummy-AFRi 0:2648 0:6601 0:5079 0:6290 0:0239 0:8837
dummy-EAPi 0:3947 0:5806 0:7235 0:5543 0:0336 0:8885
dummy-ECAi 1:2259 0:0704 2:0451 0:0965 0:2284 0:1330
dummy-MNAi 0:4327 0:5740 0:7880 0:5771 0:1288 0:5920
McFaddens R-squared 0:4039 0:3973 0:3583
No. of Countries 81 81 81

Dependent Variable = 1 57 Dependent Variable = 0 24

31

Table 2: Estimation Results - PROBIT for public strategies and benchmarks
Dep.Var. yPi Dep.Var. y

B
i

Explanatory Variable Coe¢ cient p-value Coe¢ cient p-value
GDPpci 1:4E-5 0:6618 0:0001 0:0228
Indebti 0:0010 0:8748 0:0132 0:1462
Growthi 0:1566 0:3299 0:5149 0:0107
stdev (Growth)i 0:1402 0:1568 0:4840 0:0140
stdev (inf)i 0:0014 0:0165 2:7E-5 0:9627
stdev (ER)i 1:9E-6 0:0435 0:0015 0:0032
stdev (tot)i 0:1339 0:0060 0:2324 0:0047
stdev (CA/GDP)i 0:1525 0:2200 0:0789 0:6736
coefvar(O¢ cTrans)i na na 0:7288 0:0015
constant 0:4155 0:6421 2:6197 0:0471
dummy-AFRi 1:3422 0:1840 0:7743 0:4391
dummy-EAPi 0:7398 0:3510 1:6985 0:1584
dummy-ECAi 1:5153 0:0228 0:7142 0:4200
dummy-MNAi 0:4929 0:4975 4:3164 0:0024
McFaddens R-squared 0:2055 0:4679
No. of Countries 60 58
Dependent Variable = 1 42 28
Dependent Variable = 0 18 30

32

Table 3: A list of countries included in regression analysis
ALBANIA CANADA FINLAND JORDAN NORWAY SWAZILAND

ALGERIA CHILE FRANCE KAZAKHSTAN PANAMA SWEDEN

ARGENTINA CHINA GABON KOREA PARAGUAY SYRIAN ARAB REP.

AUSTRALIA COLOMBIA GERMANY LATVIA PERU THAILAND

AUSTRIA COSTA RICA GREECE LEBANON PHILIPPINES TRINID . & TOBAGO

AZERBAIJAN CROATIA GUATEMALA LITHUANIA POLAND TUNISIA

BELARUS CZECH REP. HUNGARY LUXEMBOURG PORTUGAL TURKEY

BELGIUM DENMARK ICELAND MACEDONIA ROMANIA UKRAINE

BELIZE DOMINICAN REP. INDONESIA MALAYSIA SEYCHELLES UNITED KINGDOM

BOLIVIA ECUADOR IRELAND MAURITIUS SLOVAK REP. UNITED STATES

BOSNIA & HERZ. EGYPT ISRAEL MEXICO SLOVENIA VENEZUELA

BOTSWANA EL SALVADOR ITALY MOROCCO SOUTH AFRICA

BRAZIL EQUAT. GUINEA JAMAICA NETHERLANDS SPAIN

BULGARIA ESTONIA JAPAN NEW ZEALAND ST . VIN . & GREN.

Figure 1: Distribution of the Percentage of Countries with Strategies, and the Per-
centage of Public Strategies and Strategic Benchmarks out of Strategies Across In-
come Groups

33

Figure 2: Distribution of the Percentage of Countries with Strategies, and the Per-
centage of Public Strategies and Strategic Benchmarks out of Strategies Across Re-
gions

34

Figure 3: Distribution of the Percentage of Countries with Strategies, and the Per-
centage of Public Strategies and Strategic Benchmarks out of Strategies Across Levels
of Indebtedness

35

Figure 4: Distributions of the Basic Types of Risk Addressed in Benchmarks Across
Income Groups, Regions and Levels of Indebtedness

36

Policy Research Working Paper Series

Title Author Date Contact for paper

WPS4263HIV/AIDSandSocialCapitalina AntonioC.David June2007 A.David
Cross-SectionofCountries 82842

WPS4264FinancingofthePrivateSectorin ConstantinosStephanou June2007 S.Coca
Mexico,2000–05:Evolution, EmanuelSalinasMuñoz 37474
Composition,andDeterminants

WPS4265TheStructureofImportTariffsinthe OleksandrShepotylo June2007 P.Flewitt
RussianFederation:2001–05 32724

WPS4266TheEconomicCommunityofWest SimpliceG.Zouhon-Bi June2007 S.Zouhon-Bi
AfricanStates:FiscalRevenue LyngeNielsen 82929
ImplicationsoftheProspective
EconomicPartnershipAgreement
withtheEuropeanUnion

WPS4267FinancialIntermediationinthe HeikoHesse June2007 G.Johnson
Pre-ConsolicatedBankingSectorin 34436
Nigeria

WPS4268PowertothePeople:Evidencefrom MartinaBjörkman June2007 I.Hafiz
aRandomizedFieldExperimentofa JakobSvensson 37851
Community-BasedMonitoringProject
inUganda

WPS4269ShadowSovereignRatingsfor DilipRatha June2007 N.Aliyeva
UnratedDevelopingCountries PrabalDe 80524
SanketMohapatra

WPS4270Jump-StartingSelf-Employment? RitaAlmeida June2007 A.Bonfield
EvidenceamongWelfareParticipants EmanuelaGalasso 31248
inArgentina

WPS4271Construction,Corruption,and CharlesKenny June2007 C.Kenny
DevelopingCountries 33540

WPS4272Migration,Remittances,Poverty, DavidMcKenzie July2007 M.Sasin
andHumanCapital:Conceptualand MarcinJ.Sasin 36877
EmpiricalChallenges

WPS4273RulesofOriginandtheWebofEast MiriamManchin July2007 L.Yeargin
AsianFreeTradeAgreements AnnetteO.Pelkmans-Balaoing 81553

WPS4274AreLaborRegulationsDriving MohammadAmin July2007 S.Narsiah
ComputerUsageinIndia’sRetail 88768
Stores?

WPS4275CanForeignLobbyingEnhance KishoreGawande July2007 V.Cornago
Development?TheCaseofTourism WilliamMaloney 84039
intheCaribbean GabrielV.MontesRojas

WPS4276HumanCapital,TradeLiberalization, TomKrebs July2007 V.Cornago
andIncomeRisk PravinKrishna 84039
WilliamMaloney

WPS4277ClimateChangeAdaptationinAfrica: SungnoNiggolSeo July2007 P.Kokila
AMicroeconomicAnalysisof RobertMendelsohn 33716
LivestockChoice

WPS4278EndogenousIrrigation:TheImpactof PradeepKurukulasuriya July2007 P.Kokila
ClimateChangeonFarmersin RobertMendelsohn 33716
Africa

WPS4279TheImpactofClimateChangeon SungnoNiggolSeo July2007 P.Kokila
LivestockManagementinAfrica: RobertMendelsohn 33716
AStructuralRicardianAnalysis

Policy Research Working Paper Series

Title Author Date Contact for paper

WPS4280 GovernanceMattersVI:Aggregate DanielKaufmann July2007 R.Bonfield
andIndividualGovernance: AartKraay 31248
Indicators,1996-2006 MassimoMastruzzi

WPS4281 CreditGrowthInEmergingEurope: SophieSirtaine July2007 S.Sirtaine
ACauseForStabilityConcerns? IliasSkamnelos 87006

WPS4282 AreCashTransfersMadetoWomen NorbertSchady July2007 I.Hafiz
SpentLikeOtherSourcesofIncome JoséRosero 37851

WPS4283 InnovationShortfalls WilliamMaloney July2007 V.Cornago
AndrésRodríguez-Clare 84039

WPS4284 CustomerMarketPowerandthe NeeltjeVanHoren July2007 M.Gamboa
ProvisionofTradeCredit: 34847
EvidencefromEasternEuropeand
CentralAsia

WPS4285 PovertyAnalysisUsingAn J.A.L.Cranfield July2007 P.Flewitt
InternationalCross-CountryDemand PaulV.Preckel 32724
System ThomasW.Hertel

WPS4286 InstitutionalEffectsasDeterminants JesúsÁlvarez July2007 S.Baksh
ofLearningOutcomes:Exploring VicenteGarcíaMoreno 31085
StateVariationsinMexico HarryAnthonyPatrinos