Feeling Close and Giving More

How psychological distance affects DonorsChoose.org

February 21, 2015 by Mark

The distance between people cannot always be measured in meters or miles. Feeling close to a person entails more than simply the ability to reach out and touch them. Physical distance plays a role, to be sure, but so too do other forms of distance. Social distance matters, as anyone who has met a compatriot abroad can attest. Time, too, can separate or unite us - a friend's voicemail never brings quite the same connection as hearing their voice live.

These different types of distance share a common psychological basis deeper than mere metaphor. They shape our ability to imagine other places, times, and people in much the same way, a fact formalized in the construal level theory of Yaacov Trope and Nira Liberman (2010 review). When the mind travels across psychological distance - be it spatial, temporal, social, or counterfactual - our imagination becomes, among other things, fuzzier. Imagination, i.e. the simulation of events beyond our current experience, relies on recombining bits and pieces from memory in novel ways. However, when we try to imagine things far outside of our past experience, we run low on relevant memories to tap. As a result, our imagination for more psychologically distant events ends up less detailed and more abstract than our imagination for psychologically close equivalents.

Neuroscience research also supports the idea that the mind treats different types of distance in similar ways. Research by my former labmate Diana Tamir has shown that imagining distant events evokes more activity in the brain's default network whether the distance is spatial, temporal, social, or counterfactual. More recently, work by Carolyn Parkinson and her colleagues at Dartmouth demonstrated that the brain uses the same fine-grained neural code to track spatial, temporal, and social distance.

One possible consequence of our tendency to imagine psychologically distant events in less detail might be that we find it more difficult to imagine the needs of distant others. This in turn might make us less likely to help them. Lab research supports this hypothesis, demonstrating that temporal and social proximity between donor and recipient lead to greater willingness to donate. However, it can be hard to extrapolate the results of small laboratory studies to larger real world consequences. The analyses below tackle this problem directly by assessing the effects of temporal, spatial, and social distance on charitable donations to teachers on DonorsChoose.org

DonorsChoose is an educational charity website that applies the crowdfunding model of websites such as Kickstarter to the task of getting teachers the resources they need. Teachers post descriptions of the projects and materials they want, and donors from across the country respond to the call. The site has achieved considerable success in its mission, thanks in no small part to endorsements by celebrities such as Stephen Colbert and Oprah Winfrey. In the hopes of improving upon this success still further, DonorsChoose shares a wealth of data from its site.

In collaboration with my former labmates Diana Tamir and Jamil Zaki, I've recently been working on a Google-supported project aimed at discovering the psychological basis of successful online appeals. While ferreting out useful data for this project, I came across DonorsChoose. Perusing their data, I quickly realized that it had all the necessary ingredients for a case study on the effects of psychological distance on charitable giving. It currently includes detailed information on over 800,000 teacher projects and nearly 2.5 million individual donations.

In each case below, our measure of giving will be number of donors. Average amount donated might seem like a natural alternative, but it suffers from a pair of significant flaws. First, people's penchant for giving in nice "round" numbers (i.e. multiples of $5, $10, $25) leads to a truly ghastly distribution of donations. Neither mean, median, nor mode serves as a particularly useful measure of central tendency, and I thought it best not to digress into more complicated measures. Second, donation amounts are influenced by factors other than donors' altruistic intentions. In particular, donors' socioeconomic status likely plays a large role in how much they give, presenting a substantial confound. Thus focusing on donor number offers a much less complicated approach.

Temporal distance

We will first examine the effect of distance in time on donation rate. Unlike some events which elicit charitable giving, such as natural disasters, requests from teachers need not originate from an event at a particular point in time. Instead we treat the request itself as the key event, and measure temporal distance from the time they initially posted their project to DonorsChoose.org. The histogram below shows the distribution of donors over the first 30 days for projects that remained active at least 60 days total.

As you can clearly see, donations are by far the most frequent on the day the project is posted. They then fall off over the next 10-15 days before arriving at a relatively constant (low) rate. This result is just what we might expect on the basis of our understanding of psychological distance: as the event becomes more temporally distant, fewer and fewer people find the motivation to donate.

Spatial distance

DonorsChoose conveniently furnishes us with the longitude and latitude of each teacher's school. It doesn't provide quite the same level of detail when it comes to donors - probably a good thing for privacy's sake - but we can access many of their zip codes. Cross-referencing these with GPS coordinates using the zipcode package in R yields approximate longitudes and latitudes for each of our donors as well. Using the geosphere package, I could then calculate the Vincenty ellipsoid great circle distance - an accurate measure as distance "as the crow files" - between teachers and their respective donors. The distribution of donations as a function of spatial distance from the teachers is shown below.

Again, we see most donations originate very close to their recipients. After 100 miles or less, the giving rate evens out to a relatively low frequency with a couple of odd spikes that may correspond to the distances between major population centers. This graph nicely illustrates two different uses of DonorsChoose.org. First, it allows far away people to discover and donate to worthy projects thousands of miles distant. Second, it provides a highly convenient method of donation for people nearby.

Social distance

One thought that might have occurred to you when looking at the last graph was that it actually conflates two different types of psychological distance: spatial and social. Since many teachers are likely embedded in the communities in which they teach - and perhaps even grew up nearby - the people who are spatially close to them are also socially close to them. How then can we be certain that both spatial and social distance are contributing to the effect of distance observed above?

We could try to compare teacher and donor at the level of zip code demographics, but this would be an extremely coarse approach. Fortunately, another variable in the DonorsChoose dataset comes to our aid here: we know which teachers belong to the Teach for America program. Since this program often dispatches teachers to high-need areas far from home, it nicely disentangles social and spatial distance for us by separating teachers from their native social milieu. Below we see donor numbers grouped by spatial distance and Teach for America (TFA) status.

We use a 100 mile radius to bisect the distance distribution here because, in addition to being a nice round number, it nearly evenly divides the donations to the (much larger group of) non-TFA teachers. This allows us to clearly see that TFA teachers get a substantially larger proportion of their donations from distant donors than non-TFA teachers. The difference illustrates the effect of social distance, since it likely reflects the teachers' (socially close) friends and family giving money from (spatially) far away. However, note that even the TFA teachers get a hugely disproportionate number of their donations from nearby: if spatial distance didn't influence donations we would expect less than 10% of donations to come from within 100 miles, even in the most populous regions in the country. Thus we can safely conclude that both spatial and social proximity lead to greater giving.

Conclusion

Using data from DonorsChoose.org, we've seen how psychological distance can influence rates of charitable giving. Consistent with our scientific understanding of psychological distance, social, spatial, and temporal distance all pushed donors in the same direction. In addition to supporting psychological theory, these results might contain a practical lesson for teachers posting to DonorsChoose. Since spatial and temporal distance are inevitabilities, it suggests that emphasizing one's social similarity to a broad range of donors might be a profitable strategy to increase donation numbers.

I am sometimes asked why I use data science/big data/web-scraping techniques as a psychology researcher. The behavioral and neuroscientific tools at my disposal certainly give me much more precise means of assaying people's thoughts and neural computations, so why bother with the relatively crude measurements of the web?

Part of the answer is that I like playing with data, and the methods necessary for working with web data overlap quite a bit with the skills one develops running and analyzing fMRI experiments. However, the more substantive answer, in my opinion, is that the combination of neuroimaging and web research nicely bracket the spectrum of possible data sources. On one hand, fMRI lets me minutely examine the brains of small groups of individuals under tightly controlled laboratory conditions. On the other, web research allows me to study the real world behavior of very large groups of people. Demonstrating convergence between the two, such as we've seen here, strikes me as far more compelling evidence than either source in isolation.