Background

Vegans are the backbone of farmed animal advocacy. Even in a time of increased focus on corporate lobbying campaigns, nearly every advocacy group is putting resources toward getting people to “try veg,” “go vegan,” or “leave meat off your plate.” Many of these campaigns incorporate lessons learned through research and impact evaluation, as there is now a fair bit of well-conducted research exploring how to create new vegans and vegetarians (collectively calledveg*ns in this report).

At the same time, research about how to assist and retain new veg*ns—particularly those who adopt a dietary change outside the context of an advocacy campaign—is very limited. Improving our understanding and support of new veg*ns is particularly important as the public’s exposure to plant-based eating and veganism continues to increase. We do not want people to treat vegetarianism and veganism as fad diets to be tried and discarded.Faunalytics (2014) showed that there were far more former veg*ns in the U.S. than current ones, many of whom tried the diet for only a short time. 

Upcoming Reports

The purpose of this study is to provide solid data for advocates about how to turn those short-term veg*ns into permanent veg*ns. This is the first of several reports that will come out of this study.

• The second report, coming in a few months, will look at people’s general and specific motivations for starting the diet, how those relate to success, and their level of speciesism at the beginning and end of the six-month period. 
• The third report, coming later this year, will focus on the crucial question of the effectiveness of various strategies for overcoming barriers to staying veg*n.

This project has produced a huge amount of data, all of which will be posted on the Open Science Framework once we have completed our own analyses and publications. In the meantime, if you have additional research questions that you would like us to consider, please contact[email protected].

Participants 

This study includes 222 members of the general public in the U.S. and Canada, all of whom had started transitioning to a vegan or vegetarian diet within the past two months. 

The Level of Commitment section of this report shows that more than 90% of the sample said they would probably or definitely continue their new diet change permanently. Therefore, it may be best to consider this sample representative of people who have already moved beyond a simple interest or desire to change into the stage where they are ready to actively work toward a veg*n goal.

Key Findings

1. Most people transition to veganism or vegetarianism gradually. Just 21% of people went veg*n overnight. 38% planned to transition over a few days or weeks, 34% over many weeks or a few months, and 7% over many months. After six months, people who transitioned more slowly tended to feel less successful and be further from their goal diet than those who transitioned more quickly—but they were just as likely to continue with it as the others.
2. For people who reduce gradually, we found no significant evidence that any one reduction method works better than another. Whether people reduced their total consumption a bit at a time, their consumption of particular foods one at a time, or a combination of the two, rates of success were not significantly different. 
3. When going vegan or vegetarian, imperfection is the rule rather than the exception. On average, by the end of the six-month tracking period, 88% of participants felt successful but were still consuming 6.1 more monthly servings of animal products than they intended. Further, just 28% of participants felt completely successful, although 57% had met or surpassed their goal level of consumption.
4. The typical person reduced their animal product consumption by 42.1 monthly servings over the first six months or so of going veg*n. That is, they ate 42.1 fewer servings of animal products per month by the end of the study than they were before they began their dietary transition. For new vegetarians, this meant going from about 15 servings a week to under 6. For new vegans, it meant going from about 12 servings a week down to one serving per week and a half—not perfect, but getting close to eliminating all animal products.
5. Choosing veganism over vegetarianism appears to be more common than it used to be. In this general population sample, 41% of people were working toward veganism, 59% toward vegetarianism (1.4 times as many). We compare this against our 2014  study showing that at the time, there were 3 times as many vegetarians as vegans. (At the same time, methodological differences between the two studies could exaggerate the change, so the extent of it  should be considered uncertain.)
6. People with spouses or children tended to be further from their veg*n goals after six months compared to unmarried people without children. These findings point to some key competing demands and indicate that they may make progress more difficult. 

Recommendations

• Encourage an early, specific commitment. Participants’ level of commitment at the beginning of the study (i.e., how sure they were that they would continue the diet permanently) was a significant predictor of diet maintenance vs. abandonment. We recommend designing aneffective pledge for your outreach campaigns to encourage accountability and commitment. 
• To fit individual lifestyles and needs, use tailored pledges rather than one-size-fits-all. Individuals vary a lot. As we saw in this analysis, there are many successful routes to reducing animal suffering through diet, even for people with the same end goal. For instance, eliminating all animal products at once, eliminating foods one at a time until reaching a goal level, or increasing the number of veg*n meals per week until reaching a goal level. We recommend that you first determine individuals’ preferences, then follow the effective pledge guidelines to have them commit tothat specifically tailored pledge, not a general one. 
• Acknowledge and validate the ways that personal circumstances can make change difficult and provide appropriate support if you can. This study identified family circumstances (having a spouse and/or children) as a key barrier for many people. We suggest learning more about these barriers from those who experience them and by reading about other research on the subject (e.g.,Asher & Cherry, 2015;Greenebaum, 2018). Helping people find ways to overcome barriers is important, but it is also important just to understand and validate the difficulty they pose so that people feel supported in their journey. (This recommendation is drawn from research on how responsiveness supports goals; e.g.,Canevello & Crocker, 2010;Feeney, 2004).

Research Team

The project authors are Jo Anderson (Faunalytics), Marina Milyavskaya (Carleton University), and Marta Kolbuszewska (Carleton University). However, this project was a massive undertaking and could not have happened without the support of multiple individuals and organizations. 

We are very grateful to Faunalytics volunteers Renata Hlavová, Erin Galloway, Susan Macary, and Lindsay Frederick for their support and assistance with this work, as well as the dozens of animal advocates who helped with recruitment. We are also very thankful to VegFund, Animal Charity Evaluators, and the Social Science and Humanities Research Council (SSHRC) for funding this research. Finally, we thank all of our survey respondents for their time and effort.

Method Overview

This project focused on the experiences of new vegans and vegetarians (for simplicity referred to collectively as veg*ns in this report) in the U.S. and Canada. Participants were asked to complete a survey when they signed up to participate, as well as six follow-up surveys that were sent monthly over the next six months.

These are not people who all saw the same advocacy campaign or signed up for a challenge, so their experiences and reasons are diverse and as representative as possible of the wide range of paths to veg*nism that exist. We found that our participants’ demographics were also quite representative of the general population, but we also weighted the descriptive results to be even closer to the U.S. population. For details, see the Supplementary Materials.

We recommend that these participants be considered representative of the sorts of people who begin new veg*n diets in the U.S. and Canada, with one exception. As with any study, it is important to consider whether some people are more likely than others to participate. In this case, we strongly suspect that people who are just experimenting with a veg*n diet and are fairly uncommitted to it are probably not represented in this sample because it would feel strange to sign up for a six-month study of something you might not continue for more than a few weeks. Their reported level of commitment supports this assumption, as described on the Key Findings tab and in the Level of Commitment section below.

Representativeness And Weighting

To ensure that this sample is as representative of new veg*ns as possible, we followed apre-registered plan of comparing them against a much larger sample (n = 11,399) of veg*ns fromFaunalytics’ 2014 study. We were pleased to find that the current sample matched most of those demographics well already, but to maximize the representativeness, we weighted the descriptive results to match. For more details of the comparison and weighting procedure, see the Detailed Method section on the Supplementary Materials tab.

Attrition

Overall, 65% of the participants completed the entire study: either by making it to the end of the six-month study period still pursuing their goal diet or by completing a “quit survey” (though it wasn’t described that way to them).

As we had expected, completion rates for the individual follow-up surveys declined steadily but slowly over time, with 77% completing the first follow-up, down to 56% completing the sixth and final follow-up. Although these are good retention rates for a longitudinal study, it is always important to consider the question of whether the people who dropped out of your study differed from those who stayed in (known asdifferential attrition).

We examined the attrition carefully using apre-registered method and found that there did not appear to be significant differences between people who completed the study and those who didn’t. You can read more about this approach and the results in the Detailed Method section on the Supplementary Materials tab.

People who abandoned their veg*n goal were asked to complete a “quit survey.” Just 19 of our 222 (9% unweighted) participants indicated that they had abandoned their veg*n goal over the course of the study.

Results

This study’s pre-registration, survey instruments, analysis code, and data are available on theOpen Science Framework.

For graphs with error bars, the error bars represent the 95% confidence intervals.

Frequency of Vegan and Vegetarian Goals

Goal Diet

As you can see in the figure below, a vegetarian goal was more common than vegan but the split was more even than we observed in 2014, when there were 3 times more current vegetarians than vegans, and 9 times more former vegetarians than former vegans.

Figure 1. Goal Diet

Goals Around Other Animal Product Usage

For many people, veganism is a lifestyle, and a plant-based diet is just one part of it. Therefore, we also asked participants whether they had a goal to reduce or avoid using any other animal products. The examples provided were food additives (animal-based coloring, etc.), clothing (wool, leather, etc.), and hygiene products (glycerine, lanolin, etc.).

59.1% of participants said they also had a goal to reduce or avoid other animal products. Details are shown in the table below.

Table 1. Other Animal Product Goals

As you can see in the table above, people following a vegan diet were (unsurprisingly) more likely to avoid other animal products than vegetarians (𝛸² = 10.8,p = .001). However, not all of them were doing so. And on the flip side, vegetarians were evenly split in whether they had a goal to reduce other animal product usage.

Overall, this table shows that about half of the participants we refer to as “vegan” would be described as “plant-based eaters” by some. While we understand the distinction, it appears to be more focused on emphasizing vegans’ moral identity than providing a welcoming environment to people who are moving in the right direction, which may reduce motivation to continue. As a result, we have chosen not to distinguish these groups in this report.

Level Of Commitment

A common interpretation ofFaunalytics’ 2014 study’s high abandonment rate is that many people “try out” veganism or vegetarianism, but without a strong intention to necessarily stick with it. To get at this idea, we asked participants whether they intended topermanently avoid meat (for vegetarians) or animal products (for vegans).

As you can see in the figure below, most people didn’t specifically plan to give up the diet after a trial run, but 40% were not definite in their commitment. Even so, we interpret this as indicating that the results of this study are most representative of people who are reasonably committed to becoming vegan or vegetarian, unlike the participants in Faunalytics (2014).

Table 2. Commitment To Goal Diet

There wasn’t much difference in the commitment level of people pursuing a vegan diet versus vegetarian diet. On the scale shown in the table above, if we treatdefinitely yes as a score of 5 anddefinitely not as a 1, the average vegan’s commitment was 4.6, versus 4.5 for the average vegetarian (t = 1.89,p = .06).

Previous Diet

The figure below shows participants’ dietary origins: that is, the diet they had followed immediately prior to starting their transition to veg*nism. As you can see, there was a range, but the majority were self-described omnivores beforehand. This was lower for vegans, 30% of whom were transitioning from vegetarianism.

Figure 2. Previous Diet

The two following figures show how much of different animal products new vegans (first figure) and vegetarians (second figure) consumed prior to beginning their new diets. As you can see, there was a wide spread of consumption before they began their reduction efforts.

Figure 3. New Vegans’ Animal Product Consumption Prior To Change

Figure 4. New Vegetarians’ Animal Product Consumption Prior To Change

Transitioning To Veg*nism

As shown in the figure below, most people transition to veg*nism somewhat gradually, though the speed varies a lot.

Figure 5. Transition Speed

New vegans and vegetarians both transitioned at a variety of speeds, with no significant difference between the two groups (t = 1.20,p = .23).

Even among the people who were transitioning gradually, there was a range of approaches, as shown in the table below.

Table 3. Methods Used By Gradual Transitioners (Those Who Didn’t Transition Overnight; n = 174)

Previous Dietary Change Attempts

Previous Veg*n Attempts

In our 2014 study, we found that 31% of people had tried a vegan or vegetarian diet more than once. In this study, 40.5% of participants were attempting the diet for a second, third, or even fourth+ time, as shown in the figure below.

Figure 6. Repeat Attempts At Veg*nism

People pursuing a vegan diet were significantly more likely to have attempted a transition to a vegan or veg diet in the past and given it up, while most vegetarians were attempting the diet for the first time (𝛸² = 8.5,p = .004).

Fad Diet Attempts

Similarly, we expected that some people who try a vegan or vegetarian diet approach it as a fad health diet, similar to others like Atkins, paleo, or keto. We asked our participants whether they had tried other diets before in order to get a sense of this group. As you can see below, this was a small but notable subset of the sample.

Table 4. Fad Diet Attempts

People pursuing a vegetarian diet were marginally more likely to have tried a fad diet than people pursuing a vegan diet (𝛸² = 3.5,p = .06).

We also looked for any association between fad diets and previous attempts at veg*nism. There was no evidence that people who had previously tried a fad diet were any more or less likely to have tried to go veg*n more than once (𝛸² = 0.2,p = .68).

Successfully Transitioning To A Vegan Or Vegetarian Diet

We measured dietary success in three ways, described in more detail below:

1. Diet maintenance versus abandonment (whether participants explicitly abandoned their diet goal or not),
2. Felt success (as a self-reported percentage), and
3. Consumption success (operationalized as distance from goal diet, in monthly servings).

Diet Maintenance Versus Abandonment

In our2014 study, Faunalytics reported that there are five times as many former vegans and vegetarians in the U.S. as there are current ones. Or put another way, that 84% of people who have ever followed a veg*n diet gave it up.

In this study, we looked at success in several ways. One seemingly simple way—maintaining a diet for six months versus abandoning it within that time period—is roughly parallel to the 2014 study, though the 2014 results include people who lapsed after a longer time as well as people whose diets were temporary or not by choice (e.g., some who were unable to afford meat).

Looking just at the participants in our sample, only 19 abandoned their diets within the study timeframe. After applying weights for representativeness, this suggests an abandonment rate of 9%. However, things are not quite that straightforward because we were not able to track most participants from day 1 of their dietary transition. At the beginning of the study, just 43 of 222 participants (19% unweighted) were beginning their dietary transition that week. Most had begun it within the past four weeks (n = 102; 46% unweighted) or four to eight weeks ago (n = 77; 35% unweighted).

This is important becauseFaunalytics (2014) found that approximately a third of people who tried a veg*n diet abandoned it within the first three months and we want to account for that abandonment rate. Taking that into account, we arrive at a more conservative estimate of 43% (details are on the Supplementary Materials tab). That is, with this additional assumption, the data suggests that43% of new veg*ns abandon their diets within six months. However, it is very important to note that there are several assumptions going into this calculation no matter how we do it, so these numbers should be taken with a grain of salt.

Felt Success

In 2014, we defined the vegan and vegetarian diets strictly, with people who consumed any meat or animal products considered non-vegetarian and non-vegan, respectively. However, just as the takeaways and recommendations from that study suggest, success is not a black-and-white concept. Since 2014, both Faunalytics and the wider animal protection movement have evolved away from strict “go vegan” advocacy—in large part because that study and others suggest that it doesn’t work as effectively as a more forgiving approach.

Our measurement tools have evolved along with our recommendations. Rather than defining dietary change as just successful or failed in this study, we instead focused on measuring the extent of participants’ self-reported success—with the understanding and assumption that people are rarely perfect.

Perhaps the most straightforward way to measure success is to ask a person how successful they felt. The biggest advantage of this approach is that it allows for individual differences in transition speeds and strategies. For example, if a person is transitioning slowly to a vegetarian diet, they may feel that it has been a successful month even if they are still eating a bit of meat, because that was part of their transition plan. The main disadvantage of this approach is that the results don’t indicate the number of animals affected.

We asked participants to rate their feeling of success using a percentage. The average results are shown in the figure below. An additional analysis revealed that just 28% of people felt completely (100%) successful. This included 31% of vegans and 27% of vegetarians.

Figure 7. Average Felt Success Over Time

The figure below also shows the felt success data, but with one line per participant rather than just the average. It is included to give a sense of the variance in participants’ paths.

Figure 8. Felt Success Over Time, All Participants

Nuances Of Felt Success

On the survey, we gave participants the option to “Explain your success rating if you think it needs more context.” Most people left this blank, but there were a few common themes among the explanations that we received, as shown in the table below. People tended to use this option to explain what they perceived as slips or difficulties.

Table 5. New Veg*ns’ Explanations For Success Ratings

Consumption Success

While felt success is helpful for understanding new veg*ns’ subjective experience, it does not tell us anything about actual animal product consumption. For a more objective measure of success, we calculated the difference between participants’ goal consumption of animal products and their current level of consumption in servings. More details are provided on the Supplementary Materials tab.

The figure below shows participants’ consumption success as the distance between the number of servings of animal products they reported eating and the number they should be eating according to their goal consumption.

Using this method, a score of 0 means that the distance between their goal and actual consumption is 0, so the participant has met their goal for the month. The figure shows that the average participant was still 6.1 servings from their goal by the end of the six months. However, there was a wide range of consumption success, with many participants having achieved or nearly achieved their goal, as shown in the second figure below.

An additional analysis found that 57% of people had met or surpassed their goal level of consumption by the end of the sixth month. This included 38% of people who were going vegan and 71% who were going vegetarian.

Figure 9. Average Distance From Goal (In Servings) Over Time

Figure 10. Distance From Goal (In Servings) Over Time, All Participants

Consumption Success In Terms of Amount Reduced

Another way to look at consumption success, rather than compared to participants’ goal diets, is in terms of how many animal product servings they eliminated from their diets. (Methodological note: We have used median servings for this analysis to avoid a few outliers skewing the results.)

The median participant (the most typical person, in the middle of the range) reduced their animal product consumption by 42.1 monthly servings over the first six months or so of going veg*n. The size of the reduction was not significantly different for new vegans versus new vegetarians (K-W ꭕ² = 0.01,p > .90) because as you can see in the figure below, vegans consumed significantly fewer animal products to begin with (K-W ꭕ² = 16.2,p < .001). The paths taken by both groups look quite similar.

People transitioning to vegetarianism went from eating a median of 66.7 servings of animal products per month before beginning their dietary transition to 25.9 servings by the end of the six months. That’s about 15 servings a week down to under 6.

People transitioning to veganism went from eating a median of 51.9 servings of animal products per month before beginning their dietary transition to 3 servings by the end of the six months. That’s a starting point of about 12 servings a week down to one serving in a week and a half—not perfect, but getting close to eliminating all animal products.

Figure 11. Median Animal Product Consumption (In Servings) Over Time

Consumption Success Vs. Felt Success

As you can see from the two previous figures, consumption success and felt success both improved over time and showed a similar pattern (just reversed for consumption success because it’s measured as distance from goal).

However, individual participants’ felt success was only somewhat related to how far their consumption was from their goal, suggesting that there is more to feeling successful than just getting closer to eliminating all animal products from one’s diet. For instance, maintaining motivation or overcoming a strong craving may make people feel successful even when they have a ways to go in terms of consumption. We will consider this in later reports.

The correlations between felt success and distance from goal at each follow-up time point ranged from -0.27 to -0.41, which would classically be described as a moderate association (Cohen, 1988).

Table 6. Correlations Between Felt Success And Goal Distance

Who Are New Vegans & Vegetarians?

Demographics

The table below shows the basic demographic characteristics of our sample. All of the descriptive statistics in this report were weighted to make them more representative, using data about the breakdown of the vegan and vegetarian population fromFaunalytics (2014). However, we were pleased to see that the sample was already quite representative on those weighting variables: age, gender, race/ethnicity, and motivation for becoming veg*n. More details on the weighting procedure and comparison between the sample and population are available in the Supplementary Materials.

Table 7: Basic Demographics

The table below shows additional characteristics of new veg*ns. As shown, most were employed full-time, but the proportion of students is relatively high at almost 19%.

Table 8. Additional Characteristics

Finally, the table below shows key characteristics of participants’ households, which may also be important in their transitions to veg*nism. More than half were in a long-term relationship, about 40% had children, and three-quarters had companion animals.

Table 9. Household Characteristics

Sample Journeys

In the graphic below are real details of participants from the study, but not their real names. The goal here is to provide an idea of the individual variation observed in this study in qualitative terms.

Figure 12. Profiles Of Four Participants

The only characteristic the four people above have in common is that they were all at or beyond their goal level of animal product consumption by the end of our study. They were not randomly picked as examples because we wanted to show a range of experiences, but it wasn’t difficult to find four unique examples either. When you dig into the full six-month span of data, it is clear that every experience is different. We are only scratching the surface with this report and will continue to reveal more with the next two. We encourage advocates to remember and respect the individual journeys of the people involved, because no two are the same.

Predicting Success

One at a time, we considered each of the factors listed below as potential predictors of success. That is, whether participants’ scores on these at the beginning of the study predicted their consumption success, felt success, and/or diet maintenance vs. abandonment at the end of the six months (i.e., at the final follow-up). Additional details of the regression analyses are provided in the Supplementary Materials.

Most of the factors were not related to success, no matter which measure of success you consider. And while thep-values shown below include a post hoc correction for false discovery rate (FDR), it is clear from the figures that this did not dramatically change the interpretation of the results. Below, we consider each predictor in turn, with the meaningfully predictive ones first.

Transition Speed 

Transition speed was a marginally significant predictor of consumption success and felt success (ps = .05 & .09). People who transitioned to veg*nism faster were generally closer to their goal consumption by the end of the study and tended to feel more successful, as shown in the figures below. However, there was no association between transition speed and diet maintenance vs. abandonment (p = .85).

Figure 13. Association Of Transition Speed With Consumption Success At Final Follow-Up

Figure 14. Association Of Transition Speed With Felt Success At Final Follow-Up

Figure 15. Association Of Transition Speed With Diet Maintenance Over Six Months

Level Of Commitment To The Diet 

Level of commitment to the diet—that is, intention to continue it permanently—significantly predicted whether people maintain or abandon their diets, with more committed people being less likely to abandon it (p = .03). While this does not come as a surprise, it is crucial to remember the importance of commitment in behavior change. This is why we recommend designing aneffective pledge for your campaign. Higher commitment was also marginally related to more felt success (p = .09). However, it did not predict consumption success (p = .38): That is, while commitment may protect people against abandoning their dietary transition, it does not appear to affect how closely they follow their goal diet.

In the figures, you will note that there are noprobably not responses, although they accounted for 2.7% of responses at baseline, as noted in the Level of Commitment section above. This is because all of those participants (n = 6) either quit or dropped out of the study,

Figure 16. Association Of Commitment With Consumption Success At Final Follow-Up

Figure 17. Association Of Commitment With Felt Success At Final Follow-Up

Figure 18. Association Of Commitment With Diet Maintenance Over Six Months

Transition Method

For people who chose to reduce their animal product consumption gradually (n = 143 in this analysis), we compared the strategy of reducing overall consumption gradually against reducing consumption of particular foods and against a combination of those or other strategies. We found no significant evidence that any one reduction method works better than another. However, we suggest choosing a clear method (e.g., by eliminating foods one at a time or increasing the number of veg*n meals per week) to make it easier to commit to a specific change and follow through with it.

Figure 19. Association Of Reduction Method With Consumption Success For Gradual Transitioners

Figure 20. Association Of Reduction Method With Felt Success For Gradual Transitioners

Figure 21. Association Of Reduction Method With Diet Maintenance For Gradual Transitioners

Previous Attempts To Go Veg*n 

Previous attempts to go veg*n were not a significant predictor of any measure of success. However, we feel it is worth noting that there was a slight trend for people with previous attempts to be more likely to abandon the diet within six months. It wasn’t large enough to be statistically significant and may have occurred by chance, but it is plausible that people who have tried and failed before would be more likely to try and fail again. This is enough of a possibility to warrant mentioning the slight difference.

Figure 22. Association Of Previous Attempts to Go Veg*n With Consumption Success At Final Follow-Up

Figure 23. Association Of Previous Attempts to Go Veg*n With Consumption Success At Final Follow-Up

Figure 24. Association Of Previous Attempts to Go Veg*n With Consumption Success At Final Follow-Up

Other Animal Product Goals, Previous Diet, & Previous Fad Diets

We also examined whether each of the following factors was related to any of the three measures of success and found no evidence (even weak evidence) that they were:

• Whether participants had a goal to avoid using other (non-food) animal products
• Whether participants were previously omnivores vs. already reducing animal product consumption in some way
• Participants’ previous amount of animal product consumption
• Whether participants had previously tried fad diets like Atkins, paleo, or similar

In other words, this study suggests that success on a veg*n diet does not depend on whether participants avoid other animal products, what their previous diet was, or whether they’ve previously tried fad diets.

Demographics

We also looked at demographic and household characteristics as predictors of our three measures of success. For the most part, these characteristics were not strong predictors of success. None of them were significantly and meaningfully related to felt success or likelihood of maintaining versus abandoning the diet.

Table 10 below shows consumption success broken down by the few demographics that meaningfully predicted success.

Methodological note: We regressed each success measure on each demographic predictor using full information maximum likelihood. Demographic groups with fewer than 20 cases were combined with other groups to increase the reliability of the results. For felt success and consumption success, we then examined the variance explained (R²) to determine whether a given demographic was a meaningful predictor or not, using 4% (R² = .04) as the minimum cutoff, perFerguson’s (2009) recommendations for the social sciences. For the measure of diet maintenance vs. abandonment, we looked for significant differences.

Table 10. Demographic and Household Predictors of Consumption Success

As you can see above, several patterns appeared for consumption success:

• Younger people tended to be closer to achieving their veg*n goals than older people. (Remember that fewer servings per month indicates more success.)
• People who were married tended to be further from their goals than single and other unmarried people.
• The most striking difference was observed for people with children, who tended to be significantly further from their goals than people with no children.

Characteristics that werenot clearly related to any of our success measures included:

• Vegan vs. vegetarian goal
• Gender: man vs. woman or another gender
• Race: white vs. BIPOC
• Geographical region: Northeast USA, Midwest USA, South USA, West USA, Canada
• Employment status: full-time vs. part-time vs. not employed
• Student status: student vs. not
• Animal guardianship status: has companion animal vs. not
• Education completed or currently pursuing: While education met the “meaningfulness” criterion, accounting for exactly 4% of the variance in consumption success, it did not show a clear pattern of significance or effects. We compared consumption by people with the most common education level—a bachelor’s degree—against each of the other groups, and none of the comparisons were significant. The results were as follows: bachelor’s degree (7.4 servings from goal, ± 8.5) vs. high school or less (2.0 servings, ± 2.1), some college/university but no degree (12.8 servings, ± 15.0), associate degree / college diploma (2.9 servings, ± 2.5), or graduate degree (8.6 servings, ± 5.6).

Conclusions

This study is one of the first attempts to track early diet change over a reasonably long period of time. This initial report provides a look at the people who commit to going veg*n and how they do so in the “real world,” outside the context of an advocacy campaign.

It is clear from the results that the ways people reduce their harm to animals are nuanced, imperfect, and varied. We have provided several examples of the unique journeys people take in a graphic on the Results tab. See “Sample Journeys.”

We suggest that advocates keep an open mind and help people find the approach that works best for them. For those who advocate veganism only, that does not mean you need to compromise your morals or change your end goal, but bear in mind the wide range of paths that can all lead to the same endpoint.

More New Vegans Than In The Past?

In this study, we found that 41% of people had a vegan dietary goal, while 59% had a vegetarian one. In other words, 1.4 times as many people intended to become vegetarian. While vegetarianism was also more common in our 2014  study, the difference observed then was much more striking. At the time, there were 3 times as many vegetarians as vegans, and 9 times more former vegetarians than vegans. This suggests that choosing veganism over vegetarianism appears to be more common than it used to be—a finding that makes sense in a world where the plant-based market share has increased substantially and large-scale challenges like Veganuary have gained worldwide attention.

At the same time, it is important to note that there is a major methodological difference between the current study and our 2014 study that could exaggerate the extent of the difference. While the 2014 study only required a single survey, participants in the current study knew that the study was intended to last up to six months. Our participants may therefore be more committed than average, and if new vegans tend to be more committed than new vegetarians, they may be overrepresented.

Vegan Or Plant-Based?

The focus of this study was on diet, but we also asked about other animal product usage. Close to half (47.5%) of the participants with a vegan dietary goal also had the goal of avoiding all non-food animal products. Interestingly, more than a third of people with a vegetarian goal (37.3%) also intended to avoid all non-food animal products.

Overall then, we found that about half of the participants we refer to as “vegan” in this report would be described as “plant-based eaters” by people who make this distinction on moral grounds. As noted previously in the report, we understand the reasoning but chose not to differentiate the two in this report as our goal is to support people who have begun to reduce harm to animals with their actions. This is further justified by the literature, which shows that inclusiveness and social support have positive benefits across many domains of behavior change and well-being, while othering and exclusion have negative implications (e.g.,DiMatteo, 2004;Korpershoek et al., 2019;Lemstra et al., 2016).

Gradual And Imperfect Change Is Normal

By the end of our six-month tracking period, 57% of participants had met or surpassed their dietary goal, but the average participant was still consuming 6.1 monthly servings of animal products more than they intended. (For vegans, that would be 6.1 servings total. For vegetarians, it would be 6.1 servings on top of however many servings of dairy and eggs they intended to eat).

Even more strikingly, just 28% of participants felt 100% successful by the sixth month, and the average level of felt success at the end was 88%. It was also interesting to observe that felt success and consumption success were only moderately correlated. That is, while being close to your goal consumption tended to go along with feeling more successful, they definitely didn’t go hand-in-hand. There were people who were doing very well on their consumption but felt less successful for various reasons, and people who had a ways to go but still felt that they were doing well, perhaps because a slower transition was part of their plan.

We also saw that most people do not transition to veganism or vegetarianism overnight. Most take at least a few days or weeks (38%), many weeks or a few months (34%), or even many months (7%). Just 21% went veg*n overnight. While more gradual transitioners tended to feel less successful and be further from their goal diet after six months, there was no significant difference in their likelihood of quitting versus continuing.

Family Responsibilities Increase Difficulty

We found that married people tended to be further from their veg*n goals after six months than unmarried people, even those with long-term partners. There was an even larger difference between people with and without children, as people with children tended to be substantially further from their goals. This suggests that competing demands make progress more difficult.

However, while these individuals were further from their goals, there was no significant evidence that they were more likely to abandon their veg*n attempt or that they felt less successful than others. It is possible that they may be more flexible in their interpretation of success, taking into account the difficulties posed by their family obligations. We suggest that advocates learn about (e.g.,Asher & Cherry, 2015;Greenebaum, 2018) and acknowledge these difficulties, emphasizing that a flexible approach is possible while maintaining commitment to the goal.

Additional research on the subject points to the variety of ways that family circumstances. Helping people find ways to overcome barriers is important, but it is also important just to understand and validate the difficulty they pose so that people feel supported in their journey. (This recommendation is drawn from research on how responsiveness supports goals; e.g.,Canevello & Crocker, 2010;Feeney, 2004).

Caveats & Limitations

As with all studies, this one has some important caveats and limitations to bear in mind.

Self-Selection Into The Study

First, although this is a general population sample, participants still had to opt in to participate, and a six-month study is a big commitment. We tried to reduce the impact of that commitment by telling them from the beginning that they could drop out at any time. However, simply agreeing to participate in a six-month study suggests a degree of commitment to one’s new veg*n diet that may not exist in many people who try it. It is for this reason that, on the Key Findings tab, we describe the sample as most representative of people who have already moved beyond a simple interest or desire to change into the stage where they are ready to actively work toward a veg*n goal.

It is because of this probable selection bias that we have not put a lot of emphasis on the proportion of participants who abandon their diets within the six-month study period. We provided an estimate of 43% in the ‘Diet Maintenance Versus Abandonment’ section of the Results, but our certainty about that estimate is quite low.

Participants’ Awareness Of Being Studied

One common difficulty with research is when just the act of participating can change participants’ behavior. In this case, it is possible that participating in a study about their new veg*n diets may have inspired participants to do better on those diets than they would have if they weren’t being observed.

If it occurs, this kind of effect would mean that some of the descriptive statistics like the percentage of people who continued their diet or felt motivated to continue may be more positive than they would be otherwise. However, the results that are more useful from an advocacy perspective are the analyses examining which characteristics and behaviors arepredictive of success—and it is unlikely that these would be very affected. For instance, by the end of the study we found that the average participant was still 6.1 servings of animal products from their goal. If they hadn’t participated in this research, that might hypothetically have been an average of 10 servings instead. So it affects the overall number, but is less likely to affect something like the fact that people with kids tend to be further from their goal diets than people without kids. Maybe instead of people with kids being 11 servings from their goal versus 3 servings for people without kids, it would be 20 versus 12, or 15 versus 7, but the difference between the two groups is less likely to be systematically affected.

Self-Reported Data

All data collected in this study is self-reported by participants and is subject to the same limitations and biases that always apply when people try to reflect on their own behavior. They may not have good insight or a good memory of what happened, and they may be motivated to present themselves in a positive light. That said, similar to the previous section, self-report biases, if they occur at all, are more likely to affect the absolute values like percentages than the relationships between characteristics and outcomes, because it would have to systematically affect participants with one set of characteristics more than another. This can absolutely occur, but self-report bias doesn’t undermine all survey research the way it is sometimes described as doing.

COVID-19

About two thirds (69%) of our sample participated in the study starting in September 2019 or later, meaning that their six-month participation time frame overlapped with the presence of COVID-19 in the U.S. and Canada, where states of emergency were generally declared at the state and provincial level in March 2020. This may have made dietary change more difficult in some ways or for some people (e.g., shortages of staples like tofu) and easier in other ways or for some people (e.g., by reducing social pressures and forcing people to cook for themselves more often).

While this is somewhat unfortunate for the study, we do not see it as having major implications for the interpretation of the results. It may have increased barriers for some people and decreased them for others, but most participants (89%) were recruited into the study prior to March 2020. This suggests that our sample is not made up of people who were influenced to go veg*n by the pandemic conditions. While individual barriers may be different than they would have been without the global health crisis, it is the strategies that are used to overcome them that are of greatest interest to us (which will follow in a subsequent report).

Power Level

We preregistered an intention to recruit 400 participants in order to have at least 200 who completed the entire study by making it to the end or completing the quit survey. Due to recruitment difficulties (described in more detail on the Supplementary Materials tab), we were unable to attain that goal. However, our participant retention was fairly good so our power level did not suffer too much.

With 222 participants in this sample, we have 157 useable responses for regression analyses (138 who finished and 19 who quit). This number increases to 188 including those who dropped out but whose data can be handled using statistical techniques (full information maximum likelihood modelling or imputation) but for the purpose of power analysis, we do not include those individuals. With 157 participants, we have 90% power to detect an effect size off²= .07 at an alpha level of .05 orf²= .12 at .003. (An alpha range of .05 to .003 is what occurs when using a post hoc FDR correction with 15p-values, reflecting the 15 barriers/supports used in our largest analysis sets.)

Detailed Procedure

Participant Incentives

Participants were offered an Amazon gift card worth $5 (CAD or USD, depending on their country of residence) per survey they completed, for up to $35 value total. If they completed at least five of the six follow-up surveys, they were also entered into a drawing for one of three $300 gift cards. We used these incentives to offset the demands of the survey and keep attrition to a reasonable level.

Recruitment 

A sample of 222 participants (after data cleaning) were recruited into this study between July 2019 and March 2020. With a six-month study period per participant, end dates fell between January and September 2020. The sample was smaller than the 400 participants we had intended to recruit because it proved to be a very difficult group to recruit.

The inclusion criteria were very strict, which is the primary reason for the difficulty. Participants had to be adults in the U.S. or Canada who had started a new vegan or vegetarian diet within the past 2 months. In addition, we wanted to recruit a general population sample, not made up of people who had signed up for campaigns or been recruited by groups with a particular agenda. For the latter reason, we minimized recruiting through Faunalytics’ advocacy networks and effective altruism (EA) groups.

Instead, we used a wide range of recruitment sources and emphasized those that are neutral regarding motivation for the diet change (i.e., we focused on sources that describe themselves in terms of the diet itself rather than animal welfare, environmental protection, or health). The full list of sources is provided in the table below, along with the number of participants recruited into the final sample from each.

Table 11. Recruitment Sources

We are very grateful to the groups and individual volunteers who helped us disseminate the recruitment post and posters. The full effort included finding and contacting these and many other potential distributors.

Data Linking

We used unique, participant-generated codes to anonymously link the data from each time point. The instructions read “Please enter your first & last initials, your 2-digit birth month and the last two digits of your birth year (e.g. John Doe, born February 1987 = JD0287). The purpose of this code is to allow us to link your responses from all surveys. Please check it carefully.”

This method works reasonably well for data linking, as it means one doesn’t need to track overtly identifying information or ask participants to remember a random code that they could lose. However, some participants will inevitably enter their code wrong at one or more time points by misreading the instructions or making a typo. Where this occurred in this study, we did our best to track down the problems and link the data manually, but some missing data almost certainly occurred.

Exclusion Criteria

Our participant exclusion criteria were defined in our pre-registration on the Open Science Framework and originally included just three criteria (see original here). However, as noted there, we revised these to stricter criteria after experiencing an influx of spam. The revised criteria are availablehere and are also indicated below. The source from which the spam entries originated was removed from the dataset entirely and therefore is not reflected in the exclusion numbers listed below. For more details of that experience, please refer to this blog post about survey fraud (Anderson, 2019, Faunalytics).

The exclusion criteria we used to evaluate these checks were nota priori, as they were developed based on observation of the distribution of data for all flags listed below, to balance quality and data loss. However, we pre-registered them anyway because we made these decisions before conducting any of the analyses to test our hypotheses and research questions.

We included a survey-level flag for each of the following and adopted a three-strikes rule to exclude a survey:

1. Duplicate IP addresses;
2. “Suspect” IP addresses (e.g., known survey farms);
3. IP addresses from outside survey target area (U.S./Canada);
4. Completion times that were less than one third of the median for that survey;
5. Contradictory responses on a consistency check framed as enjoyment of the survey (e.g., choosingvery enjoyable andvery boring);
6. First attention check: Checking a box that said they had run a mile in less than 3 minutes (the world record is 3:43);
7. Second attention check: Failing to check a box that said they had used the internet, a phone, or a computer in the past month (the survey was administered online);
8. Failing a basic reading comprehension test (e.g., not correctly indicating that “Cats do not like to swim” is equivalent to “Cats dislike swimming”).

Using this rule, 2 surveys were excluded out of 1,100 total. (For comparison, a two-strikes rule would have excluded 67 surveys.)

We also added a person-level flag for each of the following and adopted a three-strikes rule to see whether we should exclude particular participants. This entailed looking at the three person-level flags below, as well as a fourth flag for anyone with an average of 1 or more strikes on each survey they completed:

1. Highly inconsistent unique identification code entered on baseline survey and first follow-up (not including those that could be attributed to typos or misreading the instructions);
2. Inconsistent age entered on baseline survey and first follow-up (not including ages that increased by 1);
3. Suspicious written response to an open-ended question on the baseline survey. Specifically, these were flagged if they duplicated another response, had poor grammar, were hard to understand (i.e., not perfectly intelligible), “clunky”, or did not answer the question asked (per ourdata quality assurance plan).

No one was excluded using this rule.

Attrition And Missing Data

As defined in ourpre-registered plan, we systematically considered whether or not missing data in later surveys could be attributed to differential attrition. The pre-registration strictly defines the following groups of participants: Completers (n = 138), Quitters (n = 19), Attriters (n = 31), and Nonstarters (n = 34). In simple terms, they are (respectively) participants who completed the study while continuing their veg*n diet, participants who told us they quit their diet, participants who dropped out of the study without explicitly quitting their diet, and participants who completed the baseline survey but nothing else.

The analysis looking for signs of differential attrition is described in detail at the link above, but briefly, we examined the associations between three participant statuses (completer, quitter, or attriter) and the 15 key barriers and supports identified as probable predictors of success. We conducted two t-tests per barrier/support: one comparing attriters versus completers and one comparing attriters versus quitters. The goal was to see whether attriters resembled completers or quitters more closely or whether the result was mixed/ambiguous.

The result was ambiguous. Specifically:

• In 10/15 cases, attriters did not significantly differ from either completers or quitters: autonomy support, societal perceptions of diet, shame/pride in diet, size of veg*n network, cultural influence on ease of following diet, personal control over food, perceived progress on health goals, feelings of motivation, frequency of cravings, and extent of habit formation.
• In 3/15 cases, attriters differed from completers but not quitters, suggesting they resembled quitters more: ability to find or prepare food, perceived healthiness on diet (marginally significant), and strength of identification with their goal diet.
• In 1/15 cases, attriters differed from quitters but not completers, suggesting they resembled completers more: concerns about dietary purity.
• In 1/15 cases, attriters differed from both quitters and completers, resembling neither: cost of diet.

We interpreted these results as indicating that people who dropped out of the study were not systematically different overall from people who did not drop out. As such, we proceeded with our pre-registered plan to treat data from attriters as Missing At Random (MAR) and used full information maximum likelihood (FIML) methods to handle the missing data for models with continuous outcome measures. The only deviation from our pre-registered plan was to use multiple imputation (via the mice package) for models with a binary outcome measure (diet maintenance vs. abandonment) because we learned after pre-registration that FIML assumes multivariate normality. The regression models used are described in more detail below.

Details Of Success Measures

Diet Maintenance Versus Abandonment

As noted in the Results section, just 19 of 222 participants explicitly abandoned their diets within the study timeframe, which provides an estimated abandonment rate of 9%. However, we know that 179 of 222 participants had been following their diet for one to eight weeks prior to the start of the study.

This is important becauseFaunalytics (2014) found that about one-third (34%) of people who tried a veg*n diet abandoned it within the first three months. For this analysis, we choose to assume that the 179 participants who had already been following their diet for at least a week when the study started represent two-thirds of the people whowould have been part of the study if we could have tracked them all from the day they started their diets.

In other words, those 179 people are the ones who maintained their new diets long enough to be included in the study—those who lapsed sooner are not counted. To add those early abandoners to the 179 people in the study, we calculated a new total that accounts for the 179 people being two-thirds of them:

179 /x= 23

x = 269

Adding in the 43 people who started their diets at the beginning of the study (222 – 179 = 43), this gives us a hypothetical starting sample of 312:

269 + 43 = 312

To get the estimated number of people who abandon their diet within the six months of the study, we calculate the number of hypothetical abandoners that come from the previous calculation (first line below) and add it to the number who explicitly abandoned their diets during the study (second line below):

269 – 179 = 90

90 + 19 = 109

This indicates that if we had started with the hypothetical sample of 312 per the calculations above, 109 would have abandoned their diets. This provides a proportion estimate of 43%, as shown below:

109 / 312 = 0.43

Consumption Success / Distance From Goal

As described in the Results section, consumption success was measured in terms of animal product servings. Specifically, the difference between participants’goal consumption of animal products and theircurrent level of consumption in servings at each time point.

Participants reported their current and goal consumption using a Food Frequency Questionnaire (FFQ) with five food groups, each rated on a five-point scale, as shown below.

Figure 25: Food Frequency Questionnaire

We converted each scale point into an average number of servings per month according topre-registered specifications, calculating the total number of animal product servings consumed. We then calculated consumption success at each follow-up time point by subtracting goal consumption from actual monthly consumption to determine the number of servings they have left to cut. A smaller number indicates more progress, with 0 indicating that they have reached or surpassed their goal.

Felt Success

Felt success, or self-perceived success, is included here for completeness, but its calculation is straightforward. Participants were asked “When you think about your dietary goals, how successful do you feel?” and chose a percentage to respond. They were also able to write an explanation of the percentage they chose if they felt it needed more context.

Detailed Analysis Method

Representativeness & Weighting

As indicated in our pre-registration, we examined the representativeness of the sample relative to the population breakdown established inFaunalytics (2014). Although the 2014 sample is getting older, it provides the strongest demographic breakdown of veg*ns that we know of because it was extracted from a sample of over 10,000 members of the general U.S. population.

As shown in the table below, the current study sample was remarkably similar in its breakdown. The only substantial difference is that our sample skews younger than the 2014 population, possibly because so much of our recruitment occurred via social media.

Table 12. Comparing Sample Demographics Against Faunalytics (2014)

The similarity of this sample to our comparison sample suggests that our wide-ranging recruitment method was successful in achieving a relatively representative sample of the general population. In this and the subsequent reports, we are weighting the descriptive statistics to improve the age-related representativeness but we have not and will not use weighting for any of the inferential statistics.

Regression Analyses

The regression analyses presented in this report were used to construct models predicting each of our three measures of success: consumption success (operationalized as distance from goal diet, in monthly servings), felt success (operationalized as a self-reported percentage), and diet maintenance vs. abandonment (operationalized as whether participants explicitly abandoned their diets or not).

These analyses were not pre-registered, as they are not part of our set of primary research questions and hypotheses, but we followed the same procedure laid out in the analysis plan for those.

Exclusions

We excluded 34 “non-starters” (participants who completed the baseline survey but none of the follow-ups) from all regression analyses, per our pre-registered plan. We had decided on this course of action because they had provided so little data on which to build a longitudinal model and because it is impossible to tell whether they quit the study, had never intended to complete more than one survey, or did not receive the follow-up invitations due to an error with their email address or linking code. We felt,a priori, that there was a substantial difference between this type of missing data and data that was missing after having completed at least one follow-up survey.

Bivariate Models & Post Hoc Corrections

To aid with interpretation, almost all analyses presented in this and Faunalytics’ other reports use a simple bivariate model with a single predictor measured at baseline and a single outcome measured at Follow-Up 6. That is, intermediate time points are ignored in these analyses.

However, it is important to note that we have also conducted a parallel set of multilevel models that predict scores at each time point from previous time points. The pattern of results is similar to what we report with these simpler analyses and will be used for academic publications.

Because using bivariate models in a study like this necessitates a large number of analyses, we applied a post hoc correction to the p-values to correct for False Discovery Rate (FDR). For any set of analyses in which a set of predictors were regressed individually on a single outcome variable (such as in the case of the 8 predictors used in this report), we applied the correction for that set of analyses.

Treatment Of Missing Data

When modelling consumption success and felt success, we used full information maximum likelihood (FIML) methods (via thelavaan package in R) to handle the missing data that arose through attrition or through missed surveys by study completers.

When modelling our binary success measure (diet maintenance vs. abandonment), we used multiple imputation (via themice package in R) to handle the missing data because FIML assumes multivariate normality.