Skip to 0 minutes and 9 secondsHello, I'm here with Professor Dan Anderberg of the Economics Department at Royal Holloway where, as you know, I also work. So, Dan, you've got a lot of interesting research on the subject of domestic abuse. That must be a difficult subject to research. It must be hard to measure this, it's something that people don't want observed, particularly. That is correct. It's a field in economics that has come up more recently over the last 5, 10 years. And the way that people are measuring domestic violence, it varies quite a lot in the literature. I guess the UK is kind of fortunate, in that it has, to some extent, some of the best measurements available, in terms of survey methods.

Skip to 0 minutes and 57 secondsBut it's not only [INAUDIBLE] it's not only survey methods that are being used. In terms of survey methods, where the UK has been, sort of, leading is in terms of introducing questionnaires where you leave the respondent alone with a laptop, as opposed to having an interviewer in the room. And that has been shown to pretty much triple the amount of abuse that people are willing to disclose. Really? Yes, so that's a clear under-reporting with an interview in place. But when they ask about specific categories where the respondent is entirely left alone, you can see that the reported abuse starts creeping up.

Skip to 1 minute and 45 secondsAnd do we have experiments where it's not done that way, where were they continue to ask the same questions face-to-face without leaving people on their own? So these surveys often have some overlap in the questions. They have actually asked both with and without interviews so you can see even the same person reporting more when they are alone. OK, so what are the different measures of domestic abuse that are used? Effectively, there are three types of measurements that are used. Either researchers tend to use measures that are not done by the respondent themselves, but things like hospitalisations, police responses to households.

Skip to 2 minutes and 32 secondsOn the argument that these ones should be more objective-- one could argue. But on the other hand, even these kinds of measures are subjective in the sense that someone took an action. Someone decided to go to hospital. Police decided to come to the scene and tick the box that this was a domestic disturbance. So someone has made the decision along the way. So these measures, by people who use them, they are argued to be fairly objective. But they are not entirely objective. And they will also have a fairly low incidence measurement, in terms of the proportion. So we believe, that in terms of physical abuse, it affects around 3% of women in the UK every year.

Skip to 3 minutes and 21 secondsBut if you use these kind of measures of a hospitalisation, police reports-- you're only capturing a fraction of that. So they are good in the sense that they are meant to be fairly objective, but they also measure only the sort of the tip of an iceberg. They tend to be popular for evaluating what is the fix up, or what is the effect of this on abuse? Because then you want to have a fairly objective measure. So even if you're under measure it, you're better off doing that and have something that doesn't have too much a shifting of composition. So you want to know that you're measuring something that is that's fairly consistent over time.

Skip to 4 minutes and 1 secondSo you can measure the instance before-- So for example, the trends might still be accurate even they're-- Right. Exactly. If you see a 10% increase in these fairly narrowly defined measure, it may reflect the 10% increase overall. But if you use a broader measure, you may be capturing more compositional changes. But there could be changes in reporting patterns over time. For example, if there's a nationwide campaign to encourage people to report abuse, and then you might see-- Oh, dear. Oh, absolutely. I mean, to some extent this is not at least the problem in our time and right now, a lot of discussions have been whether what you observe is increasing trends and the annual reported is an increase in reporting.

Skip to 4 minutes and 51 secondsAnd that maybe has to do with different cohort effect, if you ask people who were born 20, 30, 40 years earlier, they have a different opinion about what is acceptable behaviour and what is not. And they will have different reporting patterns. More recently, if you think the #MeToo kind of movement, people are saying that this may have had a big impact on what people are willing to disclose and report and what they think is acceptable or not. So definitely, we do observe changes in trends that may very well be driven by changes in reporting behaviour.

Skip to 5 minutes and 28 secondsOf course, we can circumvent part of that in research by looking, for example, at changes across areas and time to see whether in areas that they implement that something, so a different trend-- because if both have trends that are driven by changes in reporting behaviour, well, you can look at the differential trend instead. Reporting behaviour is probably also important if you're trying to do international comparisons. Yeah, that is a challenging one, indeed. First of all, just look at how things are measured in different places. I alluded to the interviews with laptops, that was first introduced in the UK, where it's been used now for 20 years.

Skip to 6 minutes and 14 secondsBut other countries are lagging and don't have as much of such survey data available. So making international comparisons is, well, challenging, to say the least. What kinds of interesting things do we find?

Skip to 6 minutes and 31 secondsWe find quite a lot, to be honest. I mean, going back to the last one. One if the projects that I've been involved in myself, and you've seen these kinds of subjective measures. You can look at the dynamic response. We can observe how women change their behaviour after saying that their partner did something that was wrong. So we can observe fertility responses. We can observe how the probability of having further children after those incidents, are really dropped substantially. And we can observe whether they go back into labour markets. So we can look at what they perceived to have happened in their life, and see how they respond to that-- their dynamic responses, in terms of their big life choices. OK.

Skip to 7 minutes and 19 secondsThank you very much. I think it's been a great interview. Thank you very much. Very nice being here.

Domestic Abuse

I was genuinely astonished by an effect that Professor Dan Anderberg mentioned at the beginning of this interview; leaving women alone to answer questions about domestic abuse triples the reported rates relative to what is measured in face-to-face interviews.

There could hardly be a better demonstration of how methods of measurement matter.


Dan Anderberg is Professor of Economics at Royal Holloway University of London. His research interests are in the areas of family economics and public economics. He is also affiliated with the Institute for Fiscal Studies (London) and a CESifo Research Fellow (Munich).


I can’t resist drawing your attention to a recent research disaster that is reminiscent of the interview technique that Dan describes in the interview. A recent book used something called the American Time Use Survey to argue that many people are very unhappy in their marriages but won’t admit this fact in front of their spouses. It turns out that this claim is based on a misunderstanding of what the phrase “spouse not present” means in the survey.

The respondents classified as spouse-not-present are actually people whose spouses were living in a different house, not people whose live-in spouses were temporarily out of the room during part of their interview. Of course, it’s not so surprising that people who have physically separated from their spouses are less happy on average than people who are still living with their spouses.

Ooops.

Again, the lesson is that careful measurement is crucial.

Discussion

Feel free to share in the comments area any interesting mismeasurement stories you may know about

Share this video:

This video is from the free online course:

Survival Statistics: Secrets for Demystifying Numbers

Royal Holloway, University of London