Summary of summary: This is a summary and interpretation of "Impacts of Animal Well‐Being and Welfare Media on Meat Demand" by Glynn T. Tonsor and Nicole J. Olynk. The authors estimated, in the US, based on the period from 1982 to 2008 and animal welfare articles in top US publications:
- For each 1% increase in the number of pig welfare media articles in a quarter (3-month period), there was a 0.0066% decrease in pork demand in the same quarter and a 0.0146% decrease in pork demand over the same quarter and the next quarter.
- For each 1% increase in the number of chicken/turkey welfare media articles in a quarter, there was a 0.0198% decrease in chicken/turkey meat demand over the same quarter and the next.
- In 2008 numbers, and assuming the article counts used in the study are exhaustive of the effects (unlikely), this would have been around 300,000 chickens spared per additional article. If we can interpret this as the causal effect, this seems like it could be competitive with corporate campaigns, with less risk of backfire (mistaking an aggregate welfare worsening for an improvement). I expect that the number of articles has increased dramatically since 2008, though, so the returns may be much worse now in the US, but may be attractive elsewhere or for more neglected animals, e.g. fishes, crustaceans and insects.
- For each 1% increase in the number of pig or chicken/turkey welfare articles, there was a 0.0001% increase in non-meat food demand, and holding prices constant, no cross-species effects from animal welfare articles were detected. So consumers were replacing some meat with non-meat foods.
- The roughly 200% increase or tripling in the number of pig and chicken/turkey welfare articles per quarter from 1999 to 2008 has counterfactually reduced 2008 demand for pig meat by 2.58% and chicken and turkey meat by 4.76% from what it otherwise would have been if the number of articles had been the same as in 1999. This is around 400 million chickens spared in 2008.
This was not an experimental study, so the usual risks of bias and confounding apply. All of the above statements come with caveats and qualifications, or ignore potentially important details.
I think the study design was useful, and we should consider replicating this study with more recent data, or conduct studies using similar designs to estimate the effects of other kinds of interventions, e.g. protests and undercover investigations (possibly through their media coverage).
EDIT: A major problem with interpretation may be that highly correlated coverage in other forms of media like TV is what cause most of the change in demand, and we're incorrectly attributing the effect to media articles. If coverage on TV and in articles are roughly proportional, then we could still have a similar percentage effect for a similar percentage change in coverage, but different absolute changes in coverage via the different media.
This article provides the first known examination of how animal welfare information provided by media sources impacts beef, pork and poultry demand. Results suggest that media attention to animal welfare has a small, but statistically significant impact on meat demand. Long‐run pork and poultry demand are hampered by increasing media attention whereas beef demand is not directly impacted. Loss in consumer demand is found to come from exiting the meat complex rather than spilling over and enhancing demand of competing meats. An outline of economic implications is provided for the broader discussion of animal welfare.
The authors use a Rotterdam model to estimate the effects of the number of animal welfare-related media articles in a quarter (3-month period), specifically about cattle, pigs or chickens and turkeys, on meat consumption, adjusting for prices, a season parameter and Divisia volume. They did not try to distinguish between articles with positive vs negative coverage.
They estimate both short-run effects, i.e. effects within the same quarter, and long-run effects, and they find that long-run effects only last an extra quarter (longer effects were not found to be statistically significant). So, short-run effects cover a 3-month period and long-run effects cover an 6-month period.
They describe the elasticities of meat demand from species X with respect to the number of animal welfare articles about X, where X=cattle, pigs and chickens/turkeys:
Table 3 suggests that media attention to animal welfare issues have not impacted beef demand, have reduced pork demand in the short-run (-0.0066) and have reduced both pork (-0.0146) and poultry (-0.0198) demand in the long-run.
In other words, for each 1% increase in the number of pig welfare articles in a quarter, there was a 0.0066% decrease in pork demand in the same quarter and a 0.0146% decrease in pork demand over the same quarter and the next quarter. For each 1% increase in the number of chicken/turkey welfare articles, there was a 0.0198% decrease in chicken/turkey meat demand over the same quarter and the next quarter.
Furthermore, the authors estimated the elasticities of non-meat food demand with respect to the number of animal welfare articles to be -0.0001 for pig animal welfare articles in the short-run and long-run, and for chicken/turkey animal welfare articles in the long-run. In other words, for each 1% increase in the number of pig or chicken/turkey welfare articles, there was a 0.0001% increase in non-meat food demand. So consumers were replacing some meat with non-meat foods, although at a low rate.
These seem like small effects, and much smaller than the price elasticities (-0.888 to -0.114). The number of articles they're looking at are on the order of 22 and 72 per quarter on average for pigs and chickens/turkeys, with maximums of 134 and 331, respectively, and around 200-300 articles about chicken/turkey welfare each quarter towards 2008, the end of the data collection period. Each article about chicken/turkey welfare would have been around a 0.4% increase in the number for its quarter, corresponding to (and resulting in, if we can interpret this as the causal effect) a 0.008% decrease in chicken and turkey meat consumption across two quarters, or half of the year. Around 8 billion chickens were slaughtered in 2008 in the US (I'm not sure if this includes egg-laying hens), so this would be around 300,000 chickens spared per additional article:
However, to interpret this causally, we also need to assume that the counts of articles used are exhaustive of the causal effect, which they are likely not. It's possible the search was too narrow. This also ignores imports and exports, and equilibrium effects with supply; because supply is not perfectly elastic (in both the short run and long run), prices should decrease and fewer chickens should be spared than otherwise. In fact, supply tends to be very inelastic in the short run, so a single additional article about chicken welfare may have little expected impact on the number of animals raised for food on its own, whereas a continued pattern of additional articles results in a long run trend in demand, to which the industry will be more responsive.
The articles were found using database searches of top US newspapers and magazines, so this might cover a substantial fraction of what people read, but these are likely correlated with the numbers in less popular outlets without necessarily causing them, indicating confounding, so if we were to target only the top ones, these elasticities may overestimate our impact. More importantly, this completely misses TV coverage, which we would also expect to be highly correlated with coverage in articles.
There may be other sources of confounding, too, since the study is non-experimental. The number of articles, rather than being primarily a driver of changes in demand, could have been the consequence of a common cause (or multiple), e.g. changing public sentiment surrounding farmed animals, political shifts. However,
- EDIT: I think it's unlikely decreases in poultry demand caused increased media coverage of chicken and turkey welfare to a significant enough degree that explains the long-run elasticity, because the same-quarter/short-run elasticity was not statistically significant (and it was positive, the wrong sign), so the estimated decrease in demand would have come after the increase in media coverage.
- They failed to reject the hypothesis that the cross-species effects from articles were 0, so it seems like changes in demand are specific to the species in the animal welfare articles. In other words, articles about chicken and turkey welfare didn't affect pig meat and cattle meat consumption; those about pigs didn't affect chicken/turkey and cattle meat consumption; and those about cattle didn't affect chicken/turkey and cattle meat consumption, at least to a statistically significant degree (under a joint test, and holding prices constant, which may not be realistic in a market). This basically rules out alternative explanations for the observed elasticities that aren't species-specific, strengthening the case for causality. This also seems to rule out significant cross-species meat substitution effects, e.g. replacing pig meat with chicken or turkey meat, other than through price changes.
They also did not consider the impacts of an extra animal welfare article on the actual number that will be published; this may increase interest further and lead to more articles, or we may hit saturation and actually get fewer than 1 extra article on the topic on average for each one we were to publish.
With animal welfare index = the number of articles they found in their search,
For instance, between 1999(1) and 2008(4) the pork and poultry animal welfare indices increased by 181.3% and 253.2%, respectively. Utilising an approach similar to Schmit and Kaiser (2004) and coupling this with the estimated long-run elasticities of our model suggests that pork and poultry demand increases over the last decade would have been (ceteris paribus) 2.65% and 5.01% higher, respectively, if media attention in 2008(4) was at equivalent levels as 1999(1). Hence, although the point estimates of these elasticities are small, over a longer period of increasing media attention to animal welfare issues, they imply notable economic impacts to the US livestock industry.
In other words, the roughly 200% increase or tripling in the number of pig and chicken/turkey welfare articles per quarter from 1999 to 2008 has counterfactually reduced 2008 demand for pig meat by 2.58% and chicken and turkey meat by 4.76% from what it otherwise would have been if the number of articles had been the same as in 1999. This is around 400 million chickens spared in 2008 (with the same caveats about broiler chickens vs egg-laying hens, imports and exports).
I didn't realize ACE had already covered this study a while back. See here.
Another major problem with interpretation may be that highly correlated coverage in other forms of media like TV is what causes the change in demand or most of it, and then we're incorrectly attributing the effect to media articles.
Since they're looking at articles and demand in the same quarter, the increase in number of articles might follow the drop in demand, too.
For what EA is already doing,
But it's certainly not just us writing for or funding writing in popular media outlets on animal welfare.
This is a very valuable comment, thanks for adding the context!
Also, Ezra Klein is leaving Vox for the New York Times. I'm not sure if he'll write about animal welfare there, but if he does, this would be huge. There's also a risk that Vox won't be as good without him.
There are also other things the animal protection movement does that often gets attention in major outlets, like
There have also been similar studies on the effects of information, media and advertisements about health and food safety on meat demand:
Thanks Michael! This is really interesting. Decreasing demand by a few percent is a pretty big deal.
My intuition is that the number of articles published isn't exact the right thing to regress on, probably instead you want something like "article views". Did the authors discuss this? I guess if all the articles are published in equally-viewed sources, looking at just the raw article count would be fine.
Ya, that makes sense. They didn't discuss this, specifically. It's worth keeping in mind that their analysis went back to 1982, so I doubt they'd have good data on number of views/reads going that far back.
They did write:
They also wrote that trying to distinguish articles by direction (positive or negative coverage), severity, and/or source would have introduced subjective elements to the analysis and wouldn't have been feasible: there were 870 articles about cattle welfare in one quarter alone, at the maximum (although this was an outlier). There were also 606 sources in the database, but I don't know how many of these ever covered animal welfare.
I wonder what events these articles were written following, since it might not be the animal welfare content that matters. When something major happens in the industry, animal welfare coverage might follow regardless just because of greater attention. If there's a disease outbreak, we might see both health/safety content and animal welfare articles, but the health/safety content might be what drives the reduction in demand. My impression is that the effects of health/safety information are stronger than animal welfare information.
Maybe a replication should include separate counts for articles about animal welfare, the environment and health/safety in the models (with articles allowed to count towards multiple), and possibly just an overall count for coverage of animal agriculture.
I made a correction: I misinterpreted their tests of cross-species spillover effects, i.e. articles about one species affecting the demand for meat from another, controlling for price (and the few other parameters). They wrote:
In other words, they couldn't tell these effects apart from 0, so the effects were negligible, at most. So the effects of the articles are pretty species-specific, which makes a stronger case for causality.