Field Experimentation Research

Field experiments are randomized interventions that take place in real-world settings. Although the term field experimentation has occasionally been used to refer to nonrandomized interventions, current usage restricts the domain of field experimentation to studies in which the units of analysis are randomly assigned to treatment and control conditions. Examples of randomized interventions may be found in a variety of social science disciplines. Working in conjunction with both state and nongovernmental organizations, scholars have randomized interventions such as job training programs, income tax schedules, health insurance plans, sex education programs, voter mobilization campaigns, school voucher programs, public-housing assignments, and narcotics enforcement.

Well-executed field experiments combine the strengths of randomized designs with the external validity of field research. While a laboratory study might examine the effects of political advertising on vote choice by assigning treatment and control groups to different advertising stimuli within an artificial setting and gauging their candidate preferences, field experiments randomly manipulate the content and timing of actual advertisement campaigns and attempt to link these varied interventions to voting patterns. Both types of studies use randomization, but the latter has the advantage of linking cause and effect in terms that have direct real-world applicability. Often, randomized interventions are unobtrusive, in the sense that the subjects remain unaware that the intervention they may receive has been assigned as part of a research effort.

The external validity of field experimentation is especially important when researchers study treatments whose effects decay over time. For example, laboratory studies that establish that political advertisements sway voters' candidate preferences within the context of a 1-hour experiment leave unanswered the question of whether in the outside world these advertisements' effects dissipate before election day. Laboratory studies serve to demonstrate that interventions are capable of exerting influence on social outcomes; field experimentation is then required to gauge whether these effects persist amid myriad competing messages characteristic of real political settings.

More marked than the contrast between field and laboratory experimentation is the difference between field experimentation and other types of fieldwork, such as ethnographic or participant-observer research. Ethnographic studies provide valuable descriptive information about cultural, economic, and political practices and how individuals understand them, but they rarely set out to trace the consequences of a randomized intervention. Participants in organizations or agencies sometimes craft randomized field experiments, but this work is typically quite different in character from conventional participant-observer research, which tends to involve a participant's reflections on a nonexperimental event. What distinguishes field experimentation is that it derives its insights from randomization procedures that ensure the comparability of treatment and control groups.

Because field experiments take place in natural settings, their features are often shaped by practical constraints. Seldom can researchers randomly assign students to different curricula; as a result, educational experiments often use the classroom or the school as the unit of assignment. Scholars seeking to assess whether voter mobilization campaigns cause citizens to vote may assign individuals to treatment and control conditions, but those seeking to study whether a campaign alters citizens' preferences for particular candidates may need to randomize at the level of the voting precinct, given the constraints of the secret ballot. Field experimentation requires a certain degree of agility on the part of the researcher, who must choose a unit of assignment that is both informative and practicable.

...