Non-governmental organizations, or NGOs, have become increasingly important in the international aid community, not only in their role of translating private and corporate funding into humanitarian projects overseas but also for their growing role as recipients of official government funding, for example from USAID and the World Bank. The growth in the latter comes largely as a result of the idea that NGOs can be more efficient providers of humanitarian services and economic development aid. As NGO activity has been dramatically evolving, it is relevant to ask how NGO behavior fits in with the recent debate on the conditionality/selectivity of aid allocation. Past economic studies have considered whether aggregate ODA, World Bank assistance and bilateral aid (by country) follow a pattern of conditionality/selectivity. In this paper this analysis is extended to country and activity choices made by US-based NGOs.
Specifically, a regression model is estimated to answer the question whether or not specific country characteristics could pull more (or fewer) NGOs into operating within a country. Characteristics of interest include need based variables (income per capita, human development index value, growth in GDP per capita and population); performance based variables, indicating areas of desired change (percent change in government consumption, percent change in foreign direct investment, percent change in trade and overall trade openness in the economy); and regional preferences (Sub-Saharan Africa and Central/South American dummy variables).
A second layer to NGO activities, however, also needs to be considered: the intersection of NGO country and activity choices and the receipt of official government donations, in this empirical study the US. The initial role of NGOs in the aid community can be described as correcting a failure of governments and international organizations to provide either the amount or diversity of aid funding that corporations or private citizens find preferable. In their role representing the median voter preference, official governments can be dramatically under-providing relative to the preferences of voters far away from the median. Also, government aid strategies often are constructed within the context of wider policy and strategic concerns and desired outcomes. NGOs provide an outlet for individual citizens, corporations and foundations to more directly address their preferences with regards to development assistance, both in terms of funding levels and project content. Thus when official funds are channeled through NGOs an inherent tension can be created: the more funding NGOs receive, the more activities they are able to conduct overseas, yet if they receive both private and public monies they are accountable to both donor categories which may not have the same outcome goals. In addition, NGO in-country identity and legitimacy depends fundamentally on their ability to act as independent providers which may also be challenged with increased accountability to government donors.
In this paper, to address the question of project content, data on self-reported NGO activity areas (for example, HIV/AIDS, Business Development, and Gender Issues) distinguishing between those receiving and not receiving government funding is analyzed. Z-statistics are then used to determine whether activity proportions are different across the two groups.
II. NGOs and the Changing Aid Agenda
Until recently the supply side of the international aid community has been primarily dominated by a small group of players: developed country governments-through the use of bilateral aid-and the World Bank and United Nations. The aid allocation decisions of these players have been based on altruistic as well as self-interested motivations including the opening up of markets overseas and national security. Ultimately aid decisions, like other policy decisions, can be modeled as representing the preferences of the median voter of either individual country populations or even multi-country populations. Thus governments and multilateral aid organizations, in their aid decisions, are clearly aware of public opinion.
In 2002 the US Agency for International Development (USAID) released a document to emphasize the overall altruism of the American people. A key goal of this document was presenting private contributions, specifically those to NGOs, as part of the overall US aid portfolio. As the primary administrator for US bilateral aid, in this document USAID also cites public polling results to help illustrate its own aid decisions. According to surveys cited in this document, US public support for international aid has always been lower than support for domestic causes; there had been a trend [pre September 11th] for greater support for humanitarian rather than security reasons; and polls taken in 2000 found that only 40-47% of Americans thought the government was spending too much on foreign aid versus 65-75% polled with that belief in the mid 1990s. [p. 130] These polling results imply that the amount and targeting of aid projects through USAID are responsive to public opinion.
Lancaster (2000) also discusses the role of the political process in US bilateral aid decisions. According to Lancaster:
The major issues of the day affect the congressional politics of any number of government programs. But this is particularly true of foreign aid, for two reasons. First, it is a set of programs with a relatively weak domestic political constituency and thus is particularly vulnerable to the prevailing political winds … Second, foreign aid is a tool of US foreign policy, so when there are controversial issues involving US foreign policy, foreign aid legislation becomes a lightning rod for criticism. [p. 46]
In addition to the political dimension, the implementation of aid through these large bilateral and multilateral bureaucracies also has certain efficiency concerns. Easterly (2002), for example, characterizes this community as a ‘cartel’ that has historically underperformed due to excess bureaucracy, poor definition of aid output (money disbursements rather then actual improvements in the target country), and excessive administration. A more recent “White Paper” investigating the link between foreign aid and security released by USAID (2004) further discusses the confusion of having too many competing goals and objectives set out by the political process. In this case, the fear is expressed that poor definitions of the goals of aid can undermine the effectiveness of foreign aid in increasing national security.
The rise of a new class of players in the community, non-governmental organizations or NGOs, was brought about, at least in part, by their ability to address two problems in the traditional supply of aid. Firstly, by donating to NGOs directly, individuals, corporations and foundations are better able to express their preferences for specific aid projects, aid goals or even country involvement. NGOs need not be accountable to whole country populations and direct political processes, but rather respond to specific population subgroups, namely to those who actively contribute resources and to those whom they work with in-country. Also, as NGOs are smaller organizations, there are smaller costs of bureaucracy and the potential for greater efficiencies in the use of donated funds. Scott and Hopkins (1999) discuss two areas where NGOs might be able to benefit from greater efficiencies than other official aid organizations.
A first potential cost savings for NGOs is in their labor costs. NGOs tend to have access to greater numbers of volunteers and their paid employees may be, in the words of Scott and Hopkins, more altruistic. Altruistic workers in their model will benefit from a ‘warm glow’ of seeing their efforts work to make others better off and therefore may be willing to work at a lower wage. A second area for greater efficiency would be in the ability to convert organizational inputs into valued in-country outputs. In this case it is argued that NGOs may have access to better information about what would be valued in-country and therefore could implement more effective projects. Structurally, NGOs are not limited to working with official state representatives, as is often the case with more official forms of aid. In addition, NGOs are also able to provide emergency relief aid and services in a more timely fashion and efficient manner, responding to acute crises overseas quickly.
This assertion of the benefits of NGOs is supported by the USAID (2002) report referenced previously. The efficiency argument is discussed as well as the ability of NGOs to work in sensitive political climates, respond quickly to humanitarian crises, and form bonds with local institutions. In addition, NGOs are praised for promoting US values: “Most importantly, they foster pluralism, volunteerism, and compassion-values that have characterized the United States throughout its history.” [p. 141] While NGOs may be presumed to be more efficient, as Edwards and Hulme (1996) discuss, substantiating this claim with empirical research has been difficult.
The rise of NGOs has occurred simultaneous to, and was certainly influenced by, what Edwards and Hulme (1996) call the ‘New Policy Agenda’ in the aid community. This agenda is defined by the increased emphasis on the efficiency and accountability of foreign aid and the use of conditional criteria in determining the allocation of aid to insure its maximum effect. In terms of this new agenda, NGOs attract official funding through their link to efficiency improvement and the NGOs, in turn, have turned increasingly to official funding as a method to increase revenues. In Edwards and Hulme (1996) the authors note “NGOs are becoming more dependent on official aid, especially during the last year or two when there has been a discernible flattening-out of voluntary income from the public in many Northern countries … it is common to find government grants making up between 50% and 90% of the budgets of major NGOs in Scandanavia, the Netherlands and Canada” [p. 962]. The second emphasis, the conditionality of aid, has a long history in the aid literature. As Easterly (2002) discusses, allocating aid conditional on a non-need based specific in-country environment in the US can be traced at least as far back as to the Kennedy Administration. A more global history of the notion of conditionality is outlined by Doornbos (2001) which in contrast to the reception in the US, seems to suggest that the rest of the world has viewed the notion of conditional aid with more skepticism.
Amongst economists and policy makers, in particular, this notion of conditionality has also been discussed and debated in terms of the appropriateness of the so-called “Washington Consensus” used to condition IMF and World Bank funding. The idea that donor countries and organizations would seek to condition their aid disbursements, as suggested above, is not a new one. Most aid conditioning is done considering political and economic progress, and top down donor self motives. While economic and political goals can benefit target countries and their populations, it can also be used to further integrate countries into the world economic and political system, a goal often of great interest to donor countries.
A more modern approach, for example in the paper by Hermes and Lensink (2001), has been to replace the word “conditional” with “selectivity” implying that aid be allocated not to condition on very specific policy reforms but rather in recognition of realized outputs with respect to poverty, macro stabilization and institutional reform. This seems to be supported in documents describing future goals of the World Bank and IMF. In addition, some would move beyond the tight definition of economic reform as presented by the ‘Washington Consensus’ and expand to a definition of conditioning on ‘good governance.’ A comprehensive study on this broader set of conditioning factors is given in Neumayer (2003) . In the US, this broader set of goals is a key part of the Millennium Challenge Corporation aid disbursement process.
As Easterly (2003) documents, the current policy environment in the US and World Bank is that of a formal policy of conditioning aid. From speeches made in 2002, he quotes the President of the World Bank, “We have learned that corruption, bad policies, and weak governance will make aid ineffective.” [p. 25] and President George Bush “Money that is not accompanied by legal and policy reform are often wasted … Sound economic policies unleash the enterprise and creativity necessary for development. So we will reward nations that have more open markets and sustainable budget policies, nations where people can start and operate a small business without running the gauntlets of bureaucracy and bribery.” [p. 25] The USAID 2002 Report on US altruism also weighs in “The emerging consensus in the development community is that aid reduces poverty only when economic policies support sustained economic growth … [although] economists and scholars have also concluded that countries implement economic reforms when they choose to-not because of aid offered or withheld.” [pp. 130-131] According to USAID, the US public supports this notion as well.
As official funding sources have become, at least rhetorically, increasingly tied to conditional criteria and NGOs have become increasingly funded by public sources, several questions can be posed: Are NGO participation decisions responsive to economic conditional criteria? Does NGO participation behave like other forms of foreign aid? Does official funding within the NGO community make it more likely to follow a conditional aid policy? If so, this issue has important ramifications for NGOs: as Edwards and Hulme (1996) state: “NGOs … also need to invest more in their own organizational development so that they are better able to identify the negative impact of changes in their funding sources or role, and act accordingly.” [p. 969] In addition, considering specific economic criteria to condition or select may be quite counter-productive to a very important role NGOs play in the aid community, namely providing humanitarian aid to countries in crisis. In addition, Guillaumont and Chauvet (2001) point out that conditioning aid on specific country-based goals may lead to biased outcomes if exogenous factors or shocks beyond a country’s control have domestic impacts. They call for an aid policy that can act as “an insurance as well as a reward.” [p. 67] In this view, a more nuanced approach would be needed for aid organizations to distinguish domestic performance differentiated from international, exogenous shocks.
Since the focus of this paper is on the US government funding of US-based NGOs, it is important to discuss the funding criteria of USAID . In order to receive maximum benefits from working with USAID, NGOs register with the organization as a PVO (private voluntary organization.) A registered PVO is required to receive funding from private sources, have overhead expenses account for no more than 40% of costs, have a board of directors, not have terrorist ties and needs to “fit[s] within USAID priorities.” [USAID (2002) p. 140] There has been a dramatic rise in the number of registered PVOs, from 138 in 1979 to 436 in 2000. [p. 141] The top twenty PVOs receiving funding from USAID include CARE, Catholic Relief Services, World Vision and Save the Children. These top twenty PVOs accounted for about two-thirds of the total PVO funding from USAID and received an average of $43 million in grants and contracts. [p. 141]
The total magnitude of aid funded through NGOs is significant. As Lancaster (2000) summarizes, funding through NGOs is an important part of the assistance allocated for development purposes: “Detailed data are not available on the proportion of US bilateral aid channeled through NGOs, but USAID officials have estimated that about a third of the nearly $2 billion Development Assistance funds are implemented by NGOs.” [p. 10]
III. Empirical Studies on the Conditional Allocation of Foreign Aid
In their allocation decisions, do countries and multinational organizations truly follow a policy of conditioning aid? There are a number of other empirical studies asking this question both with respect to wider ‘good governance criteria’ and economic criteria. In addition to these conditional criteria, the relative importance of ‘altruistic’ (need based variables for the receiving country) and ‘self-interest’ (strategic based variables of the donor country) are considered.
In their paper, Trumbull and Howard (1994), the authors promote an early panel data model for studying aid allocations to recipient countries. The dependent variable in their study is ODA per capita and they use a fixed effects model with infant mortality, political rights, population and income as explanatory variables. Their overall results are that need matters, but in their regression the need variables that were significant did not include income but were rather the infant mortality and political climate.
Later in their seminal paper, Burnside and Dollar (2000) look at the interaction of policy (justifying conditionality) and growth. While their results with respect to growth have gotten wider attention from economists, they also estimate the relationship between aid (by recipient country), policy, need and strategic variables. The goal of their estimation is to extend beyond just World Bank disbursements but also consider total bilateral, multilateral and total aid (all measured per capita by recipient country.) They use a data set including about 40 low-income countries (defined in their data set as an income per capita less than $1,900) for their aid regression and look at six four-year time periods from 1970 to 1993.
Their results with respect to an overall policy variable (which includes the ‘Consensus’ variables of openness, government surplus and inflation) is mixed: the World Bank regression and multilateral aid regression provide evidence that such aid is increased with a good policy environment. To the contrary, bilateral aid and the total of all aid do not have a significant relationship to policy. All four measures of aid have a negative and significant relationship to initial GDP per capita (measured in 1970). The dummy variable for Sub-Saharan Africa (also to be used in this study) is negative and significant in the World Bank regression only. The Central American dummy (this study will use a Central and South American dummy variable) is not significant in any of the regressions. The only other significant variables in these regressions are a negative and significant impact of population (found in all four regressions); a positive impact of a dummy variable for Egypt for the total and bilateral regressions (attributed to the bilateral aid to Egypt from the US for strategic reasons); and a negative and significant coefficient on lagged arms imports found only in the World Bank regression.
A widely comprehensive review of the literature on the impact of ‘good governance’ on aid, including such variables as a human rights index, corruption index and a measure of political stability, is given in Neumeyer (2003). In chart form, the author analyzes over forty papers and the impact on aid disbursements of ‘donor interest variables’ (summarized DI), ‘recipient need variables’ (RN) and ‘good governance variables (GG). In most cases the dependent variable is a measurement of aggregate aid per capita by recipient country. Of course, the results within these papers vary broadly but several themes are found. Neumeyer finds overall evidence that both donor interests and need variables matter. It should be noted that donor interest is broadly defined and includes religion, race, trade and investment figures.
In the case of US bilateral aid (most relevant in comparison to the US-based NGOs of this study), Neumeyer explains:
These studies differ of course in their results from each other, and sometimes
substantially so, due to different research designs. Nevertheless, most of these
studies come to the result that more respect for political freedom and, albeit less
clearly so, respect for personal integrity rights is rewarded with a higher probability
of receiving any US aid as well as with a higher level of aid allocated. [p. 30]
The issue of aid disbursements, then, has been investigated using the aggregate bilateral, multilateral and World Bank disbursements. In this paper a very different measure of aid is considered: the presence and activity of non-governmental organizations. It is believed that this is the first study to consider empirically, at the NGO and country level, how activity decisions (both in terms of country and area) are made and whether these decisions follow larger patterns of aid allocation decisions.
IV . Empirical Model
The data set for this study is built using NGO-specific data from the organization Interaction (The American Council for Voluntary International Action) and country characteristic data from the UN Development Report and World Bank Social Indicators of Development. In the Interaction data set, NGOs report (among other financial disclosures) a list of countries in which they are operating and a list of activity areas which they fund. For example, “NGO X” may report a list of ten countries and four different activity areas. The lists are independent, however, and it is impossible to say that “NGO X” is supporting HIV/AIDS activity in Botswana; it is only possibly to say that “NGO X” is active in Botswana (among other countries) and supports HIV/AIDS work (among other activities.) In the regression analysis of the paper, data are used based on individual countries: the number of NGOs active in the country as well as economic characteristics. In the later analysis based on activity, the data is based on the NGOs, means are constructed across “the percentage of NGOs” active in specific program areas.
The NGO data is taken from the “InterAction Member Profiles: 2001” and covers the period of 1999-2000. InterAction provides the data based on NGO self-reporting. The data includes 120 US-based organizations in the regression analysis and 142 for the activity analysis. Thirty-five of these organizations received no monies from the US government, either in the form of contracts or grants, eighty-five of the organizations did receive such funding. A summary of the NGO data availability is provided in Table 1.
The Member Profiles include self-reported figures for revenues: individual and corporate contributions, US government grants and contracts, an aggregate measure of other governmental and non-governmental funding (including World Bank) and foundation funding. Expenses are also reported in terms of program expenses, administrative costs and fund-raising. Most organizations also report the countries in which they have been involved. While financial statistics are reported separately for 1999 and 2000, country involvement lists are only reported once and therefore financial data is matched to 2000 country indicators. Unfortunately data does not match up specific financial amounts with actual country involvement but only whether on organization was active in a country or not. This will have ramifications for the econometrics used.
The World Bank model uses as a dependent variable total IBRD loans and IDA credits (in $US million) by country. The dependent variable for the NGO regressions is the count of relevant NGOs active in a particular country. First a model is run for all NGOs and then the sample is split to reflect those NGOs either receiving or not receiving US government funds. A summary of the data series used is given in Table 2. Summary statistics for the NGOs and the country indicators are given in Tables 3 and 4.
In terms of NGO averages, the country involvement measures the average of the actual number of countries within which the NGOs are involved. In this case, regardless of the receipt of official funding, NGOs in each group is approximately thirty countries.
The dummy variable SSA, represents Sub-Saharan African countries which represent 31.4% of the full sample of 153 countries; CSA represents Central and South American countries, 21.6% of the full sample. Both static (2000 value) and dynamic measures (% change from 1995 to 2000) of FDI, government consumption and trade are used to capture conditionality. Although all the static measures are originally included in the model selection process, only the openness of trade variable is significant. The data itself show a strong tendency toward openness in trade and FDI and lower increases in government consumption. Note that religious representation in-country (as constructed by CIA World Factbook) was tested in the NGO regressions and the only consistently significant religion was “Catholic” which turned to be highly correlated with the CSA dummy. For this reason, religion is not included in the final specifications.
An important issue in understanding NGO activity is the use of project funding for humanitarian emergencies versus longer run, development goals. Because this study requires World Bank economic data to match with the NGO data, there is a bias against including countries experiencing such humanitarian crises. For example, in this data set, neither Afghanistan nor East Timor can be included due to missing data although both had considerable NGO presence.
B. Econometric Model
Estimating the model of country involvement of NGOs brings about particular econometric issues. In one sense, the data resembles what is known as count data: positive integers, the number of NGOs involved in country “i,” and therefore OLS is not the best regression method. However, there is a natural upper-bound to the count, the number of NGOs in the data set, and therefore traditional count data techniques are not appropriate.
For this reason, a grouped logit model is used. The model is estimated in STATA using the command glogit. This routine uses weighted least squares on the following specification (in matrix form):
pj represents the number, in this case, of active NGOs divided by the total number of relevant NGOs. The dependent variable itself is the log of the odds ratio. The one drawback of this specification is that the estimated coefficients are not easily interpreted. For that reason, in addition to reporting these coefficients, marginal effects are computed within STATA. These marginal effects represent the change in the predicted estimated number of relevant NGOs active in a particular country given a one unit change in the X-regressor. These marginal effects are computed at the mean values of the X regressors. In the case of the dummy variables for Sub-Saharan African or Central/South American countries, the marginal effects represent the difference between countries in those regions and the omitted group.
For comparison purposes an OLS model is also reported with the dependent variable of IBRD funds (in millions).
C. Results—Patterns of Country Involvement
Table 5 gives the first results on the models estimating patterns of country involvement. The first and fourth columns in the Table represent a baseline regression provided for comparison with the NGO count models key to the paper. The dependent variable for this model is simply total IBRD and IDA funding and traditional OLS techniques are used.
Considering these World Bank results for the need only variables (column one), it is notable that the coefficients on per capita income, the human development index, population and the Sub-Saharan Africa dummy are all significant. The results on per capita income and the human development index taken jointly suggest that the World Bank responds to lower levels in education and health and, given similar education and health profiles, supports those countries with a higher per capita income. There is also an understandable preference to fund countries with a greater population. Finally, the significance of the SSA dummy suggests that of identically needy countries in Africa and the group not represented by a dummy variable (Asian countries, for example), the African country would receive less funding.
The second column in Table 5 represents the first group logit model with the number of NGOs in a particular country as the dependent variable. The independent variables, like the World Bank regression discussed above, represent only need based variables, or a strict altruistic model: the need variables of per capita income, the human development index, GDP growth and population are included as are the dummy variables for Sub-Saharan Africa and Central and South America. The label “Total” refers to the fact that NGOs within a country are not differentiated based on whether or not they receive government funding. In the NGO regression, the impact of per capita income is negative and significant meaning a larger number of NGOs would be involved with poorer countries. The marginal effect figure, however, indicates the effect is small: the involvement of one more NGO would require a decrease in per capita income of $715. The population term is positive and significant but again on the margin, the impact is small. It would take an increase in population of 39 million to draw in another NGO, all other variables held constant. In terms of the regional dummy variables, the SSA dummy is not significant in the case of NGOs but interestingly the CSA is positive and significant. The magnitude of the marginal effect indicates that for the same level of “needs,” between the CSA and omitted country group, almost 12 more NGOs would be involved in the CSA country. This is an interesting result highlighting the fact that the study is based on USA-based NGOs and there seem to be strong regional preferences .
When the full set of regressors is used, including the conditional variables, both the World Bank and NGO regressions show a strong, negative impact with respect to openness to trade in 2000. This seems contrary to the intent of conditioning which would reward openness. In the NGO regression a decrease in trade as a % of GDP of 5% would cause another NGO to operate in the country. The dynamic change in trade is not significant in either regression meaning that past changes in openness are also not being rewarded. The other dynamic, conditional variables are insignificant in both regressions. In the World Bank regression the per capita income, human development index, population and SSA variables continue to be significant, however in the NGO regression when the conditional variables are controlled, the need variables lose their significance.
In Table 6 similar group logit regressions are run splitting the sample across NGOs receiving or not receiving US funding. The dependent variable in columns one and five represent the number of NGOs receiving government funding active in country “i;” in columns three and seven the dependent variable represents the number of NGOs not receiving government funding active in country “i.”
In the case of the altruistic model, the impact of per capita income is only realized in the decisions of those NGOs receiving official funding: the coefficient on that regression is negative and significant. Population is significant and the magnitude of the marginal effects in both regressions suggests an increase in population would be required to pull another NGO into the country. In terms of the dummy variables, the CSA dummy is significant for both groups but has a larger impact, 8.115 versus 3.101, for the group receiving government funding. (The order of magnitude is not dissimilar, though, given the larger number of NGOs receiving funding versus those not.)
Turning to the model including the conditional variables, the group receiving official funding again has a negative impact of per capita income and trade openness. The group not receiving funding has a significant dynamic conditional variable: percent change in government spending. The coefficient of government spending is positive, contrary to the implied conditional but signals, perhaps, the response of NGOs to increased social spending by governments.
With the inclusion of the conditional variables, it is interesting to see that the NGOs receiving government aid seem to have become more regionally neutral: neither of the dummy variables is significant. In the case of the other NGOs, the SSA remains negative and significant with a marginal effect of -3.879. In the case of the non-government funded NGOs population remains positive and significant; in the government funded regression population is not significant. While the marginal effect remains small, the result is suggestive: non-government funded NGOs may have more incentives to become involved in large countries for fund-raising reasons. Potential donor individuals and foundations may be more moved to contribute to country activities in well-known and large countries.
One area of concern is the actual interpretation and computation of the marginal effects. As the descriptive statistics show, mean values of the regressors are often quite different from the median. Therefore in Table 7 the marginal effects from Table 6 (computed at the mean values) are summarized and additional marginal effects at the median values of the Xs (for countries in each regression) are computed. To aid further in interpreting the marginal effects, for each value an additional value given for “NGO + 1” is computed and represents the necessary change in the continuous X-regressor required to pull another NGO into the country. In same cases, the marginal effect is reported to a higher level of accuracy to better compute its impact. The marginal effects on the dummy variables SSA and CSA are interpreted simply as the additional (or fewer) NGOs operating in each category of country (versus the omitted group) and cannot be computed for “NGO + 1.”
Overall, the marginal effects computed at the mean and median perform similarly. Although with respect to the population variable, for example, a small difference can impact “NGO + 1” dramatically. For example in the government funded NGO regression with only the altruistic variables, an increase in population of 57.143 million would draw another NGO into the country when the marginal effect is computed at the mean versus 60.606 million when computed at the median. The impact of per capita income on government funded NGOs is significant across the regressions (and computations) and shows that a decrease of between around $770 (for the first regression) and $700 for the second is required to draw in another NGO.
The significance of the trade variable in the full regressions is also apparent. For government funded NGOs a decrease in openness of around 6.5% could draw in another NGO; for non-government funded NGOs, an increase in one operating NGO would require a decrease in openness of around 21% or 22%. The non-government funded NGOs would also require large increases in government consumption to be drawn in: around 41-45%. An increase in population of over 250 million would also be required to pull in another non-government funded NGO.
D. Results—Patterns of Activity Involvement
In Table 8, the results on the second area of analysis are reported, namely the areas of program activity. Of 148 NGOs with full data, 48 (or 33.8%) of them receive no government funding and 94 (66.2%) of them do. The table reports the number of each category involved with the individual activities such as “AIDS/HIV,” “Strengthening of Civil Society,” and then the percent of NGOs involved in each activity belonging to each respective category.
For example, according to the numbers, 17 NGOs not receiving government funding and 51 receiving such funding are involved, in some way, with “Agriculture and Food Production.” Of the 68 total NGOs involved in this area, therefore, 25% of them do not receive official funding and 75% of them do.
The column tracking “diff” indicates the difference in within group proportions of those NGOs involved. Again in the case of “Agriculture and Food Production” this tracks the fact that a higher proportion of government funded NGOs (51/94 or 54%) than non-funded (17/48 or 35%) are involved in this activity. The magnitude of 19% indicates the difference between the proportion of non-funded NGOs versus funded NGOs (or 35%-54%). The final column reports the compute z-statistic for the null hypothesis of equality in proportions active by activity across both groups. Statistically significant (at the 5% level) differences in these proportions exist for the categories “Agriculture/Food Production,” “HIV/AIDS,” “Business Development, Credit”—which all have a higher proportion of government funded NGOS—and “Policy Research and Analysis”—which has a higher proportion of non-funded NGOs.
The greatest difference (in magnitude and significance) is in the area of “Business Development and Credit.” This may reflect a strong government priority in the development of credit markets and business in the development process. Involvement in the fight against HIV/AIDS is also more proportionately being engaged by NGOs funded by official sources. The one area where non-funded NGOs are proportionately more active is in the area of “Policy and Research Analysis.” This may be due to the fact the official sources of policy research exist outside of NGOs and of direct service to the government (for example the State Department and USAID.) Thus, government officials may be less interested in funding such activities through NGOs. On the other side, independent NGOs not receiving official funding may have an interest in producing their own research.
Based of the results of this study, US-based NGOs receiving government aid seem to be basing their decisions on variables found to be significant to other forms of aid. Lower per capita income, for example, draws in more NGOs. There is evidence that policy or economic environment matters as well. Countries less open to trade draw in more of both government funded and non-funded NGOs. For NGOs not receiving government funding, interestingly, need matters less and policy matters more. Growth in government consumption is rewarded by non-funded NGOs. The regional dummy variable results are also compelling: in the fullest model the NGOs receiving funding are regional-neutral while there is a negative bias against Sub-Saharan African country involvement for those NGOs not receiving government funding. Also while funded NGOs are neutral to population, non-funded NGOs respond to countries with higher populations.
The region and population results taken together suggest an interesting interpretation. Non government funded NGOs may be more vulnerable to promoting certain “popular” regions or large countries in order to fund raise from individuals. Those NGOs receiving government funding may not be under the same constraints. In working with USAID professionals, they may be more able to fund programs in smaller countries and those countries more “neglected” or “ignored,” a critique often discussed in the context of the lack of necessary foreign aid funding to Sub-Saharan Africa. Thus, while government funding may bias the activity choice of NGOs, it may also enable them to operate in a wider range of in-country environments.
In terms of activity content, although the propensity to engage in many activities is neutral whether or not the NGO receives official monies, those NGOs that do receive government funds have a higher probability of engaging in food-related, HIV/AIDS related and business development activities. To the contrary those not receiving official funding are more likely to be engaging in policy research and analysis. This last result could be investigated further through the lens of the critique put forward by Edwards and Hulme (1998) on the legitimacy and compromise of those NGOs receiving government funding.
There are some drawbacks and potential biases to the data used in this study. All data for NGOs is taken from InterAction and therefore the sample only represents members of the organization. To the extent that this group may be self-selected (representing, for example, larger NGOs who might benefit more from membership), results may not be fully applicable to all US-based NGOs. Further, members of InterAction may be more likely than NGOs overall to receive US government funding. In addition, the data does not allow linkage between country activity and country funding. Therefore a country may have more NGOs active within its borders but overall less financial resources from the NGO community than another country. This necessitates an imperfect comparison with previous results from the literature.
Despite these shortcomings, it is hoped that this paper provides a first step in incorporating the NGO community in the overall foreign aid literature. A finding that there are some differences in the behavior of NGOs based on government funding (country characteristic and activity) means importantly that NGOs can have incentives to get involved with the US policy process. NGOs, for example, active over time in areas like agriculture and HIV/AIDS on behalf of the US government may have strong incentives to keep these funding streams. This kind of incentive was demonstrated by the NGO involvement in the US policy debate about the future of food aid. To the extent that NGOs provide contributors with the ability to express preferences outside of the standard US policy process, these differences in behavior could be of great concern. More empirical evidence on the impact of government funding on NGO behavior, however, is necessary to define more clearly the actual costs and benefits to all stakeholders.
Atack, Iain (1999) “Four Criteria of Development NGO Legitimacy,” World Development, 27 (5) 855-864.
Burnside, Craig and Dollar, David (2000) “Aid, Policies, and Growth,” The American Economic Review, 90 (4), 847-868.
Doornbos, Martin (2001) “ ‘Good Governance:’ the Decline of a Metaphor,” Journal of Development Studies, 37 (6), 93-108.
Easterly, William (2002) “The Cartel of Good Intentions: The Problem of Bureaucracy in Foreign Aid,” Policy Reform, 5 (4), 223-250.
Easterly, William (2003) “Can Foreign Aid Buy Growth?” Journal of Economic Perspectives, 17 (3), 23-48.
Edwards, Michael and Hulme, David (1996) “Too Close for Comfort? The Impact of Official Aid on Nongovernmental Organizations,” World Development, 6, 961-973.
Guillaumont, Patrick and Chauvet, Lisa (2001) “Aid and Performance: A Reassessment,” Journal of Development Studies, 37 (6), 66-92.
Hermes, Niels and Lensink, Robert (2001) “Changing the Conditions for Development Aid” Journal of Development Studies, 37 (6), 1-16.
InterAction (2002) InterAction Member Profiles 2000-2001, American Council for Voluntary International Action.
Lancaster, Carol (2000) Transforming Foreign Aid, Washington DC: Institute for International Economics.
Neumeyer (2003) The Pattern of Aid Giving: The Impact of Good Governance on Development Assistance, Routledge Studies in Development Economics, London: Routledge, Taylor and Francis Group.
Scott, C.D. and Hopkins, R. (1999) “The Economics of Non-Governmental Organisations,” The Development Economics Discussion Paper Series: 15, The London School of Economics.
Trumball, William N. and Wall, Howard J. (1994) “Estimating Aid Allocation” The Economic Journal, 104 (425), 876-882.
United Nations Development Program (UNDP) (2002) Human Development Reports.
U.S. Agency for International Development (2002) “The Full Measure of Foreign Aid,” Foreign Aid in the National Interest: Promoting Freedom, Security and Opportunity.
U.S. Agency for International Development (2004) “US Foreign Aid: Meeting the Challenges of the Twenty-First Century (White Paper)”.
Williamson, John (1990) “What Washington Means by Policy Reform,” in John Williamson, ed., Latin American Adjustment: How Much Has Happened? Washington D.C.: Institute for International Economics.
Williamson, John (2000) “What Should the World Bank Think About the Washington Consensus?” The World Bank Research Observer, 15 (2), 251-64.
Wooldridge, Jeffrey M. (2002) Econometric Analysis of Cross Section and Panel Data. Cambridge MA: MIT Press.
World Bank (2003) World Development Indicators.