GREENBERG QUINLAN ROSNER RESEARCH

How can we efficiently inform and engage an army of advocates and broaden our democracy?

Click here for an in-depth presentation documenting the research behind this groundbreaking new advocacy model.

Situation

Sometimes real change requires more than supporting a survey with progressive policy issues or even winning an election. Real change requires an informed, engaged citizenry willing to stand up and demand it from elected officials. Many groups – unmarried women, people of color, younger people – have historically been underrepresented in our democracy. It is often difficult to assure these citizens understand the significance of advocating for issues that may affect them personally and to find tools that will spur their involvement.

Recent public opinion research by Women’s Voices. Women Vote and other organizations also demonstrates that many of these same groups – unmarried women, younger voters, people of color—are at serious risk of dropping out during the 2010 election cycle. We must find a way to engage these voters now that not only adds their voice to current policy discussions but also keeps them engaged in the run-up to the 2010 election cycle. The challenge is finding a way that is cost effective.

Solution

Women’s Voices. Women Vote commissioned Greenberg Quinlan Rosner Research to build a model to predict propensity to take action. Rather than relying on stated intention to taking action in a survey (e.g., “I am very likely to write a letter to my member of Congress”), a measure that often yields unreliable responses, WVWV and GQR wanted to model “real world” advocacy. That is to say, rather than basing the model on people who say they will get involved, we based this model on people who actually become involved and call their elected officials. The modeling, therefore, was based on a program that generated actual patch-through calls to Senators’ offices.

The data collection used Interactive Voice Responses (IVR), where respondents were asked if they supported President Obama’s position on an issue and then asked if they would be willing to be connected (“patched-through”) to a U.S. Senate office to speak on that issue. Those who agreed were transferred to their U.S. Senate office. Respondents received one of four scripts where they were asked to support the president’s position on health care, equal pay, expanding the right to organize a labor union or the economic stimulus. The decision to use multiple scripts reflects the desire to build a model that could be used for different issues, rather than being specific to one issue.

The experiment was conducted among three audiences: WVWV past program participants, primary voters and registered unmarried women. In total, 130,214 contacts were made, resulting in 8,659 attempted patch-through calls for a 6.6 percent response rate. The results data were appended with consumer and census data and modeled.

In theory, the model produced a significant leap in efficiency. In the top tier—the respondents who were the most likely to patch-through—the model suggested a potential savings of 37 percent. There were different response rates in the different audiences— notably, the sample of unmarried women proved more responsive than the sample of primary voters—but the model produced savings across all three samples.

Validation

At the end of the day, however, models are predictions: they are not actual results. Therefore, Greenberg Quinlan Rosner validated this model with a separate study. This validation was between two different samples: one of registered unmarried women households and another of primary voter households. We compared the patch-through rate for a random sample of these voters with a sample predicted to be likely to take action using the model. For the latter sample only respondents predicted to have a relatively high patch-through rate were used.

The result: a 45 percent average gain in efficiency using the modeled sample. Boiling it down to dollars and cents, a successful patch-through cost of $3.86 using the random sample, compared to $2.14 using the sample predicted to take action, a savings of 45 percent. This savings varied in different states and other projects will likely see different calling costs, but we can confidently predict substantial savings regardless of the population or issue.

Moreover, unmarried women proved a more efficient target than primary voters in this validation. Assumptions about always using primary voters for this type of work should be revisited in light of this outcome.

The advocacy model can be used by groups like WVWV that educate the public and generally facilitate civic engagement or by advocacy-oriented groups that focus on generating action on particular policy topics.