Methods
Summary
For initial investigations I plan to use ordered logit regressions with standard errors clustered by MFI, controlling for year fixed effects, from 2003 to 2015 for three components of PERFORMANCE individually and simultaneously . The dependent variable, TRANSPARENCY, can take on discrete values from zero to five and is assigned a score by www.mixmarket.org. The data is unbalanced panel data with about six thousand MFI-year observations.
| TRANSPARENCY is the level of self-disclosure by a MFI annually. Transparency is assigned a discrete value of 0, 1, 2, 3, 4, or 5, with 5 being most transparent. Source: MIX Market (mixmarket.org). EFFICIENCY is the negative of the first principal component score of the four annual MFI-level factors 1) operating expense to loan portfolio, 2) cost per borrower, 3) number of borrowers per staff member, and 4) number of borrowers per loan officer. Source: MIX Market (mixmarket.org). PORTFOLIO QUALITY is the negative of the first principal component score of four annual MFI-level factors: 1) loan portfolio at risk for 90 days or more, 2) the write-off ratio, 3) the negative of loan impairment provision divided by the gross loan portfolio, and 4) the negative of risk coverage, which is impairment loss allowance divided by the 30-day portfolio at risk. Source: MIX Market (mixmarket.org) PROFITABILITY is the first principal component score of three annual MFI-level factors, 1) return on equity, 2) return on assets, and 3) yield on the nominal gross portfolio. Source: MIX Market (mixmarket.org). PERFORMANCE is the first principal component score of standardized EFFICIENCY, PORTFOLIO QUALITY, and PROFITABILITY. Control variables would include.. The annual MFI-level ratio of debt to equity.
The natural logarithm of annual MFI-level assets in U.S. dollars.
A dummy variable that is 1 if MFI is for-profit and 0 otherwise.
The degree of outreach. Outreach is assigned a discrete scalar value of 0, 1, or 2, with 2 the highest outreach among borrowers.
A dummy variable that is assigned 1 if the MFI is 100% financially self-sufficient and 0 otherwise.
A dummy variable that is assigned a value of 1 if the is regulated. This field is marked as yes if the entity submits to some regulatory authority, whether a formal banking regulator or some other financial services regulator. This most often concerns entities that are listed as banks and non-bank financial institutions (NBFIs) but may also include credit unions/cooperatives and non-governmental organizations (NGOs) in some markets.
A dummy variable that is assigned 1 if MFI accept deposits and 0 otherwise.
A dummy variable that is assigned 1 if MFI is a part of a bank, credit union/cooperative, or rural bank and 0 if part of an NBFI or NGO.
A dummy variable equal to 1 if the borrowers of the MFI are a high-end or broad business and 0 if a low-end or small business.
The natural logarithm of the GNI per capita.
The first principal component score of the six worldwide governance indicators: voice and accountability, political stability, corruption control, rule of law, regulatory quality, and government efficiency.
A dummy variable that is assigned 1 if the country where the MFI is registered/headquartered is of English legal origin and 0 otherwise.
The first principal component score of six dummy variables that represent country-level banking regulation: 1) Are banks required to prepare consolidated accounts for accounting purposes? 2) Are applicable accounting standards for banks in your country prepared in accordance with U.S. GAAP at the individual bank level? 3) Are applicable accounting standards for banks in your country prepared in accordance with IFRS at the individual bank level? 4) Are all banks operating in your country (including foreign bank branches) required to make available to the public their annual financial statements at the individual bank level? 5) Do banks disclose to supervisors fully audited financial statements? 6) Are bank directors legally liable if the information disclosed is erroneous or misleading? Source: Bank Regulation and Supervision database, 2012, World Bank.
Social trust, calculated as the percentage of people who say people can be trusted minus the percentage of people who say that you can’t be too careful. Source: World Values Survey
|
Challenges
Due to nature of the panel data it is unlikely that Granger Causality testing can be done to determine the direction of causality. This is because there are not consistent lags present in the data for particular MFI-years. In other words, MFI data is spottily reported for various years in many instances. I hope to also include structural-equation modeling to make cautious inferences regarding direction of causality. In other words, for example, does transparency adjust to performance (disclose good news) or does transparency lead to better MFI performance?
Pre Analysis Plan
H1: MFI transparency is positively related to MFI performance
H2: MFI transparency is negatively related to national governance.
H2a: MFI transparency is positively related to national governance.
H3: MFI transparency is negatively related to for-profit status.
H4: MFI performance transparency is positively related to social trust.