Syndicate content

A new overview of firm experiments

David McKenzie's picture

A number of recent field experiments have been conducted within firms and across firms. In another paper in what is shaping up to be an excellent forthcoming Journal of Economic Perspectives symposia on experiments, Oriana Bandiera,Iwan Barankay and Imran Rasul give their take of what we have learned from firm experiments so far, and their ideas on further research directions.

Field experiments within firms

In these experiments, the unit of the observation is typically the employee (or sometimes groups or divisions of employees) within firms. These experiments are designed to help firms figure out how to best motivate their employees, stopping workers shirking and increasing labor productivity. Examples include:

·         Monitoring: an experiment with workers soliciting for donations randomized the extent to which promises of donations were audited. They found considerable heterogeneity in responses, with employees who perceived the employer as unfair and uncaring were more likely to cheat, while the majority of workers didn’t shirk.

·         Pay for performance incentive schemes such as comparisons of fixed wages versus piece rate wages (piece rates increase productivity of existing workers and also lead more productive workers to apply for jobs); relative incentive schemes where workers are paid according to how their performance compares to those of others (this reduced productivity in their experiments, especially in small groups, as workers internalized the effect their production levels had on pay of others); and team incentive experiments designed to see the role of social connections.

·         Non-monetary incentives: They give the example of new work by Ashraf et al., (2011) [not yet online], who randomize 800 community agents hired to sell condoms different monetary and non-monetary rewards treatments. Agents who are assigned to the non-monetary rewards treatment, namely stars for performance plus a public ceremony for top performers, sell twice as many condoms as agents who are offered a financial margin on each pack sold.

Most of these experiments have been done in developed countries to date (the Ashraf study being one exception). Another example of such experiments underway in developing countries, such as one Nick Bloom is conducting in China to examine the impact on productivity of allowing workers to work from home.

Field Experiments across firms

These experiments are typically designed to uncover what the binding constraints are to firm growth by exogenously varying the availability of key inputs. These experiments include the experiments I’ve been involved in which give grants to firms to measure the return to capital for microenterprises;  and experiments on business training and management training – which we will be discussing more of in upcoming blog posts.

Other issues

The paper then discusses implementation issues and ethics issues. On implementation issues, one point the authors note is that when it comes to implementing within firm interventions, an important issue contains spillovers in information between treatment and control groups. They give the example of a study by Shi (2010) which switched some, but not all, workers in a firm to piece rates instead of fixed pay. Shi reports that workers in the control group became informed of the existence of a treatment group and were not pleased by it – which may have then reduced their productivity leading to an overestimate of the treatment effect although the data do not allow testing this.

A final interesting point made by the researchers is the following “Indeed one puzzling finding is that almost all the field experiments reviewed have brought large benefits to the firm. In part this is driven by endogenous experimental selection: firms would not agree to implement experiments that are expected to have detrimental consequences. Yet, the fact that in so many cases researchers have managed to increase profits appears at odds with the common assumption that firms are pressured by competitive forces to make at least close to optimal choices.” They then suggest that time constraints faced by firm owners maybe one such reason, along with delegation issues and the possibility that firm owners don’t necessarily try to maximize profits ….all in all, leading to an interesting future research agenda.


For further reading on what I think we have learnt and should be learning from research in the private sector in developing countries, see this paper (open link here) in the World Bank Research Observer.