How change sticks: How treatment effects can persist after treatment has been discontinued
usually don't observe persistence when we do field experiments, sometimes we do
Cahill & Perera, 2008 is a health example
meta-analysis of smoking cessation studies
Volpp, 2009-->6months later persistence

MT: doesn't seem like long

same with weight loss
bariatric surgery is the most persistent

OPOWER
psychology of social norms
experiments with OPOWER examining "persistence"
Alcott & Rogers (forthcoming, AER)

Social norms: injunctive and descriptive
littering Cialdini
Towel reuse Goldstein
Voter mobilization Gerber & Rogers
Tax compliance (Behavioural Insights Team, UK, 2012)-->social proof letters to increase compliance 4% - 14%

How to reduce energy use Schulz, Nolan, Cialdini, Goldstein, Griskevicius
T1 and T2 energy use measures
two universes of households: efficient, inefficient

two conditions for each: door hangers: descriptive (your neighbors use x% more/less energy than you), descr + injunctive ((your neighbors use x% more/less energy than you and that means you're a good/bad person)

T3 measure
change in energy use across the four:

Laski & Yates OPOWER
business model: utilities are regulated fixed profit per user plus additional profit from demonstrated reduction in enery use
another incentive utility companies must have twice as much capacity as used to avoid building new power plants
valuable to them to shave peak demand
OPWER exploit this niche: send Last Month Neighbor Comparison
bar chart efficient neighbors, all neighbors, YOU
"how you're doing"
roll out experiements with control groups with every client to demonstrate reduction
reduction is abt 50% of all solar production
1.9% - 3.0% reduction
equivalent to 18% - 31% short-run price increase
IPO on the horizpn
in calculating cost, they assume 0 persistence****treatment effect during the duration of the treatment

UCSB cost effectiveness of intervention planting trees with assumed 100% persistence

who the fuck thought it was a good idea to put crunchy food in lunch boxes for seminars?

3 experiements
monthly or quarterly reports
three ocnditions untreated control, treatment group (over 4-5 years) dropped group (stop sending the report after ~2yrs)
after 3 mos, no immediate effect on report receipt
over the succeeding 3 years, about 50% of the treatment effect persists
define persistence as the effect - control at end of delay pd over the effect before discontinued
was surprising bc expected it to bounce back up
also expect ppl to habituate to the report receipt
the reports are less punitive as ppl improve
why is there marginal benefit to continued treatment? perhaps bc marketing carryover effect and perhaps the diminishing punitive nature of the repots as ppl improve on energy conservation
why does seasonality increase after treatment?
MT: bc when you switch from heat to air, you think about it again
to value this treatment, persistence is very important

one implication of persistence
could imagine it will persist forever
total energy savings induced by dropped groups prior to end and after end: post-treatment is greater than during treatment
MT: but it's 3 years compared to 2 years

how can persistence arise?
surveys of 6k across two conditions and 6 sites
not due to repeated behaviors, intermittent behaviors, physical capital investments,
but maybe changed self-perception (I'm the kind of person who cares about energy use)?
maybe changed social relationships?
changed engagement by unrelated groups? third-party mailings

Craig doesn't think so
good point is that the survey goes out long after any behavior change

might have an ordinal sense of how i'm changing my behavior-->range frequency effects account for change in behavior

how could we begin to determine why the treatment effects are

Pathways to persistence (with Erin Frey)

Persistence framework
Rogers & Frey (under review)
five pathways through which effects can be made persistent

framework incorporating all sorts of persistence mechanisms
Internal to self

psychological habits
change in mental content

1. changed a belief like the law protecting privacy of vote (fixed mental stock, includes preferences)
2. changes interpretation of ambiguous stimulus/lens through which people construe the world (i'm not a failure, I have opportunity to learn)
3. changes interpretation of the self

External-to-self

1. Recursive social processes (Yaeger & Walton) -- effects can grow from this
2. Change in future costs (program your thermostat, retrofit your home, etc. one-time costs now; defaults)
3. Rip currents (discontinuity where small push can carry you into the flow of other treatments (voter participation,

Likely that the reason for persistence is due to treatments other than yours, but it may be that the person has been nudged into a targeted group

Persistence may arise for reasons entirely unrelated to initial treatment

Studying interventions framework for examining, measuring, and explaining presence or absence of persistence (also assess each pathway to understand mechanism)

Developing interventions framework for intentionally designing features to tap into pathways

People in a social network

Useful for interventionists designed to leverage additional pathways (like social network processes)

Useful for researchers developing measurements of what the possible categories for where persistence can derive from; also discriminant validity check to have this framework

Summary

OPOWER

• surprisingly potent
• surprisingly persistent
• Policy cost-benefit analyses should reflect persistence

Need to extend the period of experiementation and estimate a decay rate

similar to medical trials?

Periodic treatment? frequency cost effectiveness for optimal model

Data on the average treatment effect and what's causing it? only heterogeneity: bigger energy users show bigger treatment effects

see YOY energy inducing treatment effects and see whether they are the same group with the same effect?