Thursday, November 08, 2012

Predictive analytics and philanthropy

Political and media junkies agree that one man, whose name appeared on no ballot, really won the election on Tuesday. Nate Silver, big data political analyst, blogger, and New York Times columnist. Silver's accomplishment? Using aggregate analysis of polling data Silver was able to correctly predict the outcome of the election in all 50 states. Watch Jon Stewart and Silver discuss this accomplishment and its role in the "defense of arithmetic."

In this age of big data and increasing foundation transparency, who will be the first big philanthropist to put predictive analysis to the test in the social sector? Silver's analytic model relies on calculating the probability that each candidate will win based on the results of numerous state and national polls. The model is inherently dependent on the quality of each of those polls but it also benefits from the sheer number of polls done during a Presidential election.

As individual foundations share more of their information in more useful ways, IssueLab catalogues and organizes the research that foundations and nonprofits do, independent analysts like GiveWell or users of the Nonprofit Finance Fund's FinancialSCAN tool, and local, state and federal agencies make more data available publicly we may be reaching the point where meta-analysis of others' research - whether program evaluations, strategy plans, or needs assessments - is possible and worth considering. We are slowly building the repository of raw material - data that can be shared, compiled, compared and considered in the aggregate - that makes such research feasible. Academic centers like MIT's Jameel Poverty Action Lab have been doing this with kind of analysis of existing randomized control trials.

Who will be the first philanthropic funder to take on the role of "aggregate analytic funder?" Seen as part of the overall social landscape, with foundations often claiming the role of R and D and innovator, it is a role ripe for the taking. Now looking for the "Nate Silver of the Social Sector"
Blatant Self Promotion:
On November 27 at 11 am PDT I'll be moderating an SSIR webinar on Data and Philanthropy with Darin McKeever of The Bill and Melinda Gates Foundation, Jake Porway of DataKind and June Wang of the The William and Flora Hewlett Foundation. We'll be looking at and talking about the emerging data backbone of information from foundations as well as the "why and how"? of using this information. Please register and join us.
Please note, I'm not contradicting all my previous statements about the role of intuition and passion in philanthropy. (In fact, for an individual donor who is passionate about math, big data, and analytics, this is indeed a perfect fit.) This foundation could be a "meta sense-maker" of sector research and data, help issue-specific funders learn which information works and where there are real knowledge gaps, and help others use and make sense of the abundance of research that exists, but is sorely underused. 


Ian David Moss said...

Hi Lucy,
I'm intensely interested in this issue (especially as it relates to my field, the arts) and in fact am working on a lecture for the U. Chicago Cultural Policy Center next week on infrastructure-building in the research field.

I'm curious if you've been following some of the conversations in the science field on meta-research and publication bias (summary here: GiveWell has proposed a new focus on "meta-research" to start to address some of these problems, but I wonder if we might be too quick to embrace big data without thinking more critically about both the underlying quality and relevance of that data. As much as I heart Nate Silver (and I really do), it seems to me that he is working in a pretty well-defined context with clear feedback loops, something which is a relative luxury in the kinds of fields we're most interested in.

Lucy Bernholz said...

Right on - publication bias would be a generous term for the current state of philanthropically funded research. But we're making progress, and the work is getting better. Efforts to use the information that exists - from across many sources - are what will, I believe, point out the quality issues. This can then begin the feedback loops, improvement, use and re-use that we need. It will be harder in the social sector and across foundations - there simply are not the regulatory, market, or outcome "ties" that provide more natural incentives for individual enterprises to improve their work.