Direct and Spillover Effects of a Paperless Billing Randomized Communication Campaign

This project explores the direct and network effects of communication messages on tax compliance and paperless billing sign-up. We conduct a randomized communication campaign in a large municipality of Argentina where neighbors are required to pay a monthly fee on their real estate, locally known as Tasa por Servicios Generales (TSG), which accounts for most of the local own revenues in Argentine municipalities. The municipality where the experiment takes place has recently enabled a paperless billing option for taxpayers to receive their monthly bill by e-mail in lieu of the regular paper bill. Our campaign consists of sending letters to randomly selected dwellings where we remind neighbors about the paperless option, how to sign up, and we also include information about the status of the account, due dates, past due debt, etc. Our goal is to study the effects on monthly payments and the sign-up to digital billing, and also to analyze whether the campaign creates spillover effects on neighbors that live nearby but that do not receive a letter.

2020-09-28

2020-10-07

Indicator for whether neighbors sign up to the paperless billing option; Monthly fee payments (intensive and extensive margins); delinquency indicator; overdue days

We will run balance check OLS regressions to confirm the comparability of treatment and control groups. We will use a dependent variables account holder characteristics and dwelling characteristics listed in Section 2 of the pre-analysis PDF attached to this form.

The experiment was run on the universe of residential dwellings present in the municipality in 2020. Randomization took place in two stages (first at the block level and then at the account level). In the first stage, we randomly assigned blocks to four groups with different intensity of treatment: pure control blocks where no accounts are notified, blocks with 20% of the accounts treated, blocks with 50% of the accounts treated, and blocks with 80% of the accounts treated. In the second stage, we randomly selected accounts within the last three groups of blocks to receive the treatment letter.

Randomization was done in office by a computer using Stata software and baseline data from 2019

Randomization took place in two stages: first at the block level and then at the account (dwelling) level. In the first level, we randomly assigned blocks to four groups with different intensity of treatment: pure control blocks where no accounts are notified, blocks with 20% of the accounts treated, blocks with 50% of the accounts treated, and blocks with 80% of the accounts treated. In the second stage, we randomly selected accounts within the last three groups of blocks to receive the treatment letter.

Yes

3,982 blocks (cuadras)

68,808 accounts (cuentas)

1,102 blocks with no treated accounts (19,105 accounts)
1,100 blocks with 20% treated accounts (C = 15,060 accounts, T = 3,853 accounts)
680 blocks with 50% treated accounts (C = 5,905 accounts, T = 5,897 accounts)
1,100 blocks with 80% treated accounts (C = 3,677 accounts, T = 15,311 accounts)

The power calculations give a minimum detectable effect between 2.6 and 3.3 percentage points (see details in the attached document).

PAP direct and spillover effects