« Back to Results
Marriott Marquis, San Diego Ballroom A
Hosted By:
National Association for Business Economics
meeting of a buyer and seller takes place in a structured environment, where some infor-
mation is made available to potential buyers (and possibly to sellers), prior to a decision
being made by buyers and sellers to engage in an exchange. We are interested in evaluating
interventions that change the structure in which these meetings or interactions take place.
Traditionally we design experiments where either buyers or sellers are randomly assigned
to the intervention. The concern is that there may be spillovers both within buyers and
within sellers. The current paper addresses settings where both types of spillovers are
present. We introduce a new class of randomized experiments where we randomize both
buyers and sellers separately, and then expose pairs of buyers and sellers to the intervention
depending on the outcomes of their separate randomizations. We show how this includes
buyer and seller experiments as special cases. We also show how such experiments can
shed light on spillovers that cannot be uncovered by buyer or seller experiments. We also
develop analytic methods for such experiments.
Tech Economics
Paper Session
Friday, Jan. 3, 2020 2:30 PM - 4:30 PM (PDT)
- Chair: Michael Luca, Harvard Business School
New Goods, Productivity and the Measurement of Inflation: Using Machine Learning to Improve Quality Adjustments
Abstract
New Goods, Productivity and the Measurement of Inflation: Using Machine Learning to Improve Quality AdjustmentsDouble Randomized Online Experiments
Abstract
Consider an online market place where a large number of buyers and sellers meet. Eachmeeting of a buyer and seller takes place in a structured environment, where some infor-
mation is made available to potential buyers (and possibly to sellers), prior to a decision
being made by buyers and sellers to engage in an exchange. We are interested in evaluating
interventions that change the structure in which these meetings or interactions take place.
Traditionally we design experiments where either buyers or sellers are randomly assigned
to the intervention. The concern is that there may be spillovers both within buyers and
within sellers. The current paper addresses settings where both types of spillovers are
present. We introduce a new class of randomized experiments where we randomize both
buyers and sellers separately, and then expose pairs of buyers and sellers to the intervention
depending on the outcomes of their separate randomizations. We show how this includes
buyer and seller experiments as special cases. We also show how such experiments can
shed light on spillovers that cannot be uncovered by buyer or seller experiments. We also
develop analytic methods for such experiments.
JEL Classifications
- L1 - Market Structure, Firm Strategy, and Market Performance