Understanding non-take-up of Pension Credit and evaluating strategies to effectively boost it

We will send letters that inform recipients about their eligibility to Pension Credit (PC). The letters will vary in their content:
a) A basic letter, informing pensioners about PC, how much they are entitled to, and how to claim
b) A letter like (a) which also attempts to reduce stigma by emphasising that most eligible people claim
c) A letter like (a) except with branding from AgeUK, a highly trusted third party, rather than the borough, who may not be trusted
d) A letter like (a) which also emphasises that PC is private, so friends and family will not know the claimant gets it

Those receiving allocated to letters (a)-(c) will be sent the letter in February 2024. In 15 local authoriites, those allocated to letter (d) will be sent the letter in August 2024. In 4 local authorities those allocated to letter (d) will be sent the letter in February 2024.

The letters will specify the amount that recipients are (thought to be) entitled to. This will be rounded up to the nearest £500 per year (e.g. "You are entitled to up to £2,500 per year")

These letters will be sent in participating London boroughs (19)

2024-01-29

2024-08-31

Claiming PC six months after the letter is sent out

Claiming Attendance Allowance, poverty rates, council tax arrears, and social rent arrears.

Poverty rates will be calculated by comparing the household's income (as measured in the administrative data we are using) to the official before housing cost absolute and relative poverty line for 2024-25 (from the Households Below Average Income data).
Data on social rent arrears are only available in some boroughs. We will only examine the impact on social rent arrears for those households for whom we can observe them.
We will examine both share of households in arrears (council tax and social rent) and the average level of arrears as outcomes.

RCT with equal sized treatment arms. Letter (d) (described in the intervention field) is sent out 6 months after letters (a) to (c), meaning that recipients act as a control arm until that point. To evaluate letter (d) we will use a matched case-control approach.
We will stratify randomisation by borough and estimated PC eligibility.

To study the impact of information on outcomes, we will compare recipients of letter (a) with letter (d) each month up until the point that letter (d) is sent out, using recipients in the 15 local authorities that sent out letter (d) later than the others.
To study the impact of stigma and trust on outcomes, we will compare recipients of letters (b) and (c) with those that received letter (a), using recipients from all 19 local authorities.
We will focus on outcomes six months after the letters were sent out.

To test the impact of letter (d), we will compare outcomes of recipients to those of a control group comprised of an equivalent sample drawn from non-participating London boroughs (i.e. whose residents are not being sent any letters). The conditions for being selected for the control group will be the same as the the conditions required for being selected for treatment in participating boroughs. We will use a matched case-control approach, matching on age, sex, total income (in bands), council tax band, and separate indicators for any employment income, private pension income, state pension income, and disability benefit receipt. We will use a 1:1 match ratio.

We also conduct a regression discontinuity design (RDD) around the entitlement rounding thresholds. As described in the intervention field, the letters show estimated entitlement rounded up to the nearest £500. We will conduct a RDD around these thresholds, pooling across thresholds, using local linear regression.

To look at longer term outcomes, we will match control observations by drawing from eligible non-recipients of PC who are in the data we are using but live in metropolitan local authorities that are not one of the 19 that are part of the trial. We will match using baseline observations of key variables (as in the matched case-control exercise described above). We will then conduct a difference-in-difference analysis, allowing us to examine the evolution of these secondary outcomes over time and the impact of letter receipt. We will focus on outcomes 24 months after letter receipt.

Randomisation done in an office by a computer

Household

No

9,638 households

9,638 households

Approx. 2,409 in each of the four treatment arms (letters a to d)

When comparing the impact of the standard letter (A) vs. control, we will have 99.9% power at the 5% significance level to detect a 20% effect (that is, a 20ppt difference in take-up). When comparing the impact of the stigma or trust letter (B or C) vs. the standard letter (A), we will have 88% power at the 5% significance level to detect a 4% effect. When comparing the impact of the privacy letter (D) vs. the standard letter (A), we will have 80% power at the 5% significance level to detect a 4% effect.