Annex A. Methodology
The OECD’s first edition of Private Philanthropy for Development, published in 2018, provided data and a unique perspective on contributions to development financed by 143 large philanthropic donors over the period 2013-15 (OECD, 2018[1]). This second edition goes well beyond the first in both ambition and scope. It notably includes a more comprehensive picture of philanthropy by providing information and analysis on both cross-border financing and domestic philanthropy in developing economies, particularly in India and the People’s Republic of China (hereafter “China”). This edition surveys 205 philanthropic organisations and covers the period 2016-19. Geographical coverage is broadened with information from donors in emerging markets. This edition also introduces higher transparency standards for grants and project-level information – standards that are applicable worldwide.
The results in this report draw from a global survey conducted by the OECD between October 2020 and September 2021. The survey aimed to measure specific grants, donations and projects, as well as several organisational aspects of large philanthropic donors. To this end, two instruments were used:
A financial survey. This survey collected project- or grant-level data from each participant organisation, including project description, annual financing provided by the organisation, geographical allocation, financial instrument used, channels of delivery and modality of giving. The format and definitions used in the questionnaire were compliant with OECD-DAC statistical standards and classifications, which make the data comparable with official development assistance (ODA). Foundations that did not wish to disclose grantee-level information were able to sign a non-disclosure agreement with the OECD so that only aggregated, anonymised information about their donations would be made public.
An organisational survey. This survey was deployed through an online questionnaire that included eight thematic modules.
1. Financial instruments, income and non-financial support: financial instruments used by the foundation, all sources of income, how the foundation manages its endowment, investment strategies and the type of non-financial support provided to grantees.
2. Targeting demographics: whether the foundation targets its grants and projects according to age, gender and social or economic vulnerabilities.
3. Advocacy: the strategies and barriers faced by foundations when carrying out advocacy to change policies, practices or attitudes.
4. Collaboratives and financial sustainability of grantees: measuring how foundations co-finance initiatives through private donor collaboratives, and how they work on the long-term financial sustainability of their grantees.
5. Learning and information practices: measuring the methods foundations employ to evaluate their donations, grantees and projects, as well as all information disclosed voluntarily to assess their level of transparency.
6. Cross-cutting themes (gender equality and climate change): measuring whether foundations are including gender equality and climate change as explicit objectives of their donations and projects.
7. COVID-19 response: measuring qualitatively how foundations adapted in the short term to the COVID-19 pandemic.
8. Looking ahead to 2030: measuring expectations concerning the financial contributions that foundations plan to deliver over the next decade, with a focus on different thematic areas, the Sustainable Development Goals (SDGs) and geographies.
The OECD invited more than 400 organisations worldwide to participate in the survey. The sample targeted the largest organisations according to their annual spending in grant making or project financing, based on previous OECD research and consultations with multiple regional networks of philanthropic organisations. The targeted population consisted of foundations carrying out cross-border operations worth more than USD 5 million per year, and organisations operating domestically with spending above USD 2 million per year, or the equivalent in the local currency based on annual nominal exchange rates (Annex C). The survey was carried out in close collaboration with the Development Co-operation Directorate for the sample of foundations engaging on regular CRS data reporting as of October 2020.
This report summarises data collected for the period 2016-19 from 205 organisations based in 32 countries, including organisational data for 103 of them (Annex B). The resulting database includes over 45 000 distinct activities for the period. It was assembled using several sources of information.
OECD Creditor Reporting System: 45 of the largest foundations that annually report on their individual spending and are included in the OECD Creditor Reporting System, as of 30 June 2021, accessible at https://stats.oecd.org/Index.aspx?DataSetCode=DV_DCD_PPFD. Data sourced from the CRS represented 72% of the data pool used for this report in terms of financial volumen (gross disbursements), and 50% in terms of number of activities.
OECD financial survey: 67 foundations replied directly to the project or grant questionnaire. Of these, 9 requested that their grantee or project level data be anonymised.
OECD organisational survey: responses to this survey from 103 foundations are included in the analysis. Additional organisations replied to this survey, but their responses were excluded as they did not submit information from the financial survey.
Data collected from secondary sources by the OECD Centre on Philanthropy: for 99 foundations, the OECD recovered publicly available information from multiple sources, depending on the country where each organisation is based.
Foundations from the United States: for 19 foundations, Form 990-PF filings were used to estimate all charitable funding, and then to identify grants that correspond to the definition of private philanthropy for development. The forms were retrieved from the website of the Internal Revenue Service (IRS), according to the availability of filings as of 30 June 2021 (https://www.irs.gov/charities-non-profits/form-990-series-downloads).
Foundations from the United Kingdom: for 4 organisations, data available as of 30 June 2021 were retrieved from the GrantNav platform of 360Giving, a charity that helps organisations publish open, standardised grants data (https://www.threesixtygiving.org/).
Corporations and foundations from India: for 31 organisations, information was retrieved from the Indian Ministry of Corporate Affairs’ National Portal of Corporate Social Responsibility (CSR) as of 30 June 2021 (https://csr.gov.in/). Given that CSR provisions are allocated predominately in social sectors and financed by private corporations, they are included alongside more traditional forms of philanthropy from individual donors and foundations (OECD, 2019[2]). In addition, for all organisations that form Tata Trusts, information was collected based on the organisations’ published annual reports for the period 2016-19 (https://www.tatatrusts.org/about-tatatrusts/annualreports). Financial years are taken as the period between April and December of every year.
Foundations from China: for 45 organisations, information was compiled from each organisation’s publicly available information as of 30 June 2021, or from the People’s Republic of China Non-Profit Organisations (NPO) portal (https://cszg.mca.gov.cn/platform/login.html). The consolidated data include donations and project financing that surpassed CNY 1 million (Yuan renminbi). For the purpose of this data collection, information was considered from different types of NPOs, like civil non-enterprise institutions, social service organisations and private foundations.
What is private philanthropy for development?
The following definition was used to identify which grants, projects and activities carried out by philanthropic organisations would be included in this report:
Private philanthropy for development refers to transactions from the private or non-profit sector having the promotion of the economic development and welfare of developing countries as their main objective, and which originate from foundations’ own sources, notably: endowments; donations from companies or individuals (including crowdfunding); legacies; and income from royalties; investments (including government securities); dividends; lotteries and similar. In addition, private philanthropy for development also includes financing towards basic or applied research that directly benefits developing countries, or indirectly benefits developing countries through global public goods.
Activities not considered to constitute private philanthropy for development include:
volunteer activities of company employees that do not represent an explicit accountable expenditure on behalf of the foundation or company;
activities solely financed by the public sector, through transfers, procurement or other mechanisms;
charitable giving to religious institutions not aimed at supporting development or improving welfare.
What is official development assistance?
Official development assistance1 represents flows to countries and territories on the DAC List of ODA Recipients and to multilateral development institutions when the flows are:
Eligibility criteria for inclusion in the report
The private philanthropy financing included in this report was directed towards developing countries and territories based on the DAC List of ODA Recipients in 2020.
Cross-border vs. domestic financing
This second edition of Private Philanthropy for Development has a larger scope than the first. It is more inclusive of global philanthropy and significantly increases the data collected from large foundations and other organisations based in developing countries. These organisations operate for the most part only in the country where they are based; whether the financing corresponds to cross-border financing or domestic giving can nonetheless be determined by the data.
Defining the geographic origin of philanthropic financing
The geographic origin of private philanthropy financing follows the residence principle of an organisation’s headquarters.2 For instance, outflows from a foundation operating from a local office in a developing country, but with the main office in Paris, are considered as originating from France.
Classifying geographic origin at the country and regional levels
Except for data on the OECD CRS, the activities, projects and grants identified for this report were classified in one of three categories:
Known region or country-level allocation: all activities for which the foundation knows where the resources were allocated at a country or regional level.
Known countries but unknown distribution: all activities carried out in multiple countries for which foundations knew the countries but were uncertain about the exact share of funding that went to each individual country. For these activities, the OECD prorated the resources at the grant level in equal proportions among all countries identified by the foundation.
Global or non-localisable financing: activities that do not have a geographical dimension, such as basic research carried out in universities, funding to international organisations, or donations for which the organisation does not know the region or country of disbursement.
Thematic classifications
For all data collected from the financial survey and secondary sources, thematic classifications followed the OECD DAC Purpose Codes on sector classifications.3 The thematic classifications (sector, purpose, cross-cutting themes, etc.) were carried out using a text-based machine learning algorithm. In order to classify grants, grantees and projects, a supervised machine learning algorithm, Xtreme Gradient Boosting (XGBoost) (Chen and Guestrin, 2016[3]) was used over text included in grant/project descriptions. XGBoost is extensively used for classification tasks, and was implemented using the R interface.4
Deflators and currency conversions
Operations denominated in currencies other than the United States dollar (USD) were converted using nominal end-of-year exchange rates; country annual Consumer Price Indexes (CPI) were used to deflate financing from all organisations (Annex C). Unless otherwise stated, all monetary figures in the report are shown in constant 2019 USD.
References
[4] Bergstra, J. and Y. Bengio (2012), “Random search for hyper-parameter optimization”, Journal of machine learning research, Vol. 13/2, http://dl.acm.org/citation.cfm?id=2188395.
[3] Chen, T. and C. Guestrin (2016), “Xgboost: A scalable tree boosting system”, In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 785-794, http://arxiv.org/abs/1603.02754.
[5] Fushiki, T. (2011), “Estimation of prediction error by using K-fold cross-validation”, Statistics and Computing, Vol. 21/2, pp. 137-146, https://doi.org/10.1007/s11222-009-9153-8.
[2] OECD (2019), India’s Private Giving: Unpacking Domestic Philanthropy and Corporate Social Responsibility, OECD Development Centre, https://www.oecd.org/development/philanthropy-centre/researchprojects/OECD_India_Private_Giving_2019.pdf.
[1] OECD (2018), Private Philanthropy for Development, The Development Dimension, OECD Publishing, Paris, https://doi.org/10.1787/9789264085190-en.
[6] Sokolova, M., N. Japkowicz and S. Szpakowicz (2006), “Beyond accuracy, F-score and ROC: a family of discriminant measures for performance evaluation”, In Australasian joint conference on artificial intelligence, pp. 1015-1021, https://doi.org/10.1007/11941439_114.
Notes
← 1. See https://www.oecd.org/dac/financing-sustainable-development/development-finance-standards/What-is-ODA.pdf.
← 2. In this context, “residence” is not based on nationality or legal criteria, but on the transactor’s centre of economic interest: an institutional unit has a centre of economic interest and is a resident unit of a country when, from some location (dwelling, place of production or other premises) within the economic territory of the country, the unit engages and intends to continue engaging (indefinitely or for a finite period) in economic activities and transactions on a significant scale (one year or more may be used as a guideline, but not as an inflexible rule).
← 3. See https://www.oecd.org/dac/financing-sustainable-development/development-finance-standards/purposecodessectorclassification.htm.
← 4. The pre-processing step followed a standard Natural Language Processing pipeline. Hyper-Parameter optimisation was done using a random search approach (Bergstra and Bengio, 2012[4]), and in order to validate results, a k-fold cross validation (Fushiki, 2011[5]) approach was used. The metric performance relied on F1-score (Sokolova, Japkowicz and Szpakowicz, 2006[6]).