Microfinance in India: Analysis of Microfinance Information Exchange (MIX) data

Microfinance is an extremely ambitious (and challenging) idea to achieve financial inclusion. It is also an exciting and very interesting concept. Microfinance Information Exchange (MIX) is an organization that posts free global Microfinance data on their website, which includes Microfinance Institution (MFI) profiles and basic portfolio characteristics.

Basic MIX data is available at http://mixmarket.org/profiles-reports (top right hand corner).  For India, it has one data point for every year from 1995 to 2013, which includes performance measures such as gross loan portfolio, average outstanding balance, portfolio at risk over 30 days etc. In addition, it has MFI characteristics such as regulated or unregulated, age of the MFI, percentage of female staff. 

As I was browsing through the data, I was keeping an eye for MFIs in India and their portfolio over the last few years. Microfinance industry in India grew significantly from about 2006, as both for profit and non-profit MFIs started catering to the significant unbanked population. It was deemed as the panacea to all ills, but somewhere down the line, profit-oriented MFIs resorted to aggressive growth strategies, partly due to pressure from investors. Consequently, when customers fell back on their payments, these lenders also resorted to aggressive and unethical collection practices according to this news item.

Knowing that there are several MFI’s catering to the unbanked population in India, I decided to look at the number of loans (to collect) per loan officer from 2003 to 2013. This number more than doubled from 2009 to 2011 (going from 562 to 1433), during the peak of the crisis. Source: MIX Market

In terms of gross loan portfolio, the Microfinance sector continued to grow through 2010/ 2011, hitting a peak of $5.5 billion, with an active loan base of 82 million loans. Source: MIX Market

Surprisingly, there is only a weak correlation (0.3) between number of loans per loan officer and percentage of portfolio over ninety days due. Source: MIX Market

One possible reason could be that in India, the group loan methodology (introduced by Grameen Bank) is still very prevalent. This system emphasizes the importance of punctual payments and the group members feel obligated to repay the loan if one of their group members is unable to do so. So even though the number of loans per loan officer was high, this self-driven discipline of the group could have salvaged the MFIs from having highly delinquent accounts. 

NYC Parking Violations

Few months ago the City of New York posted Parking Violations Data. The data had over six million observations with several elements like Vehicle details including Plate ID, Plate type, Registration state, details of when the ticket was issued, where it was issued, Violation county and precinct. Because I live in Brooklyn, I was naturally curious to see how parking violations compared amongst boroughs and where Brooklyn stood in particular. On a side note, I was able to find an entry for the parking ticket that I got in September 2013 J.

The data was available in the form of a 1.3 GB CSV file. I tried reading the file directly into R, but for some reason it was unable to read beyond 450,000 rows. I processed the file separately to retain columns of interest and was able to import the whole file. The Issue Date on this data ran from 1970 to way beyond 2014, but I have only analyzed 2013 data here, which was about 63% of the total. A few entries had incorrect Borough ID’s and those have been excluded here (keeping only BX, K, NY, Q, and R). Since there were around 99 different violation codes, I have grouped similar ones (for e.g. Bike lane, crosswalk and sidewalk related violations have been grouped together, similarly, all registration related offences have been grouped together) to be able to see the broad violation categories. I have omitted Violations with negligible number of observations.   

To assess the total amount paid in fines ($) at borough level, I obtained the parking fine amount by violation code from NYC Department of Finance website. The fines are categorized by location as “Manhattan below 96”, which means areas below 96^th St in Manhattan and “All other areas”. Because there is no significant difference between these categories, for the purposes of this analysis, I have used “All other areas”. This data was matched to the parking violation data using violation code.

Percent Violations by County and Category: Almost 13% of violations pertained to ‘No Stopping Zones’ in Manhattan. About 8% offenders were parked in ‘No Parking Zones’ in Brooklyn and over 9% of them were fined for ‘Failure to Display Muni Meter Receipt’ in Manhattan. Two of the top three offending populations are in Manhattan, and the third one is in Brooklyn. 

Out of state offenders (excluding NY registration vehicles): 42% parking violations had NJ registration, 11% from PA and 7% from CT. Surprisingly, FL vehicles account for 5.5% of parking violations. No wonder I see so many FL registered vehicles in my neighborhood. 

Percentage violations by county and fines paid: Circles denote percentage violations and bars denote average fine paid. Borough of Manhattan collected $136 mm in fines in 2013. While the percent violations in Manhattan were 42% of total, the fines they collected were 47% of total. Average fine in Manhattan was $80. Brooklyn, Queens, and Bronx were all around the $65 mark, while Staten Island averaged at $74. 

There are many interesting elements in this data. For e.g. as next steps I plan to study seasonality, if any, and the effect of weather on parking violations by looking at temperature, precipitation etc. Violations by type of vehicle (Agricultural, tractor, motorcycle etc.) would also make interesting analysis.