Using Divide-and-Conquer to Improve the Effectiveness of Tax Collection Under Limited Enforcement Capacity

How to cite

Kapon, Samuel; Chassang, Sylvain; Carpio, Lucia Del. 2021. "Using Divide-and-Conquer to Improve the Effectiveness of Tax Collection Under Limited Enforcement Capacity." AEA RCT Registry. AEARCTR-0007305. https://doi.org/10.1257/rct.7305-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Samuel Kapon* , Sylvain Chassang , Lucia Del Carpio

* Primary investigator

Registry ID
AEARCTR-0007305
DOI
https://doi.org/10.1257/rct.7305-1.0
First registered
Last updated
Status
In development
Trial dates
2020-12-15 – 2021-09-30
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/7305

Abstract

We partner with a municipal district in Lima, Peru to
implement a randomized controlled evaluation of divide-and-conquer
mechanisms for tax collection on a sample of roughly 15,000 taxpayers
delinquent in paying property taxes.

Intervention

Tax payers will be assigned a numeric rank and will go through the tax collection process in order of their rank (i.e. rank 1 will always be prioritized over rank 2 when allocating processing capacity). Taxpayers will have the opportunity to settle and lock in the current tax-amount they owe. Taxpayers can either pay in full, or enter into a compromise, at a specified interest rate.

We will partition tax payers into 3 groups based on their numeric rank and inform them of their group at regular intervals. Each taxpayer is informed of a number of weeks, based on their group, after which collection procedures will be initiated immediately (which includes penalties, fees and other allowances for the cost of collection). Collection can be started at any time before this date without prior notice.

Experimental design

We partition the sample into a control and treatment group, each with approximately half the sample. The main difference between the control group and the treatment is that the control group will not be notified of their numeric rank and priority group (and so will receive no information related to these groupings). Information in the treatment will be communicated through a card as well as with phone calls.

Ranks will be assigned by the research team using a score corresponding to: [Expected Payment] / [Probability Accepts to Pay].
Expected payment and probability of payment will be estimated using previous settlement data (including tax payer characteristics, and the stage in the process at which they settle).

Design details

Randomization method
Randomization done privately
Randomization unit
Individual taxpayers
Clusters
15,000 taxpayers
Observations
15,000 taxpayers
Arms
~7500 for treatment and the same for control
Primary outcomes
Collection Amounts Time and Resources Spent on the Collection Process

Cited by

No citation data available for this registration yet.

Keywords