Loading

Current Projects

Behavior Change to Save Energy in Low-Income, Urban Households

Catherine Wolfram and Sebastien Houde

Can an energy education tool help low-income households manage their energy bills and fight climate change? Led by E2e Faculty Director Catherine Wolfram and E2e Faculty Affiliate Sebastien Houde, this recently launched randomized control trial will allow us to compare online versus in-person education to answer this question in a transparent, precise, and generalizable way.

 
Motivation

Low-income households are the most vulnerable to increasing energy prices due to climate change policies (Grainger and Kolstad, 2010). Unfortunately, there is little evidence on the full psychological, economic, demographic and social factors that influence energy choices of the low-income population. Paradoxically, households in need of energy assistance tend also to be the ones that miss out on many low-cost opportunities to save energy.

 

Broad CO2 Tax Burden by Household Income Group    
 

The figure by Grainger and Kolstad (2010) shows total incidence by income group as a percent of annual net income and current expenditures.

As you can see, a price on carbon, given the assumptions used by the authors, would be regressive, although the degree of regressivity depends on the income measure used.

Notes: Assumes CO2 price of $15/ton.

Source: Grainger and Kolstad, 2010

   

 

The research question

The fundamental question is whether low-income households:

  1. Do not know about these energy- and money-saving steps or
  2. Have the information but decide not to act on it, for example because they perceive the energy savings as minor relative to whatever psychic or other costs are involved in achieving it.

The study intends to tackle these two main questions.

 
Summary of the study

The study will be implemented in collaboration with the Fuel Fund of Maryland, a non-profit organization that has provided energy assistance to low-income households for more than thirty years. Since 2008, the Fuel Fund of Maryland has run Watt Watchers, a program that uses community based social marketing principles and tools in combination with energy education, to foster long-lasting behavior changes. This program encompasses different approaches including an online, an onsite, and a community-based version of the training. After the training is completed, the households enrolled receive a follow up service called the Energy Advisor program. Energy Advisors are trained volunteers, who work with the graduates from the Watt Watchers program. They will follow-up by phone with graduates and provide personalized energy-related guidance and counseling services.

 

 

 

 

Fuel Fund clients are households that missed electricity payments and now need financial assistance to pay their bill. In a typical arrangement, Fuel Fund covers one third of the amount owed, the utility, Baltimore Gas and Electric, covers another third, and the client is responsible for the remaining amount. In order to receive financial assistance, clients also need to enroll and graduate from one of the versions of the Watt Watchers training.


The classroom setting of Watt Watchers is 2 sessions, 1h30 each for a total of 3 hours. Sessions are scheduled 1 week apart to allow enough time for participants to try out one behavior they select at the end of the first session. The online program is much shorter and can be completed in approximately 45 minutes, from any computer, smartphone, or tablet.

 

The need for randomization

A naive comparison would likely yield selection problems since it is reasonable to believe that households who enrolled in Watt Watchers are inherently more "green" or prone to save more. This will over-estimate the impact of the program. Our study will create credible comparisons in order to avoid this issue. A randomization process will, on average, create statistically identical - thus comparable - "control" and "treatment" groups.

 
 
Design of the randomization

With a low-cost, simple intervention we plan to compare the impact of the program on energy usage (electricity and natural gas) of households who enrolled in the program with the ones who did not, as well as to directly compare the cost-efficiency of the online and onsite classes. We find this approach especially exciting given current debates about the value of online education versus the traditional onsite approaches. In our study, we have a clean, simple, precise way of measuring the difference between the two approaches, which is the energy savings achieved by households enrolled in the two classes.

We will randomize the requirement for the Watt Watchers training to clients requesting service from the Fuel Fund. Clients will be randomly placed in one of these three groups:

 

Control Group:

No training required.

Treatment Group 1 - Online Classes:

In order to receive the financial help, clients are required to take a 45 min long online training. A link to this tool is provided by the case who worker is registering the client. When the client comes back for the second interview, he must have the completed the course and gathered the money discussed.

Treatment Group 2 - Classroom Setting:

In order to receive the financial help, clients are required to participate on 2 sessions of 90 min (3 hours total) each provided by FF. These classes are face-to-face interactions in groups of 15 to 25 clients. Upon registration, the case worker will have another calendar with class dates and will inform the date/time that the classes will take place. Sessions are scheduled 1 week apart to allow enough time for participants to try out one behavior they selected at the end of the first session.

 

The study started in October 2014. All Fuel Fund locations are following a pre-defined calendar that informs the caseworker in charge of the first interaction with the client in which group he/she should be placed. Every week the assignment changes, but all clients registering in that location/week will be placed in the same group. The schedule was developed such that the cycles in all location are neither predictable nor consistent (e.g., online-onsite-control-online-onsite-control...), although we guaranteed that every location receives at least one of each type of class in every month. Each location follows an independent schedule.

 

Sample size

Our sample size will depend on the demand for FF assistance this year. In the 2013 cycle, FF provided assistance to about 6,000 clients. We should still be able to get statistical power in the case of imperfect incompliance. Our calculations show that with a 6000 participants sample and even split between the three experimental groups, the minimum detectable effect goes from 2.5% with perfect compliance, to 10% in the case of 25% compliance.