7

Proposed methodology for leafleting study

After some of the recent controversy about the state of the evidence supporting ACE’s recommendations [1] I started thinking about how best we could study leafleting, and I think there is a strong opportunity for someone to do a much higher power study than has been managed before, with relatively low effort.


Disclaimer: All of the views/plans presented therein are my own, and not officially affiliated with or endorsed by Mercy For Animals.

1. Background

1.1 A brief overview of previous research

A good review of the previous studies on leafletting effectiveness can be found here: http://veganoutreach.org/les-fall-2016/

The primary takeaway is that most of the studies conducted so far have not been controlled making it impossible to infer the effect of leafleting versus a general trend towards vegetarianism. The studies which have used controls have also always had extremely small control groups, the largest being 57, meaning none of them had the power to measure any statistically significant results.

 

This has left leafleting effectiveness estimates almost entirely dominated by personal judgements, despite the fact it is a significant tool used by effective animal charities. [2]

1.2 A summary of my plan

At a large university (pseudo)-randomly split the students evenly into two groups, and put leaflets into the pigeonholes of all members of one of the groups, and none of the other. Then send a follow up survey (possibly incentivised) to all students, a week later, asking:

  1. A question to determine if they were leafleted or not, without directly asking.

  2. If they have changed their diet in the last two weeks, and if so, how.

 

This would provide a controlled trial with a sample size of many thousands if the survey response rate was high enough (see 3.4), enough to find an effect size of approximately 1/75-1/250 leaflets creating one vegetarian (depending on sample size, as it ranges from 2,500 to 10,000, see 3.5).

1.3 My aim for this post

After contacting my university it turns out running such a study in Cambridge would not be possible (as the Student union/faculty will not email out a survey, and due to other considerations, see 3.1). However I think such a study could still be extremely valuable if conducted elsewhere.

 

By laying out the plan in detail I hope to both get feedback on areas which could be improved, and hopefully find a university at which it could be implemented.

2. Detailed plan

2.1 Preparations

Determine which students you will leaflet, either using some kind of randomisation or any natural divisions that exist (see 3.1).

 

Contact the student union/university itself and ask if they would email the survey out to the entire student population. Getting the survey to reach everyone is very important, so it would be worth working hard on this, meeting them in person if necessary etc. Also write some articles about the survey to feature in online student newspapers (the TAB etc.) and discuss getting them published in advance. For both of these playing up the fact it is a potentially large and important study being conducted by university students would probably be helpful. Incentivising the survey responses with a fairly large amount (e.g. £5 each) might also be most useful at this stage, as even if it does not boost response rate very much (see 3.5) getting the survey emailed out to everyone is very important.

 

Work with Statisticians Without Borders (or similar experts) to check the statistics behind the study all work out, and pre-register the study, including which effects we are will be looking for (such as a higher than base-rate number of people who were leafleted turning vegetarian).

Obtain as many leaflets are you are expecting to hand out (in the UK Animal Equality distribute them to students to hand out for free).

Contact/hire an online survey company to set up a survey that is linked with the university's emails or similar, so that each student can only fill it out once. The survey should also have the following questions:

 

  1. What is your last name? Or similar question designed to determine if they were leafleted or not (see 3.1).

  2. Have you changed your diet in the last two weeks?

    If they select yes them being presented with two further questions:

  3. Which label best described your diet before the change?
    Options:

  • Meat Reduction Diet (A diet reducing meat consumption, for example Meatless Mondays)

  • Pescetarian Diet (eat fish, egg, and milk products, but no other meat (including chicken))

  • Vegetarian Diet (eat egg and milk products, but no meat (including fish or chicken))

  • Vegan Diet (eat no meat (including fish or chicken), milk products, egg, or other animal products)

  • No specific diet (A diet with no specific preferences or exclusions)

  • Other __________________________

 

  1. Which label best describes your diet after the change?
    With the same options.

If the survey was incentivised there would then be a tickbox for “I would like a £5 amazon voucher” and a submit button, which would take them to a ‘thank you’ page with a referral link to send to their friends, and the information that they would get another £2 amazon voucher for each of their friends that used it and filled out the survey.

2.2 Implementation

Once all the setup is completed, find one-two days where 4 volunteers/workers are free, split into two groups and work through putting leaflets in the pigeonholes of all the students of all the colleges you arranged to. (For the practicalities of this we have found filling two suitcases with leaflets allowed two people to transport about 2,500 at once). We have been able to give out an average of 1000 leaflets an hour as part of a two person team when mass-leafleting like this, so it might be possible to do it in only one day, or two days with only two people.

One week later send out the survey (open for one week) via all your available channels, with a follow up reminder 3 days later if possible. If the response rate is too low possibly consider trying to boost it via additional methods such as facebook advertising. At the end of the week close the survey.

2.3 Analysis

Although the underlying statistics in case case seem quite simple (see 3.5), I would suggest letting statisticians without borders or other experts do the analysis, according to the pre-registered methodology. This would just involve looking for difference between the control and leafleted group in the rate veg*n dietary change, such as increased number of people reducing meat consumption. (Again see 3.5)

3.Explanation of the Details

3.1 How to split the population into two groups

I was inspired to come up with this after realising that approximately half of the Cambridge colleges allow mass-pigeonholing all of their students (and all cambridge students have a pigeonhole), and the other half none, creating a natural division. However there are actually several factors that make Cambridge colleges a non-ideal partition for this survey and whilst it turns out that the Cambridge Student union would not send out a survey in any case, these considerations also apply to any other collegiate university that such a survey might be run at:

 

i. Different colleges have different cultures and institutions, e.g. different cafeterias, which may serve differing qualities of vegetarian/vegan food, influencing the effectiveness of leafleting.

ii. Demographic confounders, as different colleges have different subject and gender ratios, which may correlate with leaflet response rates. I consider this a less important consideration as it could be controlled for with careful study design. (Such as by asking about gender and subject in the survey)

iii. An additional point against Cambridge is that quite a lot of pro-veggie leafleting has already been carried out, including mass-pigeonholing of many colleges in previous years.

iv. Any form of clustered randomisation reduces statistical power, although I am not sure about the size of this effect (see 3.5)

 

However more generally I think the important criteria for good ways of splitting the student population into two groups are:

 

i. It being possible to selectively leaflet all of the students in one group and none in the other.

ii. Which group any student falls into being easily discernible via a simple survey question which does not rely on the student remembering being leafleted/could influence the students’ later answers.

iii. Splitting the student population into approximately equal sized groups to maximise statistical power.

iv. The splitting being random, or at least not correlated with anything that should affect the response to leafleting.

 

With these conditions in mind I think the ideal method might be to find a university where all students could in theory be leafleted, and then selectively only leafleting those whose surname starts with a letter in the first half of the alphabet (e.g. a-m) or some similar system. A potential issue with this could be that last name starting letter may influence the base chance to go vegetarian (as it could reflect class or similar distinctions), however when splitting the population into only 2 groups this seems unlikely to be an issue.

 

As names are written on all student pigeonholes I would think most randomisation systems would need to depend on them, however there is a trade off as the more random a system the harder it would be to implement when actually doing the leafleting.

3.2 Survey Questions

I think the important considerations in choosing what questions to put on the survey would be:

i. Finding out something which we can actually put a value on, such as people going vegetarian (as opposed to say, changing their views about meat, where the impact is much less clear).

ii. Minimising the chance of influencing the answer, which would be an issue if you were  e.g. asking them about if they had been leafleted, making them think back to it.

iii. Within a given effect, choosing the question that maximises statistical power.
iv. Keeping the survey short, to maximise response rate.

 

In 2.1 I reused one of the diet questions from MFA 2013 study on leafleting [3], cutting out the irrelevent options and and adapting them to a much shorter (two weeks vs 3 months) timescale, which should greatly increase the statistical power. This is however at the cost of capturing longer term effects, such as the leaflets not having a direct effect but making people more susceptible to other veg*n outreach, which is discussed further in the next section

3.3 Practical survey considerations

I think actual leafleting should be carried out as rapidly as possible, ideally over a one or two day period, so that the amount of time people have to pick up and read the leaflet is as uniform as possible before the survey goes out.


One week after distributing the leaflets a survey (open for one week) should be sent out to the entire student population, consisting of one question about any changes they may have had in diet over the last 2 weeks, and then another to determine if they fall in the leafleting group or the control. I chose this as I think one week is enough for everyone to have checked their pigeonhole, and I think most people will then either read the leaflet immediately or at least in the same day, or throw it away. Any diet changes then have a few days to kick in before the survey goes out.

 

I chose this short turn around to maximise the power of the survey, as the shorter the timescale the lower the base rate of people converting to vegetarianism, and so the greater the statistical power. I suspect this is a trade off between power and looking at the long-term effects however, and probably cuts off some important effects. It might be worth conducting a follow up survey a few months later, to see both if the initial effects last, which would be very useful in estimating the value of leafleting in its own right, and also to see if any additional ones manifest.

3.4 Incentivisation

It is possible that incentivising the survey to boast response rate would be worth it, however it would significantly add to the cost of the survey, and I am unsure of the benefits. I have not been able to find any conclusive research as to how it would influence the expected response rate, with some papers even finding incentivized surveys got lower response rates.

 

Even if it does not boost response rate, it seems that incentivising the survey would increase the willingness by university-wide groups such as the student union to share the survey. This is very important as if we cannot get the survey emailed out to the entire student population evenly then this would not only reduce the sample size, but also mean the we may end up with a non-representative sample.

 

If the survey was incentivised I would suggest offering a reward of say a £5 amazon voucher to each student who fills out the survey, plus a bonus £2 for every other students you refer who fills it out. The referral bonus should hopefully engender a strong sharing amongst friends affect, which should lead to a high response rate, although I have found no studies on this, so it is only intuition.

 

A survey company may need to be brought in to create and set up the referral and reward systems, which could add to the costs. As part of this some system would need to be in place to stop people filling out the survey multiple times (such as requiring a unique university email address for everyone filling out the survey, or using a university authentication system as discussed in my plan for Cambridge).

3.5 Statistical power estimations

The power estimates are quite strongly dependant on what the base rate of students going vegetarian/reducing meat consumption are. I have looked at looked at ACEs 2013 study to get the proportion of students going vegetarian in a given two week period as 0.234% [4]

 

For all the following power calculations I used: https://www.stat.ubc.ca/~rollin/stats/ssize/b2.html

 

Note that if using university colleges or similar large groups as the way to split the students then the power would be reduced due to clustering [5], but I do not know how to estimate this effect. The following calculations are thus for a system like the name-based randomisation described at the end of 3.1

 

For simplicity this is looking only at people going vegetarian as a result of reading the leaflet, although as a study like this would likely be investigating multiple hypotheses such as the effects of the leaflets on vegetarianism and veganism separately, I have made a bonferroni correction [6] of â…• meaning the following are calculated for alpha = 0.01


To get the following figures I plugged the base rate of 0.234% and samples sizes into the above calculator, found the smallest effect size in the treatment population that gave 90% power, and subtracted the base rate to get the detectable effects of the leafleting. [7]

A sample of 10,000 would give a 90% chance of finding an 1/202 effect if it existed (i.e would provide a 90% chance to finding out if one in every 202 leaflets turned a student vegetarian)

A sample of 5,000 would give a 90% chance of finding an 1/124 effect if it existed

A sample of 2,500 would give a 90% chance of finding an 1/73 effect if it existed

(All assuming the sample was split evenly between control and treatment groups)

The average unversity size in the UK seems to be about 20,000 students making survey sizes of 10,000, 5,000 and 2,500 represent approximately 50%, 25% and 12.5% response rates respectively.

 

All these figures are exceedingly rough, and more could be calculated to find the power to detect e.g. conversions to veganism or general meat reduction,  but they serve to show that with a relatively low response rate (12.5%) we could attain significant statistical power, particularly more than enough to test some existing claims, such as the 1/50 or 1/75 figure leaflets to a vegetarian figure [8]

4. Final Thoughts
4.1 Limitations

A few things to consider are that the results of a study like this would not be obviously an immediately generalizable to other cases. Leaflets in pigeonholes could have a very different effect than when handed out normally. Student populations are also probably far more likely to respond positively to pro-vegetarian messages, although as most of the outreach currently seems to target them anyway this is not much of an issue.

 

There is also a significant range of values in which it would be effectively impossible to tell if leaflets had an effect, but they could still be considered very effective if they did. E.g. if 1/10,000 leaflets made someone turn vegetarian this would work out to approximately £600 per vegetarian created [9] which some might consider worth it (if they wanted to e.g. offset their own diet), but would also require such a large enough sample (~7 million people) [10] as to be effectively untestable.

 

As such I think a likely outcome of a study like this would be finding no statistically significant effect, but animal charities continuing with leafleting regardless, which might well be justified.

4.2 Predicted cost

If no incentive was used with the survey, then the cost of the entire could be extremely low, basically just that of the leaflets, approximately 4p each in the UK, and 20-30 person hours of doing set up and handing out the leaflets. At 10,000 leaflets and £10 an hour this would put the entire cost of the study at £700 ($900) or less.


If the suggested £5 per response and £2 referrals incentives were supplied, despite being a fairly large reward per person this would allow for a 10,000 person (representing a 50% response rate for a typical university, if all the students could be emailed the survey) study to be conducted for around £75,000 ($90,000), which would have orders of magnitude more power than any other study conducted to date.

4.3 Conclusion

Despite some subtleties about exactly how to split up the groups and maximise response the response rate to the survey I think there is a strong opportunity here, and that a study run using this methodology would be significantly valuable.

 

The main issue seems to be the availability of universities with pigeonholes and cooperative faculties/student unions to email out the survey. As such I would be interested in hearing from students at any universities they think fit the criteria.

In the meantime please give me feedback on any areas you think are flawed, or ways things could be improved.



[1] https://medium.com/@harrisonnathan/re-evaluating-animal-charity-evaluators-c164231406f7

 

[2] e.g Mercy For Animals, one of ACE’s top recommended charities pays its fellows to do this.

 

[3] https://animalcharityevaluators.org/research/interventions/leafleting/2013-leafleting-study/#main


[4] As ACE published all the data from their study [3], I was able to look at the ‘control’ responses from people who did not receive a leaflet, or who received one unrelated to reducing animal product consumption. Removing all those who did not answer the relevant questions, I was left with 356 people who filled in their present diet, as well as what it had been 3 month previously.

Counting the number of vegetarians who had reported being something other than vegetarian 3 months previously (and excluding one vegan who because vegetarian) I got a base rate of 5/356 students becoming vegetarian in a 3 month period, or â…™*5/356 = 0.234% becoming vegetarian in a two week period.

 

[5] For an explanation of clustering see e.g. http://www.healthknowledge.org.uk/public-health-textbook/research-methods/1a-epidemiology/clustered-data

[6] https://en.wikipedia.org/wiki/Bonferroni_correction There might well be better ways of doing this, I just went with something I could easily use.

 

[7] In detail for the n = 10,000 case:
I entered the figures into the calculator as so:
statistics blank.png

I then entered values for p2 until finding the lowest that gave a power of 90%, which turned out to be 0.0073:

statistics.png

This corresponds to the leaflets having an effect of p2 - p1 = 0.00496 = 1/202.


[8]https://ccc.farmsanctuary.org/the-powerful-impact-of-college-leafleting-part-1/

 

[9] When mass-leafleting colleges as in this study as a team of two we were able to hand out 1000 leaflets an hour. I have been told the leaflets cost 4p each, and assuming paying the leafleters £10 an hour this gives a total cost of £600 (2*10*£10 = £200 to pay the leafleters, and $0.04*10,000 = £400 for the cost of the leaflets).

 

[10] obtained by setting p2 in the calculator to 0.0021, 0.0001 (or 1/10,000) more than the base rate of p1 = 0.002, with a power of 90%.

Comments (17)

Comment author: jimrandomh 06 February 2017 05:09:48PM 5 points [-]

The questions about diet before and after the change seem to be pushing people strongly into claiming to be or to have been some sort of vegetarian; the only option you have there that isn't somehow anti-meat is "Other', which requires typing.

A better version of this question would have a no-dietary-restrictions option first, and a few options that aren't animal-welfare related like "low carb" and "Mediterranean".

Comment author: Alex_Barry 06 February 2017 05:33:01PM 1 point [-]

Yeah that seems right, I think it would not be too bad as they would not be shown the options until having already reported changing their diet, but it would be worth adding more options as you describe.

Comment author: jimrandomh 06 February 2017 05:05:34PM 5 points [-]

Statistics nitpick: I believe you should be using a two-sided test, as it is also possible for leafleting to reduce the rate of people going vegetarian if the leaflets alienate people somehow.

Comment author: Alex_Barry 06 February 2017 05:31:14PM 1 point [-]

Ah yeah, my mistake, thanks! I'll update the post.

Comment author: Derek 10 February 2017 03:21:12PM 3 points [-]

Good to see some thought being put into this topic. A few comments/concerns below.

To pseudo-randomise based only on names, I wouldn't use the first letter. I haven't checked, but I'd guess that some strongly correlate with confounders like ethnicity or class: many starting with an X could be Chinese, with a Z Arabic, etc.

Instead, you could assign based on whether they have and odd or even number of letters in their name. It seems unlikely that this would be correlated with confounders, or that there would be very uneven numbers in each group, but I suppose it's possible, e.g. if the most common few names all happen to be even (or odd) and also be associated with class/ethnicity/sex. Using their full name would mitigate this but would slow down leafleting compared to only using first or last, and perhaps introduce counting errors. Maybe you can get a fairly random sample of student names from somewhere and check whether using just one name is likely to be problematic.

Obviously you need to ensure that the name they use on the survey is exactly the same as the one on the pigeonhole (Elizabeth/Beth/Liz/Lizzy; middle name included or not, etc), unless you specifically ask whether they received a leaflet, which introduces other problems. I would probably suggest adding to the questionnaire something like: "(Please write your name exactly as it is on your pigeonhole.)". Presumably family names are less likely to be modified/dropped, but as noted, only using those may hinder randomisation.

I'd guess that many of the leaflets would be seen by the control group. People will show them to their friends, leave them lying around, etc. This would dilute the observed effect of the intervention. I'm not sure how to avoid it without going for cluster randomisation, which is even more difficult to get right. I suppose it would at least give you some basis for saying your findings are conservative; Type 1 errors (false positives) are generally seen as worse than Type 2 errors (false negatives), and given the state of evidence for animal-focused interventions, it is probably wise to be cautious. There would still be some factors pushing the other way though, so I'm not sure it would actually be conservative.

Opportunity permitting, I would consider some follow-up research to establish mechanisms. For example, you could ask (a random subset of?) changers why they changed, whether they remember seeing the leaflet, whether they showed it to others, what they thought about it, whether they were planning to change anyway, etc. This could increase (or decrease) confidence that the result shows a true effect. You might be able to get one or two of these into the original survey; to avoid influencing responses, use a web form that requires them to submit the 'change' answers before viewing subsequent questions - this is quite a common design in my experience. You could also include a social desirability instrument, as in the ACE study.

The use of incentives could introduce response bias. Presumably the extent to which money motivates response will be associated with SES and perhaps ethnicity and other characteristics. Could still be justified, though, in order to boost response. Not sure. Given the huge impact this has on costs, it might make sense to do an unincentivised one first, then perhaps add the incentives in later studies (obviously at different institutions). This would also function as a cheap 'test run', allowing you to refine the methodology and logistics.

I suspect the referral system would introduce considerable sampling bias. Those who pass it on, and perhaps those who are friends with referrers and who respond to a referred task, are unlikely to be representative of the study population: they'd presumably be more conscientious, have more friends, have more interest in dietary change, etc. It seems odd to go to all that effort to randomise then include a method that would undermine it so much. I'd only do it if it was impossible to get an adequate response otherwise.

Likewise publishing articles about it in advance. Those who read them may not be representative, e.g. more likely to be native English speakers, conscientious/motivated, interested in dietary change (or study design), etc. Some people's decision to go veg could also be influenced by the articles themselves. And obviously you couldn't publish them if you were trying to conceal the purpose of the study as others have suggested.

When calculating sample sizes, I'm not sure it's a good idea to rely heavily on one estimate of base rate change. Presumably it fluctuates considerably in response to time of year (new year's resolutions), news stories, other campaigns, etc., and correlates with location and many other factors. It would be a shame to waste resources on an under-powered study.

I would be a little surprised if the true effect was more than 1/200 for any dietary change (at least a durable one), and not surprised if it was orders of magnitude smaller. If you want to look at subgroups (vegan, reducetarian, etc), and maybe even if you don't, I'd guess you'd need a much larger sample than proposed. But these are just hunches.

As usual with trials (and life), it seems there would be a number of hard trade-offs to be made here. I suppose whether it's worth doing depends on what results you expect and how they would influence action. To play devil's advocate: You think a null result is likely but that it wouldn't justify stopping leafleting, in which case the study would not have been helpful in terms of allocating resources. Depending on the size and quality of the study, it would also be unwise to put much weight on a weak positive or negative effect. A strong negative is unlikely and in such a case we should probably conclude that there was a problem with these particular leaflets or with the study design (or that we just got really unlucky). So only a strong positive (>1/150?) seems like a very useful outcome. Even then, we should still suspect influence from chance, bias and/or confounding, but it would give some justification for continuing, and perhaps increasing, leaflet campaigns; and for investing in further studies to confirm their effectiveness and investigate what strategy works best, for whom, why, and in what circumstances. However, a strong positive seems unlikely (<15%?). Therefore, perhaps the question is whether it is worth doing the study if there is only a small chance of getting a result that would (or should) change our activities.

I suspect the answer to the last question is "yes", and I'm not actually sure the other results would be useless (e.g. a null or negative result might prevent inordinate future investment in leafleting based on poor quality past studies). There are also other benefits to running the study, like boosting the credibility of, and skill development within, the movement. But it's worth thinking carefully about what exactly we expect to get from the study before leaping into it.

Comment author: Peter_Hurford  (EA Profile) 08 October 2018 05:12:04PM 1 point [-]

Instead, you could assign based on whether they have and odd or even number of letters in their name.

You could SHA-256 hash the names and then randomize based on that. Doing so should remove all chances of confounding effects.

Comment author: Derek 11 October 2018 09:15:40PM 0 points [-]

It's been a long time since I wrote the comment, but I think I was under the impression the allocation had to happen at the point of distribution, using only the names on the pigeonholes. But if you could get a list of students that exactly matched the names on the pigeonholes in advance of distribution, then I agree randomising hashes would be ideal. I doubt you'd get this from administrators due to data protection issues, but presumably you could go round and manually record the names. That would be very time-consuming for a large study, but perhaps worth it to avoid any doubts about the randomisation.

Note that this would not remove all risk of selection bias because the allocation would still not be concealed, i.e. the people putting the leaflets in the pigeonholes would know who was assigned to each group. It is possible they would consequently give some of the leaflets to the wrong people, e.g. if they wanted to increase the effectiveness of the intervention they might be influenced by personal knowledge of individuals (better to give a leaflet to someone who is not already vegan) or presumed characteristics (gender, ethnicity, etc) that are correlated with likelihood of being influenced by the leaflets. This may seem far-fetched, but unconcealed allocation is associated with higher effect sizes in medical trials so we shouldn't be too quick to assume it wouldn't matter here.

https://handbook-5-1.cochrane.org/chapter_8/8_10_1_rationale_for_concern_about_bias.htm

https://handbook-5-1.cochrane.org/chapter_8/8_10_2_assessing_risk_of_bias_in_relation_to_adequate_or.htm

One solution is to give every student a leaflet inside an opaque sealed envelope, with some getting the 'treatment' and some a 'control'. But this introduces additional complexity, e.g. it could cause (further) 'contamination' as students compare what they got in that weird envelope, it reduces external validity (assuming leaflets would not normally be in envelopes), and the control leaflet would have to be very carefully designed so as not to affect the outcomes of interest while being very similar in every other respect (e.g. a leaflet promoting exercise may indirectly influence dietary choices).

Comment author: Jeff_Kaufman 08 February 2017 02:46:14AM 1 point [-]

"I reused the diet questions in my plan from MFA 2013 study on leafleting"

In my view, this study asked way too much. When you try to ask too much detail people drop out. Additionally, it asks about things like diet change, but to pick up on changes we should be comparing the experimental and control groups, not comparing one group with its (reported) earlier self.

What I'd like to see is just "do you eat meat" along with a few distractor questions:

  1. Are you religious?
  2. Is English your native language?
  3. Do you eat meat?
  4. Do you own a car?

Yes, we'd like to know way more detail than this, and in practice people are weird about how they use "meat", but the main issue here is getting enough responses to be able to see any difference at all between the two groups.

Comment author: Alex_Barry 08 February 2017 04:56:42PM *  0 points [-]

"I reused the diet questions in my plan from MFA 2013 study on leafleting"

Ah sorry again I was not quite clear, what I meant by this was the question about diet is one I had copied from the MFA study, not that I would reuse all of them. The questions I list in 2.1 are the only ones I would ask (probably with a bonus distractor question and maybe some extra options as suggested by jimrandomh).

Additionally, it asks about things like diet change, but to pick up on changes we should be comparing the experimental and control groups, not comparing one group with its (reported) earlier self.

Asking about 'change in diet' vs just diet generally is basically required to get sufficient statistical power, as the base rate of people saying yes to "have you become vegetarian in the last two weeks" is much much lower than "are you vegetarian" but the effect size we are looking for in each case is the same. One can then compare the control and experimental groups on this metric.

To illustrate the size of this effect, in the post I calculate that with a sample of 5000, by asking about change in the last 2 weeks you would have a 90% chance to find an effect of 1/124 leaflets creating one vegetarian, but if you just asked "are you vegetarian?" you would only be able to find a 1/24 effect at the same power. (Assuming a 20% base rate of vegetarianism, and using this calculator ).

When you try to ask too much detail people drop out.

I agree about using as few questions as possible, and that the MFA study asked far too much (although I think it was administered by volunteers as opposed to online, which would hopefully counteract the drop out effect in their case).

Comment author: jimblowiak 12 February 2017 12:56:22PM 1 point [-]

I was wondering if you planned on submitting this proposal to ACE?

http://researchfund.animalcharityevaluators.org/about/

They're currently funding research like this. I'm actually spending my Sunday morning trying to figure out the relative benefit of additional research (in animal welfare it would be tremendous I think, since the existing body of evidence leaves much to be desired).

The next question is how can we fund this sort of research effectively? While ACE is running their funding operation, and I think it's extremely smart of them to do so, it'd be a good hedge to have multiple gatekeepers.

Comment author: Jeff_Kaufman 08 February 2017 02:36:55AM 1 point [-]

"A question to determine if they were leafleted or not, without directly asking."

People who were leafleted but ignored it and don't remember enough to answer this one accurately is a problem here.

What would you think of: at a college that allows students to mass pigeonhole directly, put experiment leaflets in odd mailboxes and control ones in even boxes. Then later put surveys in the boxes, with different links for odd and even boxes.

Instead of having the links be example.com/a and example.com/b it would be better for them all to look like example.com/gwfr so people don't know what's going on. You could generate two piles of follow-up links and use one for the odd boxes and the other for even. QR codes might be good to add so people have the option not to type.

Comment author: Alex_Barry 08 February 2017 04:48:36PM 0 points [-]

People who were leafleted but ignored it and don't remember enough to answer this one accurately is a problem here.

Ah my wording might not be quite clear in that section, ideally the question would rely on the surveyors knowledge of who was leafleted and who was not, without the students having to remember if they were leafleted. E.g. if the leafleting was split by college, asking what college they were from.

Then later put surveys in the boxes, with different links for odd and even boxes.

A reason I would push for trying to get a survey emailed out is that a previous study on leafleting that attempted to use QR codes or links on leaflets got an extremely low (about 2% if I remember correctly) response rate.

I am not sure if giving out the surveys desperately later would boost the response rate, I think the added inconvenience of having to type in a web address or scan a QR code would significantly damage the response rate.

Still it would be worth bearing in mind as strategy that could currently be implemented at any university that allowed mass-leafleting, without being able to email all students.

Comment author: LaurenMcG  (EA Profile) 25 June 2017 05:35:58PM 0 points [-]

Do you have updates on the impact of any recent university leafleting activities?

Comment author: rossaokod  (EA Profile) 12 February 2017 05:26:46PM 0 points [-]

Thanks for writing this up! I think this type of research is very important in the animal advocacy movement.

One thing I would like to see in one of these studies is using food purchase data instead of relying on surveys. This might be possible in colleges/universities where a large proportion of students have some sort of meal plan, and pay for their food using their student ID or some other similar card. In some places the students just pay to enter a food hall, and so you wouldn't know what food they ate, but in other places I think each item is scanned individually. This approach would obviously require the university and/or food service company being supportive of this type of work, and being willing to share the data. They would of course also have to anonymize the data and it might be a challenge getting ethics approval.

I see two main benefits of this type of data. (1) it is likely much more accurate than survey data and (2) you are likely to have far more power in your statistical tests. There are a few reasons I believe (2): there would be less noise in the data (i.e. even if self reports are unbiased, they will likely have measurement error), you would likely have a larger sample size (no problems with response rates), and you could use continuous measures of meat consumption (i.e. 'number of food items bought with meat over a week/month' rather than just 'do you eat meat?')

Comment author: Bernadette_Young 06 February 2017 08:34:49PM 0 points [-]

Ethics approval would probably depend on not collecting identifying data like name, so it would be important to build that into your design. College name would work, but pseudo-randomising by leafleting some colleges would introduce significant confounding, because colleges frequently differ in their make up and culture.

Comment author: Alex_Barry 06 February 2017 09:17:15PM *  1 point [-]

Yes I mention the issues associated with college-based randomization in section 3.1.

Good point about not collecting identifying data, it should just be possible to ask for whatever information was used to decide who to leaflet, such as first letter of last name, which should avoid this issue.

Comment author: Bernadette_Young 06 February 2017 09:45:29PM 1 point [-]

Ah sorry, I missed that bit!