Neither a pure Readymade strategy nor a pure Custommade strategy fully utilizes the capabilities of the digital age. In the future we are going to create hybrids.
In the introduction, I contrasted the Readymade style of Marcel Duchamp with the Custommade style of Michelangelo. This contrast also captures a difference between data scientists, who tend to work with Readymades, and social scientists, who tend to work with Custommades. In the future, however, I expect that we will see more hybrids because each of these pure approaches are limited. Researchers who only want to use Readymades are going to struggle because there are not many beautiful Readymades in the world. Thus, researchers sticking to this pure style are either going to sacrifice quality by using ugly Readymades, or they are going to spend lots of time looking for the perfect urinal. Researchers who only want to use Custommades, on the other hand, are going to sacrifice scale. Hybrid approaches, however, can combine the scale that comes with Readymades with the tight fit between question and data that comes from Custommades.
We saw examples of these hybrids in each of the four empirical chapters. In Chapter 2, we saw how Google Flu Trends combines an always-on big data system (search queries) with a probability-based traditional measurement system (the CDC influenza surveillance system) to produce faster estimates (Ginsberg et al. 2009). In Chapter 3, we saw how Stephen Ansolabehere and Eitan Hersh (2012) combined custom-made survey data with ready-made government administrative data in order to learn more about characteristics of the people who actually vote. In Chapter 4, we saw how the OPower experiments that combine the ready-made electricity measurement infrastructure with a custom-made treatment to study the effects of social norms on behavior at a massive scale (Allcott 2015). Finally, in Chapter 5, I told you about how Kenneth Benoit and colleagues (2015) applied a custom-made crowd-coding process to a ready-made set of manifestos created by political parties in order to create data that researchers can use to study elections and the dynamics of policy debates.
These four examples all show that a powerful strategy in the future will be to enrich big data sources, which are not collected for research, with additional information that makes them more suitable for research (Groves 2011). Whether it starts with the Custommade or the Readymade, this hybrid style holds great promise for many research problems.
