Employing Behavioral Economics: Improving the Efficiency of TV License Fees Collection

We run the experiment in cooperation with the CT, the national public television broadcaster. Our experimental sample consists of subjects that are contacted by the CT for the first time. They are asked by mail to clarify why they are not paying TV license fees and suggested to register for paying. We (i) simplify the structure of the letters, (ii) vary their informational content to separately address five distinct motives of non-compliance, and (iii) in an orthogonal design to that we vary the levels of attention the envelopes raise.

2017-05-19

2017-06-06

response rate, registration rate, after 40 and 90 days from mailing the letters

We focus on two main outcome variables: any response to the letter (not necessarily resulting in a new registration) and a new registration, both being binary variables. We will measure the response at two time points, 40 and 90 days from sending out the letters. We will also differentiate with respect to different reasons of de-registration (contract update vs. statement of no TV vs. other reasons).

We employ altogether 6 text manipulations of the contact letter, including a baseline control letter (T1). The baseline letter is formulated neutrally, with no emotional or moral appeals. In comparison to the letter sent by CT in previous waves (for which we know the response and registration rates), its informational structure is significantly simplified and the amount of text is reduced so that it does not force the recipient to read irrelevant information. Treatments T2-T6 differ from the baseline letter only in one additional paragraph with specific information. Moreover, in an orthogonal design we manipulate (i) the attention level raised by the envelope and (ii) attention level and reciprocity feelings by placing a sticker with a picture of a cartoon character on the envelope (A3). The letters are sent by the CT.

2.2 Treatments and randomization
We employ altogether 6 text manipulations, including a baseline control letter (T1). The baseline letter is formulated neutrally, with no emotional or moral appeals. In comparison to the letter sent by CT in previous waves (for which we know the response and registration rates), its informational structure is significantly simplified and the amount of text is reduced so that it does not force the recipient to read irrelevant information. Treatments T2-T6 differ from the baseline letter only in one additional paragraph with specific information.
The five information manipulations are aimed at the most probable reasons for the avoidance of the fee payment. The first (labeled "Why", T2) addresses the fact that many people do not understand why the fees have been introduced in the first place instead of broadcasting being financed from the state budget directly or from private sources. The main message is that the fees secure the independence of the CT financing on the state budget and on private funding through commercials. The second manipulation ("Exchange", T3) appeals to the business-like interaction between the CT and the respondent and outlines the services and production provided in exchange for the fees. The third manipulation ("Participation", T4) is textually almost identical to the previously mentioned in that it outlines the services provided by the CT, but is framed as a question to the respondents regarding what types of program they would like to see more and invites them to participate in an online survey where they can express their preferences. Manipulation four ("Threat", T5) articulates the consequences of not paying as stated by law and puts the CT into the position of being forced by the law to demand the respondent to pay. The last, fifth manipulation ("Social norm", M6) appeals on the social norm to comply with the law, stating that 90% of the households pay as required, and labels the non-paying household as free-riders.
Moreover, in an orthogonal design we manipulate (i) the attention level by placing a label "Important" (A2) on the envelope and (ii) attention level and reciprocity feelings by placing a sticker with a picture of a cartoon character Maxidog Fik inside the envelope and the same picture also on the envelope (A3).
Finally, to measure the effectiveness of the baseline letter alone, we randomly select a second control group that does not receive any letter (T0). This no-letter group is relatively small, only 5% of the whole sample.
Overall, we have two control groups (T0, T1&A1) and seven treatment interventions (T2-6, A2-3), with T and A treatments being randomized orthogonally, resulting in 18 different cells. We stratified on a regional level (first number in the ZIP-code) and on age-groups, where we divide the sample by median age. This is because the CT allegedly faces a steady decline in the registrations of "young" households in large cities.
We expect that some respondents will not receive the letter due to technical problems, but this should be random across all treatments.

Control group T0 = 4133 Manipulation
N=82.645; N_send=78512 T1 T2 T3 T4 T5 T6
Attention Baseline Why Exchange Participation Threat Social norm
A1 Baseline 6543 6543 6543 6543 6543 6543
A2 Label „Important“ 3272 3272 3272 3272 3272 3272
A3 Sticker 3272 3272 3272 3272 3272 3272

Randomization done in Stata, also using command randtreat.

Household

No

there is no clustering employed.

82.645 households

Control group T0 = 4133 Manipulation
N=82.645; N_send=78512 T1 T2 T3 T4 T5 T6
A1 6543 6543 6543 6543 6543 6543
A2 3272 3272 3272 3272 3272 3272
A3 3272 3272 3272 3272 3272 3272

Treatment cells are balanced with respect to the first digit of ZIP code and one of two main age categories (created by median split).

The study will have 80% power to identify a 1.75 p.p. effect of T2-T6 on the probability of any response relative to T1 (assuming the effectiveness of the baseline letter T1 equal to 43%, which is based on previous waves) and 1.05 p.p. effect on the probability of new registration (assuming the effectiveness of the baseline letter at 9%). For the attention treatments A2 and A3 relative to A1, we will have 80% power to identify a 1.25 p.p. effect on the probability of any response and 0.75 p.p. effect on the probability of new registration, assuming the above-mentioned effectiveness of T1. We will report the minimum detectable effect sizes in the paper.