Melissa Gayton is a Special Projects Assistant with the Access to Justice Lab and a student at Harvard College.
In 2018, the A2J Lab completed a study with the help of the Philadelphia VIP and the Philadelphia Family Court on the accessibility of divorce in Philadelphia. The study randomized low-income people seeking a divorce into one of two groups:
- Treated group: an effort by the service provider to find a pro bono attorney to represent her;
- Control group: a referral to existing self-help resources and an offer to answer questions by telephone.
The results showed that people with lawyers were 87% more likely to successfully access divorce than people without lawyers. (You can read a longer summary of the study and its results on our website or explore the full results in the corresponding paper, Trapped in Marriage.)
We are always interested in new ways of presenting the results of our studies and of sharing the data from the results.
To that end, I created an interactive representation of some of the study’s results. Before potential clients entered the study, they underwent a 45-60-minute interview. The interactive tool offers different ways to explore the information collected during these interviews.
The interactive graphs may be difficult to understand if you’re not familiar with the statistics and the study. This post will guide you through how to get the most out of the graphic representations available.
Let’s look at a sample of some of the information you can find under the “Comparison Table” when you compare across whether or not the potential client required an interpreter and only look at statistically significant differences – differences that were probably not caused by chance.
In the rows presented here, the “No Interpreter” and “Interpreter” columns can be understood as the percent of participants who belonged to that demographic group. The “Mean Difference” is calculated by subtracting “No Interpreter” from “Interpreter.” The “P Value” indicates how likely it is that the mean difference is a result of chance. If the p value is less than 0.05, we can be confident that the differences were not the result of chance.
In this example, the rows are telling us that people who needed an interpreter were less likely to be African American or white and less likely to speak English as their primary language at home. They were more likely to be Hispanic or Latino or of some other race.
You can get a more visual representation of this information by looking at the Comparison Plot.
For this plot, how high or low the dot is on the vertical axis doesn’t matter. The most important thing is whether the dot is to the right or left of the vertical line, which is drawn at 0.05. If the dot is to the left of the line, that means the p-value is less than 0.05.
You can see that, once again, there is a statistically significant difference in whether or not the potential client is white between people who needed an interpreter and those who didn’t. However, there’s no difference when comparing across whether or not the potential client or their spouse had already filed for divorce, whether or not the potential client received legal representation or not, or whether PLA had rolled back their divorce practice.
Both the Comparison Table and the Comparison Plot can be found under the Summary tab. The other five tabs – Demographics, Income, Assets, Marriage, and Family – contain some additional information about these specific categories of variables. For example, under the Income tab, you can compare the wages reported by potential clients and their spouses. These comparisons are represented through a type of graph called a histogram.
The higher the bar on a histogram is, the more people make that wage. It’s important to note that this graph does not include the 65% of potential clients who did not earn wages. The slider on the left lets you change the number of “bins” that the wages are divided into. Let’s look at what happens when we reduce the number of bins to six.
Now, the groups are much larger. You can see that most people appear to be making around $1000 a month or less.
Other graphs are not interactive but present important information. For example, the Assets tab includes a graph breaking down what kinds of assets and liabilities that a potential client has.
We hope that these graphs are helpful in understanding the study results, and we look forward to sharing more updates on data from this and other studies.