_VPS2050
POLICY RESEARCH WORKING PAPER   2050
Measuring Aid Flows                                                    Conventional measures of aid
are not desianed to estimate
the overall aid content of
A  New   Approach                                                      financial flows. Furthermore,
they typically overstate the
Charles C. Chang                                                       grant elements of
Eduardo Fernidndez-Arias                                               concessional loans, thus
Luis Serv~n                                                            understating relative aid flows
to recipients getting mostly
grants and from donors
giving mostly grants land
loans in high-yield currencies).
A new approach to
measuring aid flows-
Effective Development
Assistance - focuses on tre
overall grant equivalent of
official financial flowvs and
allows meaningful
comparisons of recipients or
donors.
The World Bank
Development Research Group
Macroeconomics and Growth
February 1999



LPOLICY RESEARCH WORKING PAPER 2050
Summary findings
Debate about the effectiveness of foreign aid has               They implement their approach empirically using data
intensified in recent years, as budgetary pressures on aid    on some 40,000 official loans from the World Bank's
have increased in donor countries. Whatever the merits        DRS database -  virtually all of the official loans to 133
of opposing arguments, the question is: do conventional       developing countries from  1975 to 1995. T-he numerical
measures of aid (such as OECD's Net ODA), which lump          results underscore several points:
together grants and loans, accurately reflect true aid           * The conventional approach has led to systematic
flows?                                                        overestimates of the concessionalitv of official loans.
Chang, Fernandez-Arias, and Serven analyze the              This overestimate has increased significantly since the
methodological shortcomings of conventional measures          mid-1980s. Conventional methods show a rising trend;
of aid and propose a new approach, which measures             the new method shows the opposite.
official aid flows as the sum of grants and the grant-             Net ODA increasingly overstates the rtue aid
equivalents of official loans (in a new aggregate they call   content of official flows, although the divergence
Effective Development Assistance, or EDA).                    between the two approaches is somewhat muted by the
They show how results using this conceptually               rising relative importance of grants over loans in total
superior measure may differ significantly from                official flows.
conventional aggregates, providing a quite different view
on major aid trends.
This paper - a product of Macroeconomics and Growth, Development Research Group - is part of a larger effort in the
group to understand the effectiveness of development aid. Copies of the paper are available free from the World Bank,l 818
H Street NW, Washington, DC 20433. Please contact Emily Khine, room MC3-341, telephone 202-473-7471, fax 202-
522-3518, Internet address kkhine@worldbank.org. The authors may be contacted at cchang6co.worldbank.org or
lservenC@worldbank.org. February 1999. (39 pages)
The Policy Researchi Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about
developmnent issues. An objective of the series is to get the findings out quickly, even if the presentations are less than ftlly polished. The
papers carry the narnes of the authors and should be cited accordingly. The findings, interpretations, and concluslions expressed in this
paper are entirely those of the authors. They do not necessarily represenit the viewv of the WVorld Bank, its Executtive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center



MEASURING AID FLOWS: A NEW APPROACH
Charles C. Chang
The World Bank
Eduardo Fernandez-Arias
Inter-American Development Bank
The World Bank
Luis Serven
The World Bank
We would like to thank Tove Edstrand for able research assistance and Vilas Mandlekar, Gloria Reyes, Shelly Fu and
Punam Chuhan for their crucial help and advice throughout this project. The views expressed in this paper are ours
only and do not necessarily represent official opinions of the World Bank, the Inter-American Development Bank or
member countries of either institution.



2
INTRODUCTION
Many poor aid-recipient countries view foreign aid as a critical ingredient in their
development strategy, even though its development effectiveness remains in question among many
economists. At the same time, the level and trends of foreign aid are increasingly becoming sensitive
issues in donor countries' budgetary discussions, with analysts observing increasing signs of "donor
fatigue". In particular, International Financial Institutions have expressed concerns regarding the
level of overall development aid and the possible crowding out of poor traditional recipients by
former socialist economies. Whatever the merits of these views, the key issue arises of whether the
aid aggregates commonly used by policymakers and researchers in their assessments of development
aid provide an accurate measure of true aid flows. In this paper we analyze the traditional
methodology underlying these conventional measures and propose a new approach.
Official financial flows are traditionally classified as concessional on the basis of the OECD's
Official Development Assistance (ODA) classification, and aid flows are traditionally measured by
the corresponding Net ODA statistic. For example, the World Bank's Global Development Finance
uses Net ODA information to analyze trends in aid flows over time as well as across recipients and
donors. Despite its popularity, however, the methodology underlying Net ODA aggregates suffers
from a number of shortcomings. Consequently, the analysis of aid flows needs to be revisited in the
light of more satisfactory measures based on improved methodologies. In this paper we analyze the
nature of the proposed improvements and illustrate them with a comparative assessment of the
overall trends in aid flows to 133 developing economies.'
The remainder of the paper is structured as follows. Section 1 summarizes the shortcomings
of conventional measures of aid, and Section 2 proposes a new approach to overcome them. The
implementation of this new approach leads to an alternative aid measure which we label Effective
Development Assistance (EDA), which is developed in Section 3. Section 4 illustrates the proposed
method by re-assessing trends in aid flows on the basis of EDA, and comparing them with those
implied by Net ODA. Finally, Section 5 offers some concluding remarks.
1. WHIAT IS WRONG WITH CONVENTIONAL FOREIGN AID MEASURES?
Foreign aid is conventionally measured on the basis of the OECD's ODA, a concept
introduced in the early 1 970s. ODA comprises official financial flows with a development purpose in
the form of grants (inclusive of those tied to technical assistance) and highly concessional loans.
Loans are defined as highly concessional when their grant element -- i.e., the subsidy implicitly
included in the loan, relative to the loans' face value -- is at least 25 percent, as measured by a
formula to be analyzed in depth in the next section. The leading measure of foreign aid flows is the
so-called Net ODA, which is the net disbursement amount, i.e., disbursements minus amortization, of
those flows classified as ODA.
The resulting improved aid data, disaggregated by recipient country and donor type, is available on the
internet at http://www.worldbank.org/htmn/prdmg/grthweb/ddaid.htm.



3
Is Net ODA an appropriate measure of aid flows ? Conceptually, international finance
flowing to capital-scarce developing countries may involve efficiency gains even if the flows accrue
on market terms - as long as the funds are used appropriately. Such efficiency gains translate into net
financial gains for the recipient countries. The main purpose of measuring foreign aid flows is to
assess the portion of those gains that is due to a pure transfer of resources from donors to recipients
through below-market, subsidized financial terms -- i.e., to assess the donors' net financial cost,
rather than the (presumably larger) recipients' benefit.
Net ODA, however, does not accurately measure the cost that donors incur in connection
with their aid (especially debt) flows, and as a result the evolution of Net ODA over time, as well as
across donors and recipients, likely provides a distorted picture of aid trends. This distortion is due
to seven conceptual shortcomings of Net ODA that we detail below.
Shortcomings of Net ODA: Gross and Net Flows
1. Under-estimation of the aid content due to netting out. The financial cost involved in donors' aid
provided in a given year is a forward-looking concept reflected in the fractional value of the debt
service claims acquired in exchange for up-front (gross) disbursements in that period, irrespective of
the amortization of previously contracted debt obligations. Therefore, on this account, the net flow
nature of Net ODA, i.e., disbursements rninus amortization, underestimates the aid content of
disbursed flows by netting out amortization payments. For example, a constant flow of ilentical
highly concessional loans over time entails a continuous cost on the part of the donor but yields a
zero Net ODA flow, since amortization payments exactly offset disbursements. In such case, Net
ODA would completely fail to capture the aid content of flows.
Aside from this netting out involved in Net ODA, the rest of the shortcomings relate to the
flow amounts classified as ODA:
Shortcomings of ODA: Aggregation and Coverage
Design shortcomnings of ODA related to loan coverage and aggregation obscure the
interpretation of this measure. The three main conceptual problems are:
2. Over-representation of loans with high concessionality. ODA includes the full face value of both
grants and highly concessional loans without distinguishing between the two. However, concessional
loans entail repayment obligations, and, therefore, the aid they involve, i.e. the net financial cost to
donors, is only a fraction of their face value. The inclusion in ODA of the full face value of these
loans overestimates their aid content. Only grants, that is to say pure unrequited transfers, should be
accounted at full value.2
3. Under-representation of loans with low concessionality. Under the ODA definition, non-
concessional loans include loans on market terms as well as concessional loans with low degree of
concessionality. The aid content of the latter - i.e., the donors' cost involved in these loans -- is
therefore not captured by ODA.
2Assuming that the grant is not tied or subject to other financial quid pro quo.



4
4. Coverage. The inclusion in ODA of official technical assistance grants by their full value can be
seen as another shortcoming. In this case, the donor benefits from payments received in return for
the technical assistance supplied, and this may greatly reduce the donor's net financial cost.3
Shortcomings of ODA: The Grant Element
As noted above, ODA is based on a sharp distinction between concessional and non-
concessional loans, drawing from their respective grant elements. Conceptually, the calculation of the
grant element, i.e., the degree of concessionality, involves the computation of the expected present
value of the stream of debt service obligations associated with the loan under consideration. To the
extent that the discount rate utilized reflects the creditor's opportunity cost, i.e., the return it could
make on alternative investments of the same capital, this present value measures the economic value
of debt service repayments and, on this account, the financial value of the loan. The grant element of
the loan is the portion of the loan that, at a given time, is not expected to be repaid, i.e. the shortfall
of the above-mentioned present value relative to the amount disbursed.
For the purposes of ODA, loans are classified as concessional if their grant element exceeds
25 percent, and as non-concessional (and hence ignored) otherwise. The grant elements are
computed using some special assumptions, however: most importantly, loan interest rates (used to
compute interest charges) are assumed to remain constant throughout the life of the loan, and a fixed
10 percent discount rate is utilized in all present value calculations. This methodology for computing
grant elements contains a number of shortcomings, which may lead to loan misclassification and
distortion of ODA figures across time, donors, and recipients:
5. Discount Rates. In order to reflect donors' opportunity costs, the discount rates used for present
value calculations should correspond to applicable market rates. The fixed 10 percent discount rate
utilized in ODA fails that test on at least three important dimensions to which it should be sensitive,
namely time, currency, and maturity:
a) Time. Discount rates should evolve over time with market conditions prevailing at the time the
aid content of loans is estimated. For example, to measure the donors' cost as seen at the time of
loan disbursement, the market terms prevailing at that time should be used.
b) Currency. At any point in time, market rates, and therefore appropriate discount rates, are
currency specific. The discount rate should follow the currency in which debt service is payable.
c) Maturity. At any point in time and for any given currency, market rates depend on the length of
the repayment period according to the so-called yield curve. Therefore, the discount rates applied to
the debt service stream should vary over the life of the loan according to the timing of service
payments.
3The exclusion of private source financing from ODA coverage appears justified, however, because
commercial lending contains no aid by definition.



5
6. Variable Rate Loans. In the case of variable rate loans, the construction of the future debt service
stream requires a forecast of interest rate charges. This is especially important for floating rates
linked to future market conditions (e.g., indexed to six-month LIBOR). ODA makes no attempt to
predict these conditions and implicitly assumes that, like in the case of fixed-rate loans, variable rates
will remain constant at their level at the time of disbursement.
7. Credit Risk. In the absence of credit risk, the 'market rates' mentioned above are risk-free: rates.
However, credit risk is implicitly an additional source of donor financial cost from an economic
viewpoint -- as opposed to a contractual, legal perspective. As such, it should be incorporated. in the
calculation of the grant element, either through augmenting the discount rate or through the
utilization of lower expected debt service projections, or both. This is admittedly a difficult taslk, as it
would require the use of borrower-specific risk spreads and/or default probabilities. In any case,
ODA makes no attempt to adjust for credit risk.
2. A NEW APPROACH: EFFECTIVE DEVELOPMENT ASSISTANCE
On the whole, the methodological shortcomings of Net ODA just summarized underscore the
need for an alternative approach to the measurement of aid flows. Our proposed approach is based
on the grant equivalent of financial flows. We first elaborate on grant equivalents and then explain
the methodology for computing the adjusted foreign aid measure, which we call Effective
Development Assistance, or EDA. This section provides a brief overview of the key issues; the
analytical details are described at length in the next section.
Measuring the Aid Content of Financial Inflows: Grant Equivalent and Grant Element
The grant equivalent of a financial inflow is the amount that, at the time of its commitnment, is
not expected to be repaid, i.e., the amount subsidized through below-market terms at the time of
commitment. By definition, the grant equivalent of a pure grant is the amount of the grant itself In
contrast, the grant equivalent of a concessional loan is only part of the loan amount, and becomes
negligible as loan terms approach market terms. In other words, the grant equivalent, G, measures
the shortfall between the loan amount disbursed, D, and the present value of the associated expected
debt service obligations, E.4  Correspondingly, a loan can be interpreted as including two
components: a loan component, equal to the debt equivalent E, and a grant component, equal to the
grant equivalent G, so that a concessional loan disbursed by an amount D is equivalent to a non-
concessional loan by an amount E and a pure grant by an amount G:
G=D-E                                          (2.1)
By definition, the grant element, g, measures the grant equivalent as a proportion of the
inflow disbursed. Therefore, in terms of the more familiar grant element:
4Conceptually, this present value E was first made operational and utilized to reduce bond exchanges under
the Brady debt reduction program to comparable debt equivalent terms (see Claessens, Diwan, and Fernandez-Arias,
1992), and was later utilized to measure the debt equivalent of concessional debt stocks for the purpose of measuring
country indebtedness (e.g., in the Heavily Indebted Poor Countries (HIPC) initiative).



6
G                                             (2.2)
gD
In the extreme case of a pure grant, no service payments are called for (E=O), the grant
equivalent is the grant itself (G=D) and the grant element (expressed as a percentage) is g=100
percent. At the other extreme, in financially "fair" lending, debt service equals disbursements in
present value (E=D) and the grant equivalent and grant element are zero. This is the case expected
of competitive commercial lending and, at least approximately, is also the case of market-based, non-
concessional official lending.5 However, official creditors providing "soft" loans, i.e., concessional
lending, extend loans with relatively low debt service (E<D), whose grant equivalent and grant
element are consequently positive.6
We define EDA as the sum of the grant equivalents of all development flows disbursed in a
given period, thus eliminating the distortion introduced by the deduction of amortization payments in
Net ODA noted in shortcoming (1) above. This simple method of converting all inflows to a
common denominator also eliminates the over- and under-estimation of the aid content of loans as
measured by ODA, allowing for a correct aggregation of the aid included in grants and loans in terms
of comparable units. It therefore solves the ODA shortcomings (2) and (3) above.
In order to address the coverage shortcoming (4), and following traditional World Bank
methodology, grants tied to technical assistance are excluded from EDA.7 This is just a first
approximation to the ideal procedure by which the grant equivalent of tied grants would be adjusted
by deducting the benefit that the donor enjoys from the quidpro quo.
The grant content of a loan whose service due extends over T periods (e.g., years) can also
be analyzed by focussing on each period t (ranging from 1 to T) in the life of the loan and comparing
the contractually-determined interest rate it applicable to the period in question with the discount rate
dt employed to bring to present-value terms the service payments due in that period. If the interest
rate coincides with the discount rate (it=d,) in every period, the grant element and grant equivalent
are null irrespective of the debt service time profile (e.g., the grace and maturity of the loan). If the
interest rate is consistently below the discount rate (i,<dt), the grant element and grant equivalent are
both positive. The larger the gap between the interest and discount rates, the larger the absolute
value of the grant element. For any given gap between the two rates, the grant element becomes
larger as debt service payments are stretched over a longer time period by backloading amortization
payments (e.g. longer grace period and maturity), reaching its maximum in the case of a consol, i.e.,
a loan with an infinite grace period (see the details in Appendix B).
' In practice, lags between commitment and disbursement times may result in small deviations from this
benchmark. Likewise, price-smoothing practices of Multilateral Development Banks may lead to cyclical variation in
grant elements: with loan interest rates determined as a moving-average of current and past market interest rates, loan
rates tend to lag behind market rates, so that grant elements temporarily rise when market rates are rising and fall
when market rates are falling.
6 The expressions in the text implicitly assume that loans are disbursed in a single installment, and need
modification if this is not the case. The general formulas are presented in the next section.
7 However, in the data files made available with this paper, technical assistance is attached as a memo item
for interested users.



7
One implication of the above discussion is that as long as the appropriate discount rate is
below 10 percent, the grant element calculations underlying ODA - which make use of the arlbitrary
10 percent discount rate - will lead to a systematic over-estimation of the grant element and. grant
equivalent of loans. Further, such overestimation is larger the longer the duration of the loan.
Therefore, non-concessional loans would be incorrectly shown as concessional, and marginally
concessional loans would be shown as highly concessional, especially if they are long-term.8
To avoid this distortion, our proposed method to compute grant elements and grant
equivalents is based on market discount rates sensitive to currency, timing, and maturity, thus
addressing shortcoming (5) of ODA as described in the previous section. In essence, this is achieved
by extracting the discount rates from the yield curve for risk-free instruments prevailing at the time of
disbursement. We do this separately for each currency under consideration. These time-, curTency-
and maturity-specific discount rates are used to bring future debt service to present-value terms or, in
other words, to compute the debt equivalent E.
As part of this calculation, our proposed method involves the construction of the debt service
payment stream, which is accomplished taking into account the amortization schedule and interest
terms of each loan. In the (relatively frequent) case of variable-rate debt, this in turn requires a
forecast of the interest rates applicable in future years, in order to address shortcoming (6) of ODA.
As explained in detail below, we obtain such forecast making use of the yield curve of the relevant
currency.
Donors' Effort and Expected Aid
It should be noticed that the donors' "effort", that our improved aid measure tries to capture,
refers to donors' voluntary net financial costs, as opposed to the financial costs realized ex-post,
which are partly determined by the realization of market rates over time. For that reason, d.onors'
expectations should refer to the conditions at the time at which financial commitments were made.
Consequently, in this proposed approach expected interest rates are derived from the imarket
conditions prevailing at the time of commitment, rather than those prevailing at a later date or
realized ex-post.
Credit risk, i.e., the risk of default, is also a factor to consider in the measurement of dionors'
expected financial losses and, consequently, grant elements. Traditional ODA measures ignore this
factor, as it was pointed out in shortcoming (7). While we recognize that credit risk is a factor to be
taken into account in the new approach being proposed, either by a downward adjustment of the
expected debt service stream or by augmenting the discount rate to reflect a risky opportunity cost,
we have chosen not to make adjustments on this account in our EDA calculations. Correspondingly,
our EDA statistics are subject to the caveat that they reflect contractual aid, as opposed to aid
inclusive of the anticipation of failures to comply with contractual payments.
8 This explains the sometimes large ODA-based grant element of commitments of private creditors to
developing countries, as reported in the World Bank's Global Development Finance.



8
The case has been made that multilateral lending, and perhaps official lending in general,
carries a negligible risk of default. If official lenders perceive that to be the case, then the previous
caveat is not relevant. This case is supported by the fact that open defaults on official lending are
rare. However, debt reductions and condonations of various kinds are common, especially of
bilateral debts. The view that default risk is negligible is consistent with the interpretation that this
kind of debt relief is additional voluntary aid.9 Nevertheless, an alternative interpretation is that this
debt relief is the manifestation of contract breaching on the part of the debtor. Furthermore, some
analysts view credit risk on official lending higher than it appears by virtue of being hidden by new
lending at higher concessional terms to sustain full service of all debts. In the absence of an accepted
methodology, we chose to neglect credit risk in the proposed approach at this time. 10
3.  EMPIRICAL IMPLEMENTATION
We next summarize the procedure followed to construct our proposed measure of Effective
Development Assistance (EDA) defined above. We proceed in two stages: first we describe the data
used, and then we spell out in more detail the application of the methodological framework
introduced in the previous section.
3.1   Data
Construction of effective development assistance along the lines described in section 2
requires two basic pieces of information: first, detailed data on the volume and features of official
financial flows -- specifically, the characteristics of the disbursement and repayment schedules,
interest charges and other fees of official loans -- and, second, a suitable set of discount rates that
properly reflect opportunity costs.
Financial flow data
As noted earlier, official financial flows consist of loans and grants, with the latter sometimes
defined to include technical assistance (TA). Figure 3.1 .presents the historical distribution of official
flows among these three categories. As discussed in the previous section, the inclusion of technical
assistance in aid aggregates is highly questionable, due to its quid pro quo nature. For this reason,
we restrict the focus to official flows exclusive of TA.
Since we have defined EDA as the sum of official grants plus the grant equivalent of official
development-oriented loans, measuring EDA involves the computation of the grant equivalents of
official loans. This is inherently a loan-by-loan task requiring information on grace, maturity, interest,
9 Because we adopt an ex-ante perspective, our EDA calculations do not consider debt relief, be it in the form
of debt forgiveness or debt rescheduling, as new aid. Debt relief, or the accumulation of arrears for that matter,
essentially amounts to an ex-post modification of the debt service schedule that involves involuntary aid, thus outside
the scope of EDA. Notice that conventional Net ODA accounts for debt relief by recording a fictitious disbursement
matched with an equally fictitious loan repayment, with no effect on Net ODA totals.
'1 The assessment of the credit risk of official debt is mnired with a number of technical difficulties (see
Demirguc-Kunt and Fernandez-Arias 1992 for details). It is clear that such assessment requires an analysis of the
conditions of each recipient country over time. Perhaps less obvious, credit risk assessments should also discriminate
among official creditors.



9
disbursement profiles, repayment schedules, and any other contractual provision, relevant to the
expected cash flow streams associated with each official loan included in the analysis.
Such highly-detailed information was obtained from the World Bank's Debtor Reporting
System (DRS) -- the main database underlying the World Bank's Global Development Finance. The
DRS, established in 1952, is debtor-based -- i.e., the data is compiled from reports provicled by
borrowing countries. Its coverage is quite broad: it includes all bilateral, multilateral and private
creditors,'" and all aid-receiving members of the World Bank.12
The DRS is very similar to the OECD's Creditor Reporting System (CRS), which provides
the basis for the OECD's Geographical Distribution of Financial Flows to Aid Recipients. There are
some differences, however, between the two systems. The CRS encompasses some 180 recipients,
including a considerable number of dependent territories and colonies that are excluded from the
DRS. Apart from these sub-sovereign entities, the recipient coverage of both systems is nearly
identical, with the main difference being the exclusion from the DRS system of a handful of countries
that have graduated from the World Bank's lending programs.'3 Also, in contrast with the dlebtor-
based DRS, the CRS collects information directly from the creditors that it covers -- members of the
OECD Development Assistance Committee (DAC) and multilateral lending agencies. As described in
more detail below, this may lead to some discrepancies between the two sources.
In practice, our choice of the DRS as the basic source of loan data imposes some limritations
on the time dimension of the analysis. Although the DRS contains information on loans contracted as
far back as the mid 50s, loan coverage is somewhat limited prior to 1970 - the year when the system
underwent a major upgrade and expansion. Thus, we confine our sample to all official loans'4
contracted in or after 1970 -- which amount to some 40,000 loans.'5 Because of disbursement
delays, these loans account for the bulk of official loan disbursements only starting in 1975; hence the
aggregates reported below span the period 1975-95.
Lastly, we also exclude from our coverage the 1995 Tequila loans to both Mexico and
Argentina. These special loans carried hard terms that would translate into (sharply) negative grant
elements in our EDA calculations. Since disbursed amounts were substantial (on the order of $26
' Upon founding of the DRS in 1952, the system collected data only on public and publicly guaranteed debt.
In 1970, the system was expanded to cover private non-guaranteed debt.
12 Almost all developing countries are members of the World Bank. Current membership, including
developed countries, is 181.
13 See Appendix A for a complete listing of the countries included in our analysis.
14 However, it is important to note that our database lacks loan-by-loan information on IMF flows. These
consist of standard INF loans, which are typically excluded from conventional aid aggregates, and highly-concessional
loans (such as SAF and ESAF), which are not. Further, the terms attached to the former loan category display
considerably variation over time and across loan types, while those on the latter category do not. This means that an
approximate calculation of loan grant elements based on average (rather than loan-specific) IME terms, which are
available, should be reasonably accurate for the concessional-window loans, and much less so for the non-concessional
ones. In view of these facts, we opted for retaining the concessional-window EMF flows in our calculations, and
discarding the non-concessional window loans.
15 In practice, this choice of starting period becomes almost unavoidable given the very limited availability of
interest rate data for non-dollar currencies prior to 1970 , which would prevent construction of appropriate (liscount
rates even if complete loan data were available.



10
billion), their inclusion would result in significant changes in 1995 EDA measures. We chose not to
include Tequila loans in our coverage because this kind of financial rescue package is atypical in our
sample, both in its financial terms and its volume, and would distort the aid trends we analyze in this
paper. 16
Having defined our loan universe, an important step before embarking on the calculation of
grant elements is to verify the consistency of the DRS loan data with the CRS data underlying the
conventionally-used Net ODA. This is important because we will later compare our EDA with the
OECD's Net ODA, and we need to ensure that divergences between the two aid measures are not
just due to discrepancies between their respective data sources. To verify this, we used data from the
DRS to construct a DRS-based aid aggregate comparable to Net ODA. One problem in doing so is
the stated purpose of official loans: ODA includes only those concessional loans with a development-
related objective, while DRS includes all concessional loans regardless of their purpose. To follow
the ODA definition, we opted for removing from our DRS-based loan data those loans with the
clearest non-developmental purpose -- military and defense-related loans. 17
A second problem has to do with the treatment of debt forgiveness. As noted earlier, the
OECD accounts for debt forgiveness by recording a fictitious grant matched with an equally
fictitious loan repayment. In contrast, the DRS system does neither. Thus, to ensure comparability
between OECD ODA and DRS-based ODA, we deducted debt forgiveness from the grant totals
when adding grant flows to DRS loan flows to construct our DRS-based ODA.18
Figure 3.2 plots the nominal differences between DRS-based ODA constructed in this way
and conventional OECD ODA. In general, the differences are quite small. Mean differences and
mean absolute differences between both measures, computed for each recipient country over the
period 1975-1995, cluster around zero for most countries. The average across countries of the mean
differences is just $ 2.4 million. By contrast, mean absolute differences are much larger -- their cross-
country average is $29 million. This suggests timing mismatches between creditor and debtor
reporting, which cancel out in the average differences but not in the mean absolute differences.
Indeed, mean absolute differences of three-year moving averages are substantially lower. Relative to
the recipient economy's size (in terms of GNP), the mean discrepancy is less than 1 percent of the
recipient's GNP for 125 of the 133 countries in the sample. The mean difference in the total sample
is -.07 percent of GNP, with a standard deviation of .77 percent.'9
16 However, with the onset of financial crises in emerging markets in 1997-98, there appears to be an
institutionalization of this kind of lending , both in volume (rescue and emergency lending packages will approach
$200 billion in the period) and in the hardness of the terms. Future revisions of EDA methodology will need to
incorporate these new developments in some way.
17 This purpose group includes some 350 loans.
"' Since we are comparing ODA totals, we adjusted for ODA debt forgiveness only, rather than for all
forgiveness, as defined by the OECD. The main difference between the two is the forgiveness of two large non-ODA
loans to Egypt in 1991 and 1992.
19 We also examined the correlation between the OECD and the DRS-based ODA measures. The correlation
was computed on the loan component only, since both measures share the same grant component. On a nominal basis,
the sample correlation coefficient is .89, and in three-year moving averages, the sample correlation coefficient is .93.



11
The conclusion from these consistency checks is that despite the differences in lender
coverage and reporting methodology, the World Bank and the OECD's databases are actually quite
similar for most countries in our sample. Having comparable underlying loan data indicates that any
major divergences between our EDA and the OECD's Net ODA should be driven mai:nly by
differences in methodology, which is precisely the focus of our subsequent analysis.
Interest Rate Data
The other key ingredient for the computation of the grant element is a set of appropriate rates
to discount the expected cash flow streams associated with each loan. As discussed in the previous
section, the conventional approach underlying ODA uses an arbitrary rate of 10 percent for all
discounting, a practice that completely ignores market conditions and loan features relevant to the
opportunity costs that discount rates ought to capture. Proper valuation should employ instead
discount rates that reflect the time, currency and maturity characteristics of the relevant cash flow.
Such rates therefore need to vary along all these dimensions.
To this end, we use as discount rates the market yields on government securities of various
maturities denominated in the six major lending currencies -- U.S. dollar, yen, deutsche mark, French
franc, British pound and Swiss franc. Together (by themselves and as part of currency baskets) these
six currencies account for 88.8 percent of the total volume of loan disbursements in our loan sample.
We take the market yields on government securities of a given maturity denomninated in these
currencies as a measure of the risk-free rates of return on debt of the corresponding mraturity
denorninated in each respective currency.
From the 1980s on, the bulk of our interest rate data are collected from Bloomberg's bond
indexes. The indexes track the composite price and yield of a basket of government bonds considered
to be benchmark issues in each currency. For each currency and maturity, we use the annual
averages of all available monthly data. A major shortcoming of this source, however, is the very
limited availability of data prior to 1980 for all currencies other than the US dollar. To complete the
time series, we resorted to other sources - mostly central bank bulletins and the IMF's International
Financial Statistics.20
With the raw annualized rates, we constructed yield curves for each of the six currencies and
each year of the period 1970-1995. Where yields are unavailable for a given maturity, we interpolate
across maturities using the observed rates for each year; hence, the interpolation derives thLe term
structure of interest rates implied by the observed benchmark rates. We experimented with both
linear and logarithmic interpolation. The latter involves fitting a log-linear curve across the observed
rates, while the former involves fitting piece-wise straight lines across the same observations. Figure
3.3 plots the matrix of our interpolated rates for the US dollar.
20 Following the IMF's practice, where rates on T bills or other short term instruments are not available, 3-
month deposit rates are used instead. See Appendix A for a complete listing of available rates and their sources.



12
In the end, we opted for the logarithmically-interpolated curves for several reasons.  First,
linear interpolation is problematic in cases where the rates at the long end of the yield curve are
missing. The reason is that the linear method would essentially imply a flat term structure beyond the
last available short rate, or alternatively, dictate that the tail of the curve adopt the same slope of its
shorter end -- both highly questionable assumptions. The logarithmic approach offered more
flexibility in dealing with this problem. Second, as will become clear later, we utilize the yield curves
to project future short-term interest rates, which serve as the basis of interest service on floating-rate
loans. Since these future short rates are derived from the slope of the yield curve, they take on
extreme values from abrupt changes in slope that occur around observed rates on linearly-
interpolated yield curves.
Lastly, to discount the flows associated with loans denominated in multiple currencies, we
constructed yield curves for various currency baskets -- including the Special Drawing Rights (SDR),
European Currency Unit (ECU), and the lending baskets of major multilateral institutions (IBRD and
1DB). Currency weights in each basket are applied to the discount rates available in our six
currencies to build yield curves for each basket.22  Underlying weights for the baskets were
constructed from publicly-available sources. In particular, the ECU, the SDR, the IiBRD and the
IDB baskets are collected from the respective institutions.23
3.2 Computing Effective Development Assistance
Using the World Bank's DRS loan data and the currency-specific yield curves summarized
above, adjusted grant elements and grant equivalents were computed for all official (i.e., bilateral and
multilateral) non-military loans in the DRS database contracted from 1970 onward. The actual
calculations take into consideration the features of each individual loan as specified in the loan
contract. Thus, they allow for multi-year disbursement periods, commitment fees and one-time
charges, various amortization profiles, variable interest rates and so on. Formally, the general
expression used to compute the grant element of a given loan is:
PV(D   "     P       A,   )V({C  }t )]                                               (3.1)
g                           PPV({D }t= 0)
21 However, all the calculations in the paper were performed with both sets of rates. The differences between
the respective results were significant for some loans, but negligible for the broad aggregates presented below.
22 Historical interest rate data were not available for all underlying currencies in some baskets. In such cases,
we construct the full-basket rates using ouly the available currencies and rescaling their weights to add up to 100
percent.
23 For loans denominated in currencies other than the major six (and the baskets just mentioned), we adopt
the following procedure. We construct generic baskets whose currency composition reflects the general currency
profile of multilateral and bilateral loans. We derive the basket weights each year from bilateral and multilateral
disbursements made in each currency in that year (scaled to exclude currencies with unavailable rates), and use the
weights to aggregate the discount rates for each currency. The resulting basket rates represent the weighted average
opportunity costs of bilateral and multilateral lending made during the year, and the associated discount rate is used
for loans denominated in currencies for which market interest rates are unavailable.



13
where {D} is a sequence of disbursements, {A} is a sequence of principal amortization payments, {I}
is a sequence of interest payments, {C} is a sequence of other charges such as commitment fees, and
PV denotes the present value. The time indices 0, 6 y and T respectively denote the date of first
disbursement, the number of periods between first and last disbursement, the grace period, and the
maturity of the loan. To keep the calculations manageable, annual periods were used; this obviously
represents an approximation in the (relatively frequent) case of loans specifying semi-annual service
payments. Notice that the term in square brackets in the numerator of (3.1) above is just the present
value of debt service obligations - i.e., the debt equivalent E introduced in the preceding section.
Present values of disbursement and service payment streams are computed using the discount
factors derived form the currency-specific yield curves just described. Hence, for any sequence {X},
T  T    xt
PV({X,}t0)=O (            )                                             (3.2)
where Ro, is the (annualized) interest rate on t-year maturity instruments prevailing at time 0; it is
also currency-specific, an issue which for notational simplicity is ignored here. Hence, in accordance
with the ex-ante perspective adopted in this analysis, the interest rates used for discounting are those
prevailing at the time of loan agreement, rather than those effectively observed ex-post 4
Some remarks on the construction of the disbursement and loan service streams mlay be
useful. As with discount rates, the different components of loan service were constructed on the basis
of the ex-ante contractual information -- and not from the interest and amortization payments
actually made ex-post by the borrower. Regarding the amortization schedule, the arrangement: most
frequently found in the data is that of equal repayments following a grace period. However, a wide
variety of amortization profiles exist in the data, ranging from annuity-based amortization schedules
to "balloon" principal repayments (whereby the loan is amortized in full in one single payment at
maturity).
In turn, interest charges are typically accrued at a contractually-determined rate on the loan's
outstanding balance. While a time-invariant interest rate is by far the most common arrangemernt, in a
large number of cases the loan contract sets interest rates that change over the life of the loan, either
in a pre-specified manner or following a reference rate such as LIBOR. For the calculations, interest
charges were generally computed as
it      ( it<  )(D,-A)j   for O<t< T                                                (3.3)
24 For simplicity, in the text it is implicitly assumed that the dates of loan agreement and first disbursement
are the same. While this is indeed correct for the majority of loans, it is not invariably the case -- the first disbursement
often takes place one or more years after loan agreement. The discount rates used in the calculations in such cases
were those observed at the time of agreement. Experiments were also made using instead the rates prevailing at the
time of first disbursement. This of course affected the estimated grant elements of specific loans, but had a virtually
negligible impact on the aggregate annual grant equivalents accruing to each recipient country.



14
where it is the interest rate contractually specified for period t, and the term in large brackets is the
loan's outstanding principal as of the end of period t-1. The determination if it for the case of loans
with variable interest rates deserves comment. For those loans with interest rate linked to some
short-term market rate (e.g., 6-month LIBOR), a forecast of the latter was constructed using the
anticipated one-period interest rate derived from the yield curve described earlier, adjusted for a
margin calibrated from the historical data. Thus, the contractual interest rate was projected as:
t =           I     (I ~~~~+ Ro0t)t+1                                  34
it =bot +s  where   bo0 =   1--(I    t R -',    j(1+Y)                        (3 4)
where bo, is the projected short-term market rate (e.g., LIBOR); s is a contractual spread, and the
(proportional) margin ,u was computed through a univariate regression of the short-term market rate
25
on the one-year rate from the yield curve.
A different type of variable interest-rate arrangement concerns the loans granted by some
multilateral institutions since the early 1980s, whose interest rates are determined as a spread over
the lender's average (as opposed to marginal) cost of borrowing. This can be viewed as a weighted
sum of current and lagged interest rates on various currencies, and therefore follows market trends
with a substantial lag. Projection into the future of the cost of funds is further complicated by the fact
that the weights of the various currencies in the lender's borrowing basket can, and do, change over
time (albeit gradually). To simplify the calculations, however, anticipated changes in the currency
composition of the borrowing basket were ruled out. Thus, the cost of funds relevant to each loan
was determined by aggregating the yield curves of the various currencies making up the lender's
borrowing basket, using as weights their shares in the basket on the year of the loan's first
disbursement.
Finally, loans from major multilateral institutions typically carry commitment charges on the
undisbursed principal, that accrue periodically (commonly starting 60 days after commitment).
Hence, commnitment fees were computed as:
Ct = c    D, -  D,) = c ,Dt , 1<t<e5                                            (3.5)
Z-0    -r=O          -r= t
The term in brackets captures the undisbursed loan amount as of the end of period t-1, and c is the
contractually-fixed rate at which commitment fees accrue.
In contrast with service payments, the contractual information usually does not specify a
disbursement schedule. Thus, actual disbursements had to be used instead. This means that in our
discussion we use the term "loan principal" to refer to the total amount effectively disbursed, which
might differ from the contractually-agreed amount. Note also that it is not uncommon for actual
disbursements to lag behind the originally planned schedule, and this could lead to an
25 As a robustness check, all calculations were repeated using an additive, rather than multiplicative, margin
pt. The impact on the estimated grant elements was negligible.



15
underestimation of the grant element relative to its true ex-ante value (i.e., as viewed at the time of
agreement).26
Having obtained grant elements as just described, the grant equivalent of each disbursement
is computed by imputing to the disbursement amount the loan's overall grant element:
Gt = g g D;  0 < t� '                                                              (3.6)
It follows from (3.1) above that, in present value terms, PV({GJ) = PV(DJ) - [PV({Ad) -+ PV(It})
+ PV({CtQ)] . In words, the loan's overall grant equivalent (given by the discounted sum of the grant
equivalents of the different disbursements) equals the present value of the disbursement stream minus
the present value of the loan's debt service stream.
The above expressions are quite general and allow for a broad variety of contractual
arrangements regarding amortization schedules, interest payments and other charges. In general,
however, they do not yield a closed-form expression for the grant element, which is only available
under additional assumptions. Specifically, Appendix B shows that if (i) the loan is disbursecd in full
at the time of commitment; (ii) the interest rate is fixed at i; (iii) the amortization profile involves
equal payments after a grace period y; (iv) the discount rate R is constant; and (v) commitment fees
and other charges are ignored, then (3.1) simplifies to:
1         1
gz2           I               1+)                                   (3.7)
R1z)            R. (T- Y                                      37
Conventional grant element calculations are based on (3.7) setting R equal to 10 percent.
In reality, of course, the assumptions just listed do not hold, and the "approximate" grant
element obtained from (3.7) could differ greatly from the correct one obtained from (3.1). As will be
shown below, however, in practice the main source of discrepancy turns out to be the assumption of
a constant discount rate; the other assumptions have a comparatively more modest impact on the
grant element calculation.
4     RE-ASSESSING TRENDS IN AI) FLOWS
The equations outlined in the preceding section were numerically evaluated for each one of a
total of over 40,000 loans27 in the DRS database (of which some 24,000 were from bilateral lenders,
26 In addition, comprehensive disbursement infonnation is available only in US dollars. This might introduce
another distortion in the computed grant component of loans denomninated in other currencies: even if their
disbursements proceed on schedule, the time profile of disbursements expressed in dollars could be altered ex-post due
to unanticipated exchange rate changes during the disbursement period.



16
and the rest from multilateral lenders).28 All results are aggregated and summarized next, focusing
first on the broad aggregates and turning then to their breakdown by recipient and donor.
4.1    Measuring grant elements
The improved methodology just described has a major impact on the estimated aid content of
official loans. Figure 4.1 plots the average grant element of official loans as obtained from three
different approaches. First, the conventional (i.e., the OECD's) approach, using the 10-percent
discount rate and imposing restrictive assumptions on the disbursement, interest and amortization
schedules (as needed to arrive at the simplified expression (3.7) described earlier). Second, the
adjusted methodology proposed above, imposing no restrictions on loans' disbursement, interest
and amortization schedules, and using the year-, currency- and maturity-specific discount rates
constructed in the previous section. Third, for the purpose of comparison, the graph also presents an
intermediate approach allowing for unrestricted disbursement and service schedules but still using the
10-percent discount rate. In each case, the annual average of the grant elements of individual loans is
constructed using as weights the share of each loan in the year's total disbursements
The four panels of Figure 4.1 respectively show the average grant element of all official
loans, all multilateral loans, multilateral concessional-window loans, and bilateral loans. They bring
out three important facts. First, the conventional method leads to a systematic overestimation of the
average grant element, as shown by the fact that the adjusted grant element is almost invariably lower
than the conventional one; the exception are the early 1980s in the case of bilateral and multilateral
concessional loans. Second, the gap between the grant element implied by the two approaches shows
substantial variation over time: it declined during the late 1970s, became almost negligible in the
early 1980s, and has widened dramatically since 1985. In fact, after that date the conventionally-
measured grant element has shown an upward trend, while the adjusted grant element has plunged --
particularly in the case of bilateral loans, whose concessionality fell to historical lows in the 1990s.
Third, the overestimation of the grant element by the conventional method appears to be due to its
assumption of a 10-percent discount rate. As the figures show, if all other restrictive assumptions
underlying the conventional method are removed, but the 10 percent discount rate is retained, the
resulting grant element is closer to the adjusted grant element, but the improvement is indeed rather
minimal.
The main lesson from these facts is that the use of an arbitrary 10-percent discount rate
clearly distorts the conventionally-measured grant elements. In the sample, there is a positive bias
that varies considerably over time. A natural question is: how does this arbitrary 10-percent rate
differ from the "right" one? Conceptually, as argued above, the discount rate should be not only
time-varying, but also currency- and maturity-specific -- precisely the approach underlying the
calculations of the adjusted grant elements above. For the purpose of illustration, however, it is
possible to construct an ideal "average" discount rate, which is the one implied by the loan-by-loan
27 Of this total, some 50 loans had to be discarded because their contractual information regarding interest,
amortization or both was incomplete and did not allow calculation of the corresponding grant element.
28 For concessional IMF loans, the loan-by-loan information is not available, as noted above. Hence we used a
simplified approach: we plugged into (3.7) the average terms of concessional IMIF disbursements in each given year,
using as disbursement figure the corresponding annual total, and allowing the discount rate to vary over time.



17
calculations of the adjusted grant elements. One simple way to do this is to solve the converntional
formula (3.7) for the discount rate that would yield the average adjusted grant element just presented
in the above graphs, for an ideal loan whose features (grace, maturity and interest rate) match those
of the average loan in the sample. Notice that, by doing this, all the other subtleties found in the
contractual loan information (multi-year disbursements, time-varying interest rates, irregular
amortization schedules, various loan charges...) are ignored.
Following this procedure for every sample year, a time-series of implicit discount rates can be
constructed.29 The results are presented in Figure 4.2, which, for the purpose of illustratiorn, also
plots the interest rate on 10-year U.S. government securities. The figure shows that the irnplicit
discount rate comes close to the horizontal 10-percent line only during the first half of the 1980s
(and the former even exceeds the latter in 1982). In the remaining years, the implicit rate is well
below the 10 percent benchmark, and especially so in the early 1970s and the 1990s. This, of course,
is just the underlying reason for the overstatement of the conventionally-measured grant elements in
those years, as described above. Finally, it is important to note that the time pattern of the irnplicit
discount rate is governed by the changing conditions in world financial markets. As the figure also
shows, the swings of the former track those in U.S. long-term interest rate, although, with
considerably reduced magnitude -- reflecting the (increasing) role of non-dollar-denominated official
loans in overall lending flows.
4.2    Reassessing time trends in aid
So far the discussion has focused on the concessionality of official loans. What are the
implications for measured aid flows? To answer this question, the first step is to assess the aid
volume implicit in official loans according to their adjusted grant elements; we shall refer to such aid
volume as the effective development assistance of official loans. This can be constructed by summing
their adjusted grant equivalents, computed according to equation (3.6) above. The results are shown
in Figure 4.3, which also presents the grant equivalents derived from the conventional method as well
as those arising from the intermediate method that uses the general expression (3.1) with a 10-
percent discount rate. To facilitate comparability across years, the resulting totals have been
converted to 1995 U.S. dollars.
The graph again reveals a stark contrast between the conventional and adjusted figures after
1985: while the conventional grant equivalent has risen steadily to a record-high 20 billion by t]he end
of the sample period, the adjusted grant equivalent has changed little over the last decade, and at
present amounts to some 8 billion. In other words, in recent years the conventional grant equaivalent
overestimates the aid content of official loans by 150 percent ! Finally, like with the grant elements,
the chart again confirms the relatively secondary role of factors other than the discount rate in
explaining the discrepancy between conventional and adjusted grant equivalents.
29 It is worth noting that the weights used in each year to construct the average loan features (i.e., the average
grant element, grace, maturity and interest rate) for this calculation from those of the individual loans are given by the
face value of each loan committed in the year in question. This follows logically from the use in the calculations of
equation (3.7), that assumes full loan disbursement at the time of comnmitment. However, it differs from the weighting
scheme used in Figure 4.1 which, as noted in the text, is disbursement-based rather than commitment-based.



18
Conventionally, however, the aid content of official loans is measured by the net
disbursement of ODA loans. Figure 4.4 compares such conventional measure with the EDA of
official loans just described.30 Unlike Net ODA, EDA is based on gross flows, and hence it is not
surprising that the levels of the two measures should display persistent differences. However, the
pattern is indeed very similar to those of the preceding figures: ODA overstates systematically the aid
content of official loans and, except for the early 1980s, the overstatement is quantitatively very
large. Moreover, because of the diverging pattern of both measures since 1985 -- rising Net ODA
loans and declining EDA -- the former at present exceeds the latter by around 100 percent.
The above discussion has been confined to official loans. What is its relevance for total aid?
Adding official grants to the adjusted grant equivalent of official loans, one can construct a measure
of total EDA, analogous to ODA. Since grants have, by definition, a 100 percent grant equivalent,
any discrepancy between ODA and EDA is just due to their different measurement of the aid content
of official loans. As just shown, such discrepancy is quite considerable; relative to total ODA or
EDA, however, it is somewhat lessened by the large (and increasing) magnitude of official grants
documented earlier. This is illustrated by Figure 4.5, which compares Net ODA and total EDA over
the last two decades. As with official loans, it is apparent that since 1985 Net ODA considerably
overstates aid flows -- by as much as 25-30 percent in recent years.
4.3   EDA vs. ODA by recipient
Because EDA is based on a consistent measure of the concessionality of financial flows, it
allows meaningful comparisons of aid flows across donors and recipients. This is not the case with
comparisons of ODA flows, however, which can be misleading for a number of reasons: 1) because
of the potentially large differences in the grant/loan composition of flows from different donors or to
different recipients, 2) because of the potentially different degree of concessionality of the loans
involved in those flows, or 3) because of differences in the time profile of lending flows impacting the
netting out of amortization in ODA calculations.
Indeed, there are systematic differences in the concessionality of official flows across both
recipients and donors. Figure 4.6 focuses on the recipient side, grouping aid-receiving countries into
six geographical regions, according to the World Bank's definition. The figure presents the average
grant element of official loans to each of the regions over the period 1990-95, computed according
to both the conventional method and the adjusted method introduced earlier.
As the figure shows, both methods reveal a considerable variation of grant elements across
regions. Conventional grant elements range from 18 percent in Europe and Central Asia to 49
percent in Sub-Saharan Africa. In turn, adjusted grant elements range from virtually zero in Europe
and Central Asia, as well as Latin America (for practical purposes, official loans to these two regions
are not concessional), to 47 percent in Sub-Saharan Africa. Interestingly, the regional rankings by
grant element are similar under both measures. However, the extent of divergence between them
varies considerably across regions -- from just over 2 percentage points in Sub-Saharan Africa, to 12
in South Asia and some 20 percentage points in the other developing regions.
30 As noted earlier, the loan database used here does not allow construction of a satisfactory DRS-based ODA
measure prior to 1975, and hence the chart begins in that year.



19
How do these diverging assessments of loan concessionality across recipient regions impact
on measured regional aid flows? Figure 4.7 presents the ratio of Net ODA to total EDA in recent
years for each of the six developing regions above; the underlying flows are expressed in 1995 US
dollars. The figure illustrates eloquently how the degree of distortion embodied in conventional aid
measures can vary across aid recipients. For East Asia and the Pacific, and Latin America and the
Caribbean, the contrast between both Net ODA and total EDA is dramatic: during 1990-95, Net
ODA to these two regions exceeded effective assistance by around 60 percent. At the other end, for
Sub-Saharan Africa and South Asia (who receive mostly grants), the divergence is much smaller -
around 10 percent in both cases.
4.4   EDA vs. ODA by donor
Like with recipients, EDA also permits consistent comparisons of aid flows across donors.
By way of example, we conclude this section with a brief look at the comparative performance of
bilateral donors.
Conceptually, it should be clear from the earlier discussion that, since conventional aid
measures typically overstate the grant elements of concessional loans, such measures will tend to
exaggerate the contribution of donors whose disbursements take the form mostly of loans. Further,
the overstatement will be more severe for donors lending in low-yield currencies, because the interest
rate on their loans will tend to be below the arbitrary 10 percent discount rate used in conventional
grant element calculations. The opposite will happen with donors giving mostly grants, or
concessional loans in high-yield currencies - conventional measures will tend to overstate their
contribution to a more limited extent.
Figure 4.8 ranks DAC donors according to their respective volume of aid relative to GNP,
using both Net ODA and EDA, for the periods 1981-85, 1986-90 and 1991-95. Two qualifications
are in order. First, the figures underlying the rankings only include donors' direct aid to developing
countries and therefore exclude their contributions to multilateral institutions. Second, because we
lack detailed debt forgiveness data by donor prior to 1991, the figures for 1989 and 1990 underlying
the rankings are not adjusted for ODA debt forgiveness - i.e., donors' grant totals (and hence total
aid) in those years mnight include forgiveness of ODA debt.3"
With these caveats, the figure reveals interesting contrasts between the ODA and EDA--based
donor rankings. Most remarkable is the fact that a few donors -- notably Japan, but also France and
Spain (the latter after 1981-1985) -- rank systematically higher under ODA than under EDA. This
likely reflects the high reliance of these donors on loans rather than grants and, at least in the case of
Japan, also the fact that yen interest rates have remained well below the 10 percent benchunark.
Correspondingly, other donors find themselves in the opposite situation, ranking higher under EDA
than ODA. This is the case of Canada (especially in 1986-90) and, more recently, the Netherlands as
well.
31 Prior to 1989 ODA debt forgiveness is negligible.



20
5. CONCLUDING REMARKS
Conventional aid aggregates such as the OECD's Net ODA provide a distorted measure of
true aid flows. The reason for this distortion is the weak methodological foundation of such
aggregates. They lump together the net increase in loans, which entail future interest and repayment
obligations, with grants, that do not; further, they include certain loans at full face value and totally
exclude others; finally, the selection of which loans to include is based on a calculation of their grant
elements that, among other simplifications, makes use of an arbitrary discount rate set at 10 percent.
This paper has proposed a new approach that corrects most of the major methodological
shortcomings underlying conventional aid measures. The approach is based on the use of the overall
grant equivalent of official disbursements as the key measure of aid, resulting in a new aggregate that
we label Effective Development Assistance (EDA). The use of grant equivalents as the basic aid
measure has the virtue of allowing comparability between aid flows implicit in grants as well as loans
with varying degrees of concessionality. Most importantly, EDA is based on adjusted grant
equivalents constructed using discount rates sensitive to market conditions along several dimensions
- timing, currency and maturity of the loans --, unlike the conventional grant elements based on the
arbitrary 10 percent discount rate.
The paper has implemented empirically this new approach using detailed data on some
40,000 official loans from the World Bank's DRS database -- virtually all the official loans during the
period 1975-1995 to thel33 developing countries included in the DRS system. The numerical results,
which are available to all interested researchers, underscore several important facts. First, the
conventional approach to grant element calculation has led to a systematic overestimation of the
concessionality of official loans. Second, this overestimation has increased significantly since the mid-
1980s. After that date, conventionally-computed grant elements have shown a rising trend, while the
conceptually superior adjusted grant elements show, if anything, the opposite time pattern. Grant
equivalents obtained by both methods reveal the same contrasting pattern. Third, as a consequence
of these results, Net ODA increasingly overstates the true aid content of official flows, as more
accurately measured by EDA - even though the large divergence between both approaches in their
assessment of the aid content of official loans is somewhat muted by the rising importance of grants
over loans in total official flows.
Since conventional aid measures typically overstate the grant elements of concessional loans,
they tend to exaggerate comparatively more the aid flows to those recipients getting mostly loans.
This is so especially if the loans are given in low-yield currencies, because their interest rate will tend
to be further below the arbitrary 10 percent discount rate used in conventional grant element
calculations. The same happens with donors: the aid contribution of those giving mostly grants (and
loans in high-yield currencies) will be understated relative to the others. The paper presents
numerical findings confirming these intuitive facts. This in turn implies that conventional aid
measures can be very misleading for purposes of comparison across recipients and/or donors.
While we believe that these results are highly relevant to the current policy debate on aid,
some caveats remain. First, our approach to the measurement of aid adopts an ex-ante perspective in
order to capture donors' effort, and it would be interesting to compare our results with those



21
obtained from an ex-post approach - based on actual rather than contractual debt service flows.
Second, and perhaps more important, while our analysis addresses most of the key limitations of
conventional aid measures, it still does not incorporate default risk in the calculations of loan
concessionality. Doing so would basically require construction of the risk spreads applicable to each
borrower each year for each creditor class. However, in the absence of markets for official debt to
price directly such risks, this would be a rather problematic task, well beyond the scope of this paper.



22
Appendix A
This appendix lists the countries included in our analysis and the sources of our raw discount rate data.
I. INCLUDED RECIPIENT COUNTRIES
Recipient countries included in our analysis are listed by region in the box below:
East Asia and Pacific   Europe and Central   Latin America and        Middle East and        Sub-Saharan Africa
Asia                  the Caribbean            North Africa
Cambodia               Albania               Argentina                Algeria                Angola
China                  Armenia                Barbados                Egypt                  Benin
Fiji                   Azerbaijan             Belize                  Iran                   Botswana
Indonesia              Belarus                Bolivia                 Jordan                 Burkina Faso
Korea                  Bulgaria               Brazil                  Lebanon                Burundi
Laos                   Czech Republic         Chile                   Malta                  Cameroon
Malaysia               Estonia                Colombia                Morocco                Cape Verde
Mongolia               Georgia                Costa Rica              Oman                   Central African Republic
Myanmar                Hungary                Dominica                Syrian Arab Republic    Chad
Papua New Guinea       Kazakhstan             Dominican Republic      Tunisia                Comoroas
Philippines            Kyrgyz Republic        Ecuador                 Yemen, Republic of     Congo
Solomon Islands        Latvia                 El Salvador                                    Cote d'lvoire
Thailand               Lithuania              Grenada                                        Djibouti
Tonga                  Moldova                Guatemala               South Asia             Equatorial Guinea
Vanuatu                Poland                 Guyana                                          Eritrea
Viet Nam               Romania                Haiti                   Bangladesh              Ethiopia
Western Samoa          Russian Federation     Honduras                Bhutan                 Gabon
Slovak Republic       Jamaica                 India                   Gambia, The
Tajikistan            Mexico                   Maldives               Ghana
Turkey                Nicaragua                Nepal                  Guinea
Turkmenistan          Panama                   Pakistan               Guinea-Bissau
Ukraine               Paraguay                Sri Lanka               Kenya
Uzbekistan            Peru                                            Lesotho
St. Kitts and Nevis                             Liberia
St. Lucia                                      Madagascar
St. Vincent and the Grenadines                 Malawi
Trinidad and Tobago                            Mali
Uruguay                                        Mauritania
Venezuela                                      Mauritius
Mozambique
Niger
Nigeria
Rvvanda
Sao Tome and Principe
Senegal
Seychelles
Sierra Leone
Somalia
Sudan
Swaziland
Tanzania
Togo
Uganda
Total Included: 133                                                                           Congo, Dem. Rep (Zaire)
Zambia
Zimbabwe



23
II. SOURCES OF DISCOUNT RATE DATA32
Bloomberg:    The bulk of our data after the mid-80s is collected from Bloomberg. Bloomberg maintains
indexes on benchmark treasuries in all six major lending currencies. These Bloomberg "generic" treasury
indexes track the composite price and yield of a basket of bonds considered to be benchmark issues in each
market. Price of a "generic" bond, note or bill of a specified maturity is an average of a number of market
maker bid-side quotes updated hourly, several times throughout the day, or at closing, depending on the
market.
For consistency, all monthly data available in Blooomberg are collected and used. Annual rates are
calculated by taking simple arithmetic averages of monthly yields. The resulting data is comprehensive for the
90s for all currencies. A major shortcoming of these Bloomberg rates is the lack of historical data for all
currencies except the US dollar. To complete the series, we collected data from other sources - mostly central
bank bulletins and the IMF's International Financial Statistics.
Listed below are sources from which we collected historical yields and interest rates. Following the
IMF's practice, where rates on T bills or other short term instruments are not available, 3 month deposit rates
are used instead. Note that yields on Bloomberg generic government bills, notes or bonds are simply referred
to as "Bloomberg."
US RATES
3 month T bills      IM1, International Financial Statistics, discount on new issues of three month T bills,
annual data are averaged weekly rates.
6 month T bills      1983-1996: Bloomberg, generic 6 month T bills.
2 year treasury      1977-1996:Bloomberg.
3 year treasury      1970-1996: Bloomberg.
5 year treasury      1970-1996: Bloomberg.
10 year treasury     1970-1996: Bloomberg.
30 year treasury     1970-1979: Ibbotson Associates.
1980-1996: Bloomberg.
UK RATES
3 month T bills      1970-1996: IMF, International Financial Statistics (as reported by Bank of England,
Quarterly Bulletin). Rates represent tender rates at which 91 day T bills are issued,
calculated from amount of discount. Monthly data are weighted averages of Friday clata.
Annual figures are averaged monthly rates.
2 year UK Gilt       1992-1996: Bloomberg.
3 year UK Gilt       1992-1996:Bloomberg.
4 year UK Gilt       1992-1996:Bloomberg generic 4 year.
5 year UK Gilt       1970 to 1991: Annual Abstract of Statistics, Central Statistical Office (CSO) of Great
Britain. Rates are par (gross redemption) yields calculated from yield-maturity curves
fitted mathematically. Until 1979, annual rates are averages of Wednesday yields. From
1980 on, the averages are of all observations (usually 3 per week); and from 1982 forward,
the figures are averages of working days. See Bank of England Quarterly Bulletini for
descriptions of actual methods of calculation. Note that all CSO rates are calculated in the
same fashion for the various maturities.
1992 to 1996: Bloomberg.
32 Discount rate data is available upon request.



24
7 year UK Gilt       1992-1996:Bloomberg.
10 year UK Gilt     1970 to 1991: Annual Abstract of Statistics, Central Statistical Office of Gieat Britain.
1992 to 1996: Bloomberg.
15 year UK Gilt     1994-1995: Bloomberg.
20 year UK Gilt      1970 to 1991: Annual Abstract of Statistics, Central Statistical Office of Great Britain.
1992 to 1996: Bloomberg generic 20 year.
30 year UK Gilt      1996: Bloomberg.
JAPANESE RATES
3 month deposit rate  IMF, International Financial Statistics. From June 1992, rates are average interest rate for
the last week of the month on unregulated three-month time deposits. Prior to that, rates
are end of month guideline rates on three month time-deposits set by the Bank of Japan.
Annual rates are averaged monthly rates.
1 year              1995-1996: Bloomberg.
2 year               1979 to 1989: Economic Statistics Monthly, Research and Statistics Department, Bank of
Japan (BOJ). Rates are yields to subscribers of interest-bearing govermment bonds of the
specified maturities. Data is available for monthly frequency. Annual figures are
averaged monthly rates.
1990 to 1996: Bloomberg.
3 year               1978 to 1988: Economic Statistics Monthly, BOJ.
1989 to 1996: Bloomberg.
4 year               1980 to 1987: Economic Statistics Monthly, BOJ.
1988 to 1996: Bloomberg.
5 year               1988-1996: Bloomberg.
6 year               1994 to 1995: Economic Statistics Monthly, BOJ.
1996: Bloomberg.
7 year               1970 to 1986: Economic Statistics Monthly, BOJ.
1987 to 1996: Bloomberg.
10 year              1970 to 1986: IMF, G7 database. Data collected on a monthly basis. Before (Nov. '84)
end-of-month; thereafter, average of over-the-counter 10-year govermnent bonds, with
longest residual maturity; from January 6, 1992, the yield refers to benchmark 10-year
government bond. Annual figures are averaged monthly rates.
1987 to 1996: Bloomberg.
15 year              1992-1996: Bloomberg.
20 year              1986 to 1991: Economic Statistics Monthly, BOJ.
1992 to 1996: Bloomberg.
GERMAN RATES
3 month deposit rate   1970-1996: IMF, International Financial Statistics (IFS).  Monthly rate represent
arithmetic average of daily quotations reported by Banks for three month interbank deposit
rates.
1 year T bill        1975-1996: IMF, IFS, rate on 12 month Federal debt register claims.
2 year               1977 to 1989: Statistische Beihefte ze den Monatsberichten der Deutschen Bundesbank,
or Kapitalmarkt-statistik after 1990, Deutsche Bundesbank (Statistical Supplements to the
Monthly Reports of the Deutsche Bundesbank, or Capital Market Statistics after 1990).
Rates are yields on fully taxed public bonds outstanding with remaining maturity of over
one up to two years. Annual rates are averaged monthly rates. Rates from this source
reported below are referred to as Deutsche Bundesbank.
1990 to 1996: Bloomberg.
3 year               1977 to 1989: Deutsche Bundesbank, remaining maturity of over 2 up to 3 years.
1990 to 1996: Bloomberg.



25
4 year              1977 to 1989: Deutsche Bundesbank remaining maturity of over 3 up to 4 years.
1990 to 1996: Bloomberg.
5 year              1977 to 1989: Deutsche Bundesbank remaining maturity of over 4 up to 5 years.
1990 to 1996: Bloomberg.
6 year              1977 to 1992: Deutsche Bundesbank remaining maturity of over 5 up to 6 years.
1993 to 1996: Bloomberg.
7 year               1977 to 1989: Deutsche Bundesbank, remaining maturity of over 6 up to 7 years.
1990 to 1996: Bloomberg.
8 year              1977 to 1992: Deutsche Bundesbank, remaining maturity of over 7 up to 8 years.
1993 to 1996: Bloomberg.
9 year              1977 to 1992: Deutsche Bundesbank remaining maturity of over 8 up to 9 years.
1993 to 1996: Bloomberg.
10 year             1970 to 1989: IiMF, G7 Database. Yield on government bonds with maturities of 9-10
years.
30 year             1994-1996: Bloomberg.
FRENCH RATES
3 month T bill      1970-1996: IMF, IFS, 3 month T bill rate.
2 year              1990-1996: Bloomberg.
3 year              1990-1996: Bloomberg.
4 year              1990-1996: Bloomberg.
5 year              1990-1996: Bloomberg.
7 year              1990-1996: Bloomberg.
10 year             1970 to 1989: IMF, G7 Database. Rates are long-term (7-10 yr.) govt. bond yields
(Emprunts d'Etat a long terme TME).
1990 to 1996: Bloomberg.
20 year             1990-1996: Bloomberg.
30 year             1990-1996: Bloomberg.
SWISS RATES
3 month deposit rate   1973-1996: Swiss National Bank Monthly Bulletin (Banque Nationale Suisse, Bulletin
Mensuel). 3 month deposit rates offered by large banks. Up to June 1989, monthly rates
represent averaged daily values. From July 1989, monthly rates are end of month rates.
Annual rates are averaged monthly rates.
2 year              1994-1995: Bloomberg
3 year              1994-1996: Bloomberg
5 year              1990-1996: Bloomberg
7 year               1970 to 1984: JvIF, IFS, weighted average yield to maturity (if below par) of ten
government bonds with at least five years to maturity, callable. Annual yields are
computed from daily averages of prices.
1985 to 1993: Swiss National Bank Monthly Bulletin, Federal government obligations of
more than 7 years maturity.
1994 to 1996: Bloomberg.
9 year              1996: Bloomberg.
10 year             1990-1996: Bloomberg.



26
Appendix B
This appendix shows how the methodological approach used in the paper to compute the
adjusted grant elements contains as a special case the conventionally-used grant element formula.
The starting point is provided by equations (3.1)-(3.3) in the text, to which the following
simplifying assumptions are added:
Assumption 1: The loan is disbursed in full at the time of commitment
Assumption 2: The principal is repaid in T-y equal amortization payments, starting in year y+l and
ending in year T (where T > 7 2 0)
Assumption 3: The interest rate is fixed at i throughout the life of the loan
Assumption 4: The discount rate is fixed at R
Assumption 5: The loan involves no charges other than interest and amortization
The first assumption implies that the present value of the disbursement stream is just PV(PtD) =
Do, where t=O is the date of commitment. In turn, Assumption 2 implies that amortization payments
are of the form
-   D= (T � )fory<t<T                              (A.1)
(T - y)
and zero otherwise. Hence, using Assumption 4, the present value of the amortization stream equals:
PV({A~ })  [T~Yr] tL+, (1+R)t - T-r [R   (I+ R)r  (1 +R)TI               (A.2)
Using (A. 1) and Assumption 3, the sequence of interest charges can be constructed as follows.
First, the loan's outstanding principal, equal to the cumulative sum of disbursements minus the
cumulative sum of amortization payments, is given by:
Do                 forO<t<y;
Do- (t-Y)T   l  fory<t<T                                  (A.3)
and interest charges therefore follow the path
I, = i*Do                                  forO<t<y+1
=i D   T-t+1]                           fory+l<t<T                (A.4)
Thus, the present value of the stream of interest payments is:



27
_____      i.Do.T-t+lI
PV[{11}]    E (l1�R) +  R)'  t=2    (1 + R)t              (A.)
where the two summation signs refer to the interest charges incurred during and after the grace
period respectively (the second sum only applies if T - r > 1, i.e., whenever the loan is not
amortized in full in a single payment). Tedious manipulations permit simplifying this expression to:
PVR   -  i*0 ri+                                          (A. 6)
R     T-r R(I+R)   (I+R)y
Replacing (A.2) and (A.6) into the grant element equation (3.1) in the text,
rDo    I      I                        Do  I  I    I           1)X
Do-  TI- r  R (1�+1R)r  (1 +R)T          R       T-y R  (1+R)T  (1+RY]]]
Do
This can be easily simplified to:
(1-   R1+Y)  (l + R)T
g   (I   R      R-(T-yr)
which is equation (3.7) in the text. This simplified expression allows some useful insights. First, the
sign of the grant element depends only on the relation between the interest rate and the discount rate.
In particular, if the discount rate equals the interest rate, then the grant element is zero regardless of
loan grace and maturity (notice, however, that this would not be the case if the loan entailed any
service fees in addition to interest charges). Second, the lower the interest rate i for a given discount
rate R, the larger the grant element. Third, for a given configuration of interest and discount- rates,
the absolute value of the grant element rises with the grace and maturity periods, and reaches a
maximum (equal to 1 - i/R) when both grace and maturity approach infinity -- i.e., in the case of a
consol.



28
References
Asti Demirguc-Kunt and Eduardo Fernandez-Arias, August 1992, "Burden-sharing among Official
and Private Creditors", World Bank PR WP #943.
Stijn Claessens, Ishac Diwan, and Eduardo Fernandez-Arias, October 1992, "Recent Experience with
Commercial Bank Debt Reduction", World Bank PR WP #995.
The World Bank, Global Development Finance (formerly World Debt Tables), various issues.



Figure 3.1 Official Development Assistance by Type
1995 $US Billion                                 1975-1995
70
60                        ___
50   -_                                                            _   _ _ _ _
10
10OD Lon
o
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
Note: Total for 133 countries in the sample.



Figure 3.2 DRS-Based ODA versus OECD ODA
US$ millions
800 -
600         _     -                                                         Average Difference
o Mean Absolute Difference
400 -U111 China
U India Korea
200  -                         _      ;                                              _
U Egypt   ��lndonesia
%        m                                  E Poland
0  Egypt                                                                       - 0
-200 -                                                 ___   __
O China
-600 -__
-800
Note: DRS-based ODA is official development assistance constructed with loan-by-loan data from the World BanKs Debtor Reporting System.



Figure 3.3 US Dollar Yield Curves
12.0~~~~~~~~~~~~~~~~2
10.~~~~~~~~~~~~(
400)



Figure 4.1 Average Grant Element: Adjusted vs Conventional
a) Average Grant Element of All Official Loans
45% -
25%
--oAdiustedd
15%-         Conventional
_10%
5%-,,,,. 
U)    r-    r)      s      c) 0          r-    CD     o       )    Uc
0     0-     0      0)     U      0)    0)    0)       0     0      0)
C)    U)    C)    C)    C)~   co    co    co    C)    U      CD
b) Average Grant Element of Multilateral Loans
45% -
35% -                                                              �
25% -
- Conventional
s% E  ,1                     
U)  N    C)            !C)    U)     N      C     -      ()
N     N      N     U)    U) Uco co c           00                  C)
C) Average Grant Element of Multilateral
Concessional Window Loans
90% 
85%       --
80% -
7_Adjusted
70% -                   Conventional
*10%
65% -                       I                 
U)    N      c)     -      (      U)     N      C)     _     rQ     U
N     N      N      0)    0)    0)    0)        0       )    U      C)
D                   0      0)    CD    0)    0)    0)         0     0
d) Average Grant Element of Bilateral Loans
45%-
35%
25% -
z-Adjusted
15% -Conventional                          _      _
.    -10%                               -



Figure 4.2 Conventional Discount Rate and Adjusted Discount Rates
15%- 
15%- -_________                  _            _______                       _ _ _           _ _ _            ___            _ ___  _
14%                                                                                            Implicit Market-Based Discount Rate
13% -                       _                    _
/    \    A        ~~~~USD IO year yield
12%-                              ___           _                                       __            _    _ _ _
11%                                -_.__                                               _             _           _,
10%
9%     - _,                                                      _      _       _
8%                -                                                      __________
7%  -                                                                               -       
5%   -         I  .TI
O   (-  cNI  C'   1Q          (0   1-        C)  d 0                        L)  (0  I-       ao 0)   0       o (\   C')   ' IT   (O
X    X- 1       -  N-   N-  N-  N-  N-  N-  (0   ao  OD   CO  (0  (0  (0  (0  (0  (0   0)    )   0    CD    )   0)    )
0)  CD   0)   0)  cn   0)0)         )   0)   0)   0)  0y)   )    )   0)  0)  0)    )   0)   0)   0)   0)   0)   0)    )   0)   cz



1995 $US Billion                      Figure 4.3 Grant Equivalent of Official Loans
22.0 -_ 
20.0-
18.0 -
~Using adjusted grant elements
16.0 -                                                                                                  _
-0--Using conventional grant elements                                      -
14.0 -              tUsing 10%  discount rate
10.0
8.0- \ / -
I O
6.0     -,               ,     ,             ,     ,      ,      ,  _         ,            ,      ,     ,      ,,,, 
LO)   co      -    a:    0)    0       co      N     C)    "t    LO    (0    C- OD    O) 0                i     (N    CV)           LO
N-    N-     N-    N-    N-    (0    o0(0    00    (0    (  0(0    (0    (0    co    0)    OD    0)    0')   0) cn)
0)ote:0)    C)    0)    C)    0)    0)    C)    0)    a)    0)    CY)                  )    0)    0)    0)    a)    0)    a)    0)
Note: Total for 133 countries in the sample.



z
CD      X0         ONA0                                    O D        o .
@   151975
1976
D   1977                                                                       a
CD   1978 
U)>
FD   1979
1980
1981 
1982                                                                            l
1983                                                                           0
1984
0
1985
ch
1986m
1987
1988                                                                           Z
1989                                                                           0
1990
1991
1992
1993
1994
1995           1_           _  _ _   _ _  _ _ _ _  _ _  _ _ _  _



1995 $US Billion                                  Figure 4.5 Net ODA vs Total EDA
45.0 -
40.0 -Effective Development Assistance
400 Net Official Development Assistance
35.0                                                  ---                          _  _ _  _
30.0 
25.0 -  
20.0 -__ =_ _
LO    co .0-    co    0)    0                         )           LO     ( -- r-                   0 -             E             a r
-    N       1-  a  o-    r-    COC   CC)   cO    CO    co    co            C O C o         C      0      a)    0)    0)    0       0L
0)    0)    CD    O)    0)    0)    0         0)   C0)    0)    0)    0)    0         0)    0)    cn    0)    0        0      0)    0)
Note: Excluding technical Assistance and debt forgiveness. Total for 133 countries in the sample.



Figure 4.6 Average Grant Element by Recipient Region 1990-95
60%                             .
50%       El Using adjusted grant elements                                                    _
1 Using conventional grant elements
40%                                         _  _ _   _   _      _  _
20  %i   20Il
Europe and    Latin America and  Middle East and    East Asia and  South Asia     Sub-Saharan
Central Asia     Caribbean       North Africa       Pacific                           Africa
-10% -



Figure 4.7 Net ODA Relative to EDA by Recipient Region,
(1990-95 Totals)
East Asia and Pacific
Latin America and Caribbean
Middle East and North Africa
Europe and Central Asia
Sub-Saharan Africa
South Asia
100%        110%         120%         130%         140%         150%         160%         170%
Note: Excluding TA and debt forgiveness.



FIGURE 4.8 Ranking of Bilateral Aid by Donor: ODA/GNP vs EDA/GNP
1981-85                                                1986-90                                                   1991-965
RANK                ODA/GNP                     EDAIGNP                         ODAIGNP                     EDAIGNP                      ODAIGNP                    EDAIGNP
1     NORWAY         0.366%      NORWAY         0.366%         NORWAY           0.417%     NORWAY           0.415%       NORWAY          0.386%     NORWAY         0.382%
2     NETHERLANDS  0.351%        SWEDEN         0.347%         SWEDEN           0.378%     DENMARK          0.351%       SWEDEN          0.318%     DENMARK         0.327%
3     SWEDEN         0.349%      NETHERLANDS   0.313%          DENMARK          0.364%     SWEDEN           0.338%       DENMARK         0.315%     SWEDEN         0.314%
4     AUSTRALIA      0.238%     AUSTRALIA       0.240%         NETHERLANDS   0.312%        NETHERLANDS   0.295%          FRANCE          0.244%     FINLAND         0.220%
5     DENMARK        0.221%      DENMARK        0.217%         FINLAND          0.208%     CANADA           0.209%       FINLAND         0.235%     NETHERLANDS  0.218%
6     CANADA         0.184%     CANADA          0.190%         CANADA           0.207%     FINLAND          0.196%       AUSTRIA         0.214%     FRANCE          0.195%
7     JAPAN          0.175%      BELGIUM        0.151%         JAPAN            0.200%     AUSTRALIA        0.169%       NETHERLANDS   0.190%       AUSTRIA         0.190%
8     BELGIUM        0.158%      UK             0.123%         AUSTRALIA        0.182%     SWITZERLAND    0.130%         AUSTRALIA       0.180%     AUSTRALIA      0.157%
9     FRANCE         0.136%      JAPAN          0.121%         FRANCE           0.164%     ITALY            0.123%       JAPAN           0.160%     SWITZERLAND   0.148%
10    UK             0.115%      FRANCE          0.112%        ITALY             0.145%     FRANCE          0.104%       SWITZERLAND   0.152%       GERMANY         0.141%
11    SWITZERLAND  0.113%        SWITZERLAND   0.109%          SWITZERLAND    0.135%        BELGIUM         0.102%        GERMANY         0.146%    CANADA          0.127%
12    UK             0.115%      GERMANY         0.092%        GERMANY           0.114%     JAPAN           0.100%       CANADA           0.133%    BELGIUM         0.121%
13    USA            0.086%      USA             0.081%         BELGIUM          0.104%     GERMANY         0.097%        BELGIUM         0.117%    LUXEMBOURG   0.117%
14    FINLAND        0.082%      FINLAND         0.080%         UK               0.077%     UK              0.084%        LUXEMBOURG    0.117%    JAPAN             0.084%
15    ITALY          0.061%      AUSTRIA         0.052%         USA              0.060%     USA             0.061%        ITALY           0.104%    ITALY           0.084%
16    NEW ZEALAND  0.043%        NEW ZEALAND   0.051%           PORTUGAL         0.051%     PORTUGAL        0.048%        SPAIN           0.089%    UK              0.074%
17    AUSTRIA        0.041%      ITALY           0.050%         SPAIN            0.040%     NEW ZEALAND    0.036%         UK              0.068%    USA             0.067%
18    SPAIN          0.028%      SPAIN           0.029%         AUSTRIA          0.040%     AUSTRIA         0.035%        USA             0.065%    SPAIN           0.048%
19    IRELAND        0.020%      IRELAND         0.020%         NEW ZEALAND    0.036%       SPAIN           0.022%        PORTUGAL        0.042%    PORTUGAL        0.039%
20    PORTUGAL       0.002%      PORTUGAL        0.011%         LUXEMBOURG    0.019%        LUXEMBOURG    0.019%          NEW ZEALAND   0.039%      NEW ZEALAND   0.039%
1. Adjusted for ODA debt forgiveness



Policy Research Working Paper Series
Contact
Title                           Author                  Date              for paper
WPS2034 Information, Accounting, and the  Phil Burns             December 1998      G. Chenet-Smith
Regulation of Concessioned      Antonio Estache                            36370
Infrastructure Monopolies
WPS2035 Macroeconomic Uncertainty and     Luis Serven            December 1998      H. Vargas
Private Investment in Developing                                           38546
Countries: An Empirical Investigation
WPS2036 Vehicles, Roads, and Road Use:    Gregory K. Ingram      December 1998      J. Ponchamni
Alternative Empirical Specifications  Zhi Liu                              31022
WPS2037 Financial Regulation and          James R. Barth         January 1999       A. Yaptenco
Performance: Cross-Country      Gerard Caprio, Jr.                         38526
Evidence                        Ross Levine
WPS2038 Good Governance and Trade Policy: Francis Ng             January 1999       L. Tabada
Are They the Keys to Africa's Global Alexander Yeats                       36896
Integration and Growth?
WPS2039 Reforming Institutions for Service    Navin Girishankar  January 1999       B. Casely-Hayford
Delivery: A Framework for Development                                      34672
Assistance with an Application to the
health, Nutrition, and Population
Portfolio
WPS2040 Making Negotiated Land Reform     Klaus Deininger        January 1999       M. Fernandez
Work: Initial Experience from                                              33766
Brazil, Colombia, and South Africa
WPS2041 Aid Allocation and Poverty Reduction Paul Collier        January 1999       E. Khine
David Dollar                              37471
WPS2042 Determinants of Motorization and    Gregory K. Ingram    January 1999       J. Ponchamni
Road Provision                  Zhi Liu                                    31022
PWS2043 Demand for Public Safety          Menno Pradhan          January 1999       P. Sader
Martin Ravaliion                          33902
WPS2044 Trade, Migration, and Weifare:    Maurice Schiff         January 1999       L. Tabada
The impact of Social Capi'.al                                              36896
WPS2045 Water Challenge and Institutional    R. Maria Saleth     January 1999       F. Toppin
Response (A Cross-Country       Ariel Dinar                                30450
Perspective)
WPS2046 Restructuring of Insider-Dominated   Simeon Djankov      January 1999       R. Vo
Firms                                                                      33722
WPS2047 Ownership Structure and Enterprise  Simeon Djankov        February 1999     R. Vo
Restructuring in Six Newly                                                 33722
Independent States



PoiHcy Researeh Working Paper Series
Contaict
dri^Se  Author          Date              for paper
WV,-, 82 D,orruntion in Economic           Shang-Jin Wei           February 1999       C. Bernardo
Development: Beneficial Grease,                                              31148
Minor Annoyance, or Major Obstacle?
WPS2 r:  Household Labor SUDply,           Kaushik Basu            February 1999       M. Mason
lJnemployment, and Minimum Wage Garance Genicot                              30809
Legislation                      Joseph E. Stiglitz