Receive Grants or Perish? The Survival Prospects of Ugandan Non-Governmental Organisations

Abstract

This study examines survival patterns in a large, representative panel of Ugandan non-governmental organisations (NGOs) between 2002 and 2008. It finds no evidence that more effective or more altruistic NGOs have a greater likelihood of survival. The main determinant of survival appears to be access to grants, and NGOs without grants struggle to survive. An investigation of the grant allocation mechanism suggests that effectiveness does not increase an NGO's likelihood of receiving a grant. Grant allocation appears to be neither fair nor effective, but rather to be awarded on the basis of habit rather than merit: once a grant has been allocated there is a strong tendency for it to persist. The odds are stacked against small NGOs that have not previously received grants. A picture emerges of two parallel NGO worlds: one where revenues are small, variable and hard to come by and survival is not very likely, and the other where revenues are high, more stable and more accessible and survival is more likely. The study suggests it may be difficult for an NGO to move from the former to the latter.

Acknowledgements

This work was supported by the British Academy [50401], Commonwealth Secretariat [ZACA-2006-116], the South African National Research Foundation [SFH2005082600083] and the University of Nottingham [NRF4322].

We would like to thank Fred Matovu, Patrick Birungi and their team of enumerators for their diligent efforts in tracing the NGOs and administering the questionnaire. For useful comments we would like to thank D. Bonbright, R. Burger, A. Dreher, M. Gugerty, O. Morrissey, H. Öhler, L. Salamon, W. Sokolowski and F. Teal. We are also grateful to participants of the following conferences and seminars who gave us useful feedback on earlier drafts: the International Society for Third Sector Research, Stellenbosch; Centre for the Study of African Economies, University Oxford; and the Association for Research of Nonprofit Organizations and Voluntary Action, Washington DC; and seminar presentations at the University of Goettingen; University of Nottingham; and the Centre for Civil Society Studies at Johns Hopkins University.

Notes

An Online Appendix is available for this article which can be accessed via the online version of the journal available at http://dx.doi.org/10.1080/00220388.2012.754430

1. This is presumably at least partly because appropriate data are lacking. Existing work on the survival of non-profit organisations (e.g. Tuckman & Chang, 1991; Bielefield, 1994; Fernandez, 2008; Wollebaek, 2009) relates mostly to developed countries.

2. See Callamard (2006); McGann and Johnstone (2006); Burger and Owens (2010).

3. For a full account of the original survey see Barr, Fafchamps, and Owens (2005).

4. Since 1989, the NGO Registration Statute requires all NGOs in Uganda to register their organisation with the NGO Registration Board in the Ministry of Internal Affairs (MIA). The initial intention was to draw the sample from this list, but it became evident that a number of NGOs in the register had registration certificates that had elapsed. We began by verifying the official register and found that only 15–30 per cent of the NGOs on the register were actually in operation. In some cases it was known that the NGO was still in operation but simply had not renewed its certification; in other cases the NGO was known to have ceased operation; but in other cases it was not clear that the NGO had ever begun operating. There was also concern that many NGOs had changed their core functions and location without informing the Registration Board.

5. For a detailed discussion of the community focus group interviews, see Barr and Fafchamps (2006).

6. In some cases, the NGO was not known to the focus group and therefore could not be matched back to the intended NGO. In these cases, the focus group assessed an alternative NGO, which was often not part of the sample for the first module and subsequently had to be dropped from the analysis.

7. Investigation of bias introduced by the loss of observations due to matching is available on request. The analysis of a number of key observables such as income and expenditure shows that there is little evidence of such bias. This does not, however, exclude the possibility that there may be bias as a result of unobservable factors.

8. Responsiveness is measured on a five-point Likert scale as a consensus response from the community focus group to demonstrate their agreement with the statement that the NGO is always quick to respond when inhabitants of this parish or the parish as a whole ask for help.

9. Accessibility is captured in three categories, i.e. (i) the NGO has an office in the community, (ii) the NGO does not have an office in the community, but it visits the community regularly; (iii) the NGO does not have an office in the community and community members have to visit the NGO.

10. The community participation variable is constructed based on two variables, namely (i) Does the NGO ask the community about their needs before they initiate a project? (a binary variable compiled from three underlying questions) and (ii) Does the NGO ask the community for feedback after concluding a project? (a binary variable constructed from three questions to the focus group).

11. Decisions had to be made for exceptional cases. Five NGOs that had become private enterprises were removed from the baseline sample as being rare and special cases comparable neither to mortality nor to survival. Eight were classified as closed: four that still existed in name but had stopped operating, and four whose leaders had moved and which had either ceased to exist or undergone such radical transformation (e.g. geographical focus, activities and staff) that they could be regarded as entirely new NGOs.

12. Other measures of staff that included volunteer as well as salaried staff gave similar results.

13. Reinikka and Svensson's (2011) research on Ugandan not-for-profit health facilities finds that altruism (measured by religion) has an effect on service delivery that matters quantitatively. We therefore include this variable in the specification.

14. Although not shown in the table, the result remains when we include these variables in other specifications, and when they enter these models individually.

15. For robustness, we estimate both specifications using an instrumental variable linear regression which reports the same signs and levels of significance.

16. These test statistics are obtained from the 2SLS estimation, assuming a linear model. We are only interested in the first stage results from this specification, which makes this approach valid.

17. The Moreira p-value correction only exists for linear models. The values reported in Table 3 are thus based on linear implementation of the regression.

18. Not reported here, but available on request for our larger sample, we exclude our perceived altruism and effectiveness variables. They remain insignificant. The coefficients of the funding variables do not change from those reported in Table 3.

19. Columns 1 to 3 use probits. However, with such a specification there is reason to be concerned about selection bias, because NGOs that did not survive are missing from our 2008 grant variable and the unobservable factors that are important for survival may also be important for grant allocation. We therefore also estimate a Heckman probit in columns 4 and 5. Table A1 [Online Appendix] shows that the hypothesis of independent equations cannot be rejected, thus the discussion below concentrates on the first three models. The patterns of significance are reasonably stable, especially when considering the drop in the sample size when including our altruism and perceived effectiveness variables.