Amy Heineike - Data Scientists at Work - page 238

Gutierrez: What was the first data set you remember working with?
Heineike: Perhaps not my very first data set, but in my economics work,
I think some of the first I played with were actually government data sets. So
you would log onto a portal and download economic indicators or census
data for Britain, or similar kinds of data. The data was very tightly controlled—
you actually had to get vetted by an agency in the UK that would decide
whether you could get access to the economic statistics. If you were approved,
then you could download the data in spreadsheets, depending on your query.
So you would only ever see bits and pieces of it. Those were probably the first
data sets I came in contact with.
These data sets were very much small data, in a way what we would maybe
now think of as older-school data. So this was data that was actually collected
for the purpose that people were using it for and it could it on a spreadsheet.
And not only that, it would have a bunch of statisticians arguing about exactly
how to present it before you even got to see it. This is very, very different
from the kinds of data that I'm using every day and the kinds of data that we
think of as “big data” now.
Gutierrez: When did you realize the power of data?
Heineike: I worked for a small company called Volterra Consulting in the
UK, which has done some really interesting analysis and some really interest-
ing, different kinds of work. There were several fascinating studies that that
company worked on while I was there. As one example, Paul Ormerod, who
was one of the two directors, published papers on his finding that if you look
at whether or not people have bank accounts in the UK, it's actually very hard
to predict whether someone will have a bank account based on just their
economic status, their income, or how much they earn. It's actually a very
important thing for us to understand, because people who don't have bank
accounts are typically excluded from financial systems and have a much harder
time plugging into how our economy works. Through his research, Paul found
that you could predict much more effectively whether or not someone has a
bank account if you have information about whether their friends have bank
accounts. As soon as you know that, then you realize this is obviously a kind
of a network effect. Bank accounts are one of those things where people need
to understand, “Oh, I should get one of these things. And this is how I would
use it, and I shouldn't be scared of getting a bank account. I can go to a bank.”
And so then it makes you realize their communities really matter for transmit-
ting this knowledge.
However, you just can't get this insight from high-level statistics. You need the
data on connections between people. To understand this issue, you've got to
figure out which data to look at. You might have to get it from surveys, or else
be very creative about where to hunt signals down. People's interactions drive
economic activity in profound ways.