Gender Gaps in Depression Diagnosis

The study uses a within-subjects design wherein all participants are exposed to two treatment blocks. In each of these blocks, participants will be shown one or more hypothetical scenarios. A scenario involves a list of depression symptoms and the frequency of their occurrence. The study uses a within-subjects design wherein all participants are exposed to two treatment blocks. In each of these blocks, participants will be shown one or more hypothetical scenarios. A scenario consists of a list of depression symptoms and the frequency of their occurrence over the past two weeks. The displayed symptoms are chosen from the PHQ-9 questionnaire, which is a widely used (and often self-administered) instrument that screens for depressive disorders. The list of symptoms and frequencies are randomly chosen to meet certain thresholds which indicate different levels of depression severity.

After reviewing the hypothetical scenario, participants are asked to state their views on: (i) whether the symptoms indicate that the subject of the scenario suffers from depression (recognition), (ii) perceptions of depression severity, (iii) the likelihood of seeking help from different sources, and (iv) beliefs and attitudes regarding depression and mental health.

[NEW October 2026] A similar experiment will be conducted with mental health specialists recruited from the members of the American Psychological Association to establish whether there are differences in how specialists diagnose and recommend further steps for male and female patients presenting similar symptoms.

The two treatment blocks, Self and Other, differ only in the subject of the hypothetical scenario. In Self, the subject of the hypothetical scenario is the participant. In Other, the subject of the scenario is a hypothetical individual (who is either Male or Female). Participants will evaluate 2 scenarios in each treatment block and the order of the blocks will be randomized. The symptoms in the second scenario in the Other treatment is fixed and the subject of that scenario is the same as the subject of the first scenario in that treatment. This common scenario will be used to elicit second order beliefs by asking participants to provide their best guess of the percentage of participants in the study who thought that the scenario described a person with depression. This guess will be incentivized based on the actual responses from participants in the study.

In the vignettes, the scenario including the subject of the scenario (Self or Other) is introduced. Then three or more symptoms are shown to the participant in a tabular form, along with a frequency which is one of Several days, More than half the days, Nearly every day. The list contains the questions exactly as stated in the PHQ-9 instrument. Then, the participants respond a series of questions regarding the scenario:

Q1. Suppose that [you/NAME] were/was experiencing the hypothetical issues at the frequencies listed above, do you think [you/NAME] would have depression? [Definitely yes, probably yes, probably no, definitely no]
Q2. How severe do you think the depression would be if [you/NAME] were experiencing these issues in real life? [None or minimal, Mild, Moderate, Moderately Severe, Severe]
Q3. If [you/NAME] were experiencing these hypothetical issues, how likely do you think it is that [you/she,he] will seek help from the following sources? For these questions, imagine that there are no constraints on the time or money that has to be spent, and no problems relating to health insurance coverage for these options. [Very unlikely, Somewhat unlikely, Somewhat likely, Very likely]
a General Practitioner solely for this purpose
a General Practitioner during a visit for another purpose
a Psychologist or a therapist
a counselor at your workplace or university
a close friend or relative
an AI-enabled mental health chatbot
Q4. Based on the hypothetical issues and their frequencies experienced by [you/NAME], please indicate the extent to which you agree or disagree with the following statements. [Strongly disagree, Somewhat disagree, Somewhat agree, Strongly agree]
\item The issues would go away by themselves in some time
The issues are not very serious and do not require treatment
I/[NAME] worry/worries about what others would think of me/her/him if they became aware that [I/she,he] had these issues
I/[NAME] would rather deal with these issues [myself/herself,himself] than rely on help from others
I/[NAME] do/does not think that the available treatments for these issues are effective
I/[NAME] am/is worried about the side effects of medication for depression
Q5. We conducted a similar survey with a sample of 100 Americans. The composition of respondents in this survey was broadly representative of the American population. Participants in that survey evaluated the following exact hypothetical scenario and answered whether or not they thought [NAME] would have depression if he/she were experiencing these symptoms.

List of symptoms

Of the 100 Americans who participated in that survey, how many do you think answered Definitely yes or Probably yes to the question: Suppose that [NAME] were experiencing the hypothetical issues at the frequencies listed above, do you think [NAME] would have depression? [Number between 0 and 100]

[NEW October 2026] The sample of mental health specialists from the APA will evaluate three scenarios each, where some of the scenarios are about women and some about men.

2024-10-23

2024-10-31

The three primary outcomes are Recognize, Correct and Seek help. Recognize is a binary variable equal to 1 if the answer to Q1 is definitely yes or probably yes, and 0 otherwise. Correct is the outcome to assess whether participants assess the severity of depression correctly. It takes the value of 1 if the severity level is identified correctly and 0 otherwise. Seek help captures the willingness to seek help after seeing the list of symptoms and answering Q1 and Q2. We will use the responses to Q3 to generate an outcome variable which equals 1 if the response is somewhat likely or very likely for one or more of the options presented in Q3 and 0 otherwise.

[NEW October 2026] For the APA sample, we have three main outcomes:
(1) Depression labeling: an indicator equal to 1 if the respondent answers "Definitely yes" or "Probably yes" when asked whether the hypothetical patient would have depression, and 0 otherwise.
(2) Accurate severity assessment: an indicator equal to 1 if the respondent correctly identifies the PHQ-8 severity category of the hypothetical scenario, and 0 otherwise.
(3) Recommendation for professional assessment: an indicator equal to 1 if the respondent is "Somewhat likely" or "Very likely" to recommend that the hypothetical patient seek a professional evaluation for depression, and 0 otherwise.

The first set of secondary outcomes contained whether help would be sought from a mental health specialist or from non-specialists. The specialist variable will be equal to 1 if the response is somewhat likely or very likely in at least one of options 1-3 in Q3 and 0 otherwise. The non-specialist variable will be equal to 1 if the response is somewhat likely or very likely in at least one of options 4-6 in Q3 and 0 otherwise.

The second set of secondary outcomes we intend to capture three different factors that could underlie gender differences in recognizing depression and seeking help. We refer to these three factors as perceptions, psychic costs and norms.

To capture gender differences in perceptions we will generate binary outcomes from each of the answer options in Q4. Each of these variables would be equal to one if the response is somewhat or strongly agree and 0 otherwise. We will also generate an index summarizing the perceptions in a single variable.

To study psychic costs, we will use the Self and Other scenarios and create a binary variable equal to 1 if the scenario being evaluated corresponds to Other.

To study norms, we will survey responses to Q5.

[NEW October 2026] For the APA sample, secondary outcomes capture recommendations to seek help from different sources and respondents' perceptions of whether professional evaluation or help is warranted.

The experiment involves randomly choosing a list of depression symptoms and severity of symptoms from the PhQ-9 questionnaire to build a scenario that participants are prompted to assess. The two treatment blocks refer to who is the subject of the scenario (Self or Other). we choose symptoms (i.e., items form the PHQ-9 scale) and frequencies so that the combinations of symptoms and frequencies gives an overall score between 5 and 27.

All participants see scenarios in both Self and Other in randomized order. The list of symptoms and frequencies is randomly selected based on a pre-defined algorithm.

[NEW October 2026] For the APA sample, we use the PHQ-8 questionnaire (eliminate the suicidal ideation question from the PHQ-9 list). We randomize the gender of the subject of the subject of the scenario, which is our main focus. We also randomize the attributes we mention about the subject like age and occupations, keeping these attributes as close as possible.

Individual - within subject.

No

Non-clustered design.

300 individuals who evaluate 4 scenarios.

1200 observations from 300 individuals who evaluate 4 scenarios each.

[NEW October 2026] For the APA sample, we aim to recruit at least 100 respondents. Recruitment will occur through relevant professional psychology listservs and networks.