Disparity Visualiser: A Tool for Illustrating Group-Level Disparities in Experimental Findings

The Disparity Visualiser is based on considerations outlined in Pietrzyk and Jacob (2026). It allows you to connect your experimental results to macro-level inequalities. You can illustrate how group-level disparities might change under varying levels of group-specific treatment prevalence, group-specific treatment effects, or group-specific baseline values.

You can use the Disparity Visualiser as a tool for sensitivity analysis to check how your experimentally established group-specific treatment effects contribute to group-level disparities under varying levels of group-specific treatment prevalence. If you choose to use the Disparity Visualiser in your research, please cite it as follows: Pietrzyk, Irena and Marita Jacob (2026): Disparity Visualiser. A Tool for Illustrating Group-Level Disparities in Experimental Findings. http://doi.org/10.5281/zenodo.18627891.

The Disparity Visualiser offers visualisations of two different outcomes: group-level disparity (Section 1) and the change in group-level disparity based on a treatment (Section 2).


Analysis Options

1. Group-Level Disparity

This outcome depends on group-specific baseline values, group-specific treatment effects and group-specific treatment prevalence.

The following formula is utilised:

$$\begin{aligned} Group\text{-}Level\ Disparity = &\left( Baseline\_Value_{1} + Treatment\_Effect_{1} \cdot Treatment\_Prevalence_{1} \right) \\ &- \left( Baseline\_Value_{2} + Treatment\_Effect_{2} \cdot Treatment\_Prevalence_{2} \right) \end{aligned}$$

For further information, see Pietrzyk and Jacob (2026) and Pietrzyk and Erdmann (2020).

1.1 For Interval-Scaled Data

You can use the Disparity Visualiser with your own data.

For easier understanding of the application, the input fields are pre-filled with an example from Gërxhani, Kulic, Rusconi, and Solga (2025). In this survey experiment, the researchers tested whether the effect of publication record (mostly single-authored publications compared to mostly co-authored) on the invitation propensity differed between female and male scientists. Invitation propensity was measured on a seven-point scale.

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the baseline values for the groups (i.e., mean values for each group in the absence of treatment) and the treatment effects.

The predefined value range is -1.5 to 1.5 for baseline values and -2.5 to 2.5 for treatment effects. Your values should be within these ranges when using z-standardised values. If your values do not fall within the predefined range, you can use the Disparity Visualiser with custom input (Section 1.3).

Women: 0.00
-1.501.5

Men: 0.00
-1.501.5
Women: 0.00
-2.502.5

Men: 0.00
-2.502.5

Results

Impact of Treatment Prevalence in Publication Record on Group Disparity in Job Invitation
Treatment PrevalenceMen
10.750.50.25
Treatment PrevalenceWomen
00.250.50.751
in favor of women
in favor of men

1.2 For Percentages

You can use the Disparity Visualiser with your own data.

For easier understanding of the application, the input fields are pre-filled with an example of an RCT conducted by Pietrzyk, Erdmann, Schneider, Jacob, and Helbig (2025). In this RCT, the researchers tested whether the effect of an educational intervention on university enrolment rates differed between students from low and high social origins.

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the baseline values for the groups (i.e., the rates for each group in the absence of treatment) and the treatment effects.

Low Social Origins: 0
050100

High Social Origins: 0
050100
Low Social Origins: 0
-30030

High Social Origins: 0
-30030

Results

Impact of Treatment Prevalence of Educational Intervention on Group-Level Disparity in Enrolment Rates
Treatment PrevalenceHigh Social Origins
10.750.50.25
Treatment PrevalenceLow Social Origins
00.250.50.751
in favor of low social origins
in favor of high social origins

1.3 For Any Type of Data (Custom Input)

If your values fall outside the predefined ranges for interval-scaled data or percentages, you can use the Disparity Visualiser with custom input.

You can use the tool with your own data. You can enter values between -200 and +200, with up to three decimal places. Both comma and period can be used as decimal separators.

For easier understanding of the application, the input fields are pre-filled with the example from Pietrzyk, Erdmann, Schneider, Jacob, and Helbig (2025), who have estimated the impact of an educational intervention on university enrolment rates for students from low and high social origins.

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the baseline values for the groups (i.e., the rates for each group in the absence of treatment) and the treatment effects. You can enter values between -200 and +200, with up to three decimal places. Both comma and period can be used as decimal separators.

Results

Impact of Treatment Prevalence of Educational Intervention on Group-Level Disparity in Enrolment Rates
Treatment PrevalenceHigh Social Origins
10.750.50.25
Treatment PrevalenceLow Social Origins
00.250.50.751
in favor of low social origins
in favor of high social origins

2. Change in Group-Level Disparity

This outcome represents the change in group-level disparity due to a treatment. It depends on group-specific treatment effects and group-specific treatment prevalence. Baseline values are not relevant for calculating changes in disparity.

The following formula is utilised:

$$\begin{aligned} Change\ in\ Group\text{-}Level\ Disparity = &\left( Treatment\_Effect_{1} \cdot Treatment\_Prevalence_{1} \right) \\ &- \left( Treatment\_Effect_{2} \cdot Treatment\_Prevalence_{2} \right) \end{aligned}$$

For further information, see Pietrzyk and Jacob (2026) and Pietrzyk and Erdmann (2020).

2.1 For Interval-Scaled Data

You can use the Disparity Visualiser with your own data.

For easier understanding of the application, the input fields are pre-filled with an example from Gërxhani, Kulic, Rusconi, and Solga (2025). In this survey experiment, the researchers tested whether the effect of publication record (mostly single-authored publications compared to mostly co-authored) on the invitation propensity differed between female and male scientists. Invitation propensity was measured on a seven-point scale. The change in disparity reflects the difference between the gender disparity in invitation propensity without and with consideration of publication record.

The predefined value range is -2.5 to 2.5 for treatment effects. Your values should be within these range when using z-standardised values. If your values do not fall within the predefined range, you can use the Disparity Visualiser with custom input (Section 2.3).

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the treatment effects.

Women: 0.00
-2.502.5

Men: 0.00
-2.502.5

Results

Change in Group-Level Disparity in Likelihood for Job Invitation by Treatment Prevalence of Mostly Single-Authored Publications
Treatment PrevalenceMen
10.750.50.25
Treatment PrevalenceWomen
00.250.50.751
in favor of women
in favor of men

2.2 For Percentages

You can use the Disparity Visualiser with your own data.

For easier understanding of the application, the input fields are pre-filled with an example of an RCT conducted by Pietrzyk, Erdmann, Schneider, Jacob, and Helbig (2025). In this RCT, the researchers tested whether the effect of an educational intervention on university enrolment rates differed between students from low and high social origins.

The change in group-level disparity reflects the change in enrolment inequality based on the educational Intervention.

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the treatment effects.

Low Social Origins: 0
-30030

High Social Origins: 0
-30030

Results

Change in Group-Level Disparity in Enrolment Rates by Treatment Prevalence of Educational Intervention
Treatment PrevalenceHigh Social Origins
10.750.50.25
Treatment PrevalenceLow Social Origins
00.250.50.751
in favor of low social origins
in favor of high social origins

2.3 For Any Type of Data (Custom Input)

If your values fall outside the predefined ranges for interval-scaled data or percentages, you can use the Disparity Visualiser with custom input.

You can use the tool with your own data. You can enter values between -200 and +200, with up to three decimal places. Both comma and period can be used as decimal separators.

For easier understanding of the application, the input fields are pre-filled with the example from Pietrzyk, Erdmann, Schneider, Jacob, and Helbig (2025), who have estimated the impact of an educational intervention on university enrolment rates for students from low and high social origins.

Defining the parameters

First, you need to specify your treatment.

Second, you need to specify the outcome.

Third, you need to define your groups.

Fourth, you need to define the treatment effects.

Results

Change in Group-Level Disparity in Enrolment Rates by Treatment Prevalence of Educational Intervention
Treatment PrevalenceHigh Social Origins
10.750.50.25
Treatment PrevalenceLow Social Origins
00.250.50.751
in favor of low social origins
in favor of high social origins

In developing the Disparity Visualiser, the authors utilised technical assistance from artificial intelligence.