You are on page 1of 9

Joseph Millum, Ph.D., M.Sc.

joemillum@gmail.com

NHANES: Non-vegetarian vegetarians and other interesting diets

The National Health and Nutrition Examination Survey (NHANES)


The National Health and Nutrition Examination Survey (NHANES) is a series of
surveys sponsored by the Centers for Disease Control and Prevention (CDC).
They collect data on a nationally representative sample of the civilian, non-
incarcerated US population. This data is gathered through extensive interviews
and medical examinations. It covers demographics, diet, and many measures
relating to health and health behaviors. Since 1999, NHANES has surveyed
approximately 5000 people per year and releases data in two-yearly cycles.

The 2007-8 and 2009-10 cycles included the question: “Do you consider yourself
to be a vegetarian?” All cycles include various questions about dietary habits
(these vary between cycles) and two 24-hour dietary recall interviews. The
interviews compile everything that the respondent recalls eating over the past 24
hours.

Previous analyses using NHANES


1. My previous blog post reported on the demographics of self-reported
vegetarians using the NHANES dataset. Approximately 2.2% of the non-
incarcerated, civilian U.S. population self-identify as vegetarian. Self-identified
vegetarians are more likely to be female, college-educated, born in a non-Spanish
speaking country outside the U.S., and identify as “other” race/ethnicity—
meaning not white, Black, nor Hispanic.

2. Juan et al. analyzed the day 1 dietary interview data and estimated that 52% of
self-identified vegetarians reported eating meat or fish in the first dietary recall
data.1

3. Saulius Šimčikas and Harry Burdon analyzed the NHANES dietary recall data
from 2003 – 2014 to estimate the proportion of adults and children over age 8

1 Juan, WenYen, Sedigheh Yamini, and Patricia Britten. "Food intake patterns of self-identified
vegetarians among the US population, 2007-2010." Procedia Food Science 4 (2015): 86-93.
2 https://animalcharityevaluators.org/blog/is-the-percentage-of-vegetarians-and-vegans-in-the-u-
2

who ate vegetarian and vegan diets during the two 24 hour periods.2 They also
note: “Out of the people we classified as vegetarians based on dietary recalls,
only 41% self-identified as vegetarians in 2007–2010.”

Motivations for this analysis


The dietary interview data comprises probably the most reliable record we can
get of what people actually eat, and it shows that many self-described
vegetarians actually report eating meat or fish. This prompts the question of
what, if anything, is different about true vegetarians versus others who answer
the vegetarianism question positively. Is it a misunderstanding or a language
issue? Are there different conceptions of what it means to be vegetarian? Are lots
of the meat-eating vegetarians actually pescetarians? Should we focus less on the
vegetarian label and more on what makes people eat little or no meat?

I attempted to make some headway on answering these questions by breaking


out the NHANES respondents into some different groups.

Data sets
1. Data on 20,686 individuals from the Demographic Variables & Sample
Weights questionnaires and the Diet Behavior & Nutrition questionnaires were
combined for the 2007-8 and 2009-10 NHANES datasets. Individuals who
refused (n=2) or responded don’t know (n=7) for the question about self-
identified vegetarian status were recoded as missing (for a total of 920 missing).
Dietary data from the day 1 and day 2 dietary interviews were combined for the
2007-8 and 2009-10 NHANES datasets. These were then merged with the
demographic dataset. Individuals for whom the food data was incomplete (i.e.
did not complete a dietary interview on both days) were dropped from the
sample. This left demographic and dietary data on 16,241 individuals.

2. USDA foodcodes were obtained from the USDA’s Food Surveys Research
Group.3 I classified them according to whether they contained meat (includes
mammals and birds) or fish (includes fish, amphibians, crustaceans, etc.). This

2 https://animalcharityevaluators.org/blog/is-the-percentage-of-vegetarians-and-vegans-in-the-u-
s-increasing/
3 Available at: https://www.ars.usda.gov/northeast-area/beltsville-md-bhnrc/beltsville-human-

nutrition-research-center/food-surveys-research-group/docs/dmr-food-categories/. I color-coded
the 2013-14 codes because that’s what I tracked down first… this contains more food codes than
earlier versions, but food codes that already existed in the earlier versions should still be the same
(for very limited exceptions, see: https://www.ars.usda.gov/ARSUserFiles/80400530/pdf/1516/15-
16_Changes%20in%20WWEIA%20Food%20Categories%20between%20survey%20cyles.pdf)
3

involved a number of judgment calls, since some of the foods are not described
in ways that specify whether they contain meat products.

Analysis
1. First, I attempted to replicate the analyses conducted by Juan et al. and
Šimčikas and Burdon. I used the dataset I’d constructed but just using Day 1 data,
as Juan et al. did. 43% (141/327) of self-identified vegetarians in my sample did
not report eating meat or fish on Day 1. Applying sample weights gives an
estimate of 51% of the population as self-identified vegetarians who did not eat
meat or fish on Day 1, which is very close to Juan et al.’s 52%.

Using the combined Day 1 and Day 2 data I estimated the proportion of adults
who did not consume meat on either day as 2.0% of the population and the
proportion of children aged 8-18 who did not consume meat on either day as
2.37% of the population. Šimčikas and Burdon’s population estimates were given
as ranges depending on how strictly foods were classified as vegetarian or not.
For adults they estimate 1.99% - 2.12% in 2007-8 and 2.07%-2.46% for 2009-10. For
children aged 8-18 they estimate 1.92-2.60% in 2007-8 and 3.08-3.81% in 2009-10.
These estimates seemed close enough to mine that the differences were likely
caused by minor differences in how we classified foods. Of 191 adults who I
classified as not eating meat on either day, 77 self-identified as vegetarian (40%).
This is very close to Šimčikas and Burdon’s 41%. Šimčikas and Burdon also
analyzed a question asking respondents if they ate finfish or shellfish in the last
30 days. Including these responses cut the estimates for people eating vegetarian
diets by half again.

2. I applied sample weights according to NHANES guidelines to allow estimates


that reflect the non-incarcerated, civilian U.S. population aged one year and
above. (Lots of babies (479) are reported as not eating meat, but that’s not a
helpful indicator for dietary trends… So, I excluded infants.) These were the
results:
4

Diet Description Number Population 95%


(N=15,495) estimate confidence
interval
Sayveg Self-described 3274 2.1% 1.7 – 2.5%
vegetarian
Trueveg Self-described 108 0.9% 0.6 – 1.2%
vegetarian and report
no meat or fish
consumption either
day
Nonvegveg Self-described 219 1.2% 1.0 – 1.5%
vegetarian and report
some meat or fish
consumption
Nomeat Report no meat or fish 309 2.2% 1.7 – 2.7%
consumption either
day
Meatveg Self-described 196 1.0% 0.8 – 1.3%
vegetarian and report
some meat
Fishveg Self-described 23 0.2% 0.1 - 0.4%
vegetarian and report
some fish but no meat
Fishnomeat Report some fish but 155 1.3% 0.9 - 1.7%
no meat
Vegdays = Report at least one 1962 12.2% 11.8 - 13.3%
1 or 2 day without meat or
fish

Conclusions. 2.1. Although 2.1% of the population claim to be vegetarian, only


0.9% of the population bear that out when asked about their food consumption.
This result is consistent with the findings of Juan et al. and other studies that
combine questions about self-identified vegetarian status and actual food
consumption. For reasons given earlier, I think this is a particularly well run
survey and so I have quite a lot of trust in its results. However, we are dealing
with very small numbers and so we shouldn’t put too much weight on exactly
what the tiny percentage in the overall population is. We should also expect that
the number of “true vegetarians” is lower than NHANES allows us to estimate,
since respondents who reported no meat or fish consumption on either of these

4 N=15,489 because this excludes those for whom there was no answer to the self-described
vegetarian question.
5

two days might still have eaten meat or fish on some other day (as Šimčikas and
Burdon note).
2.2. One might wonder if the reason for lots of self-identified vegetarians
reporting eating meat or fish is that there are lots of pescetarians who call
themselves vegetarians. The data suggests this is not the case. A tiny fraction of
self-described vegetarians ate fish but not meat (23/327). Moreover, among the
whole sample—vegetarian and not—only 1.2% reported eating fish but not meat
on the dietary interviews for the two days. In sum, in the US, the fish-only crowd
is very small.
2.3. By contrast lots of respondents reported no meat or fish consumption.
An estimated 12.2% of the population (1962/15,495 of the sample) ate vegetarian
for at least one of the two days that they were asked about. And 2.2% of the
population (n=309) ate vegetarian both days.

3. Another possibility is that self-identified vegetarians who eat meat,


nevertheless eat less meat than those who do not identify as vegetarian. I tested
this hypothesis by comparing the number of meat or fish items reported by
respondents on the dietary interviews over the two days combined.5

Diet Average animal items


nonvegveg 2.95
population 3.84

A t-test indicated that the mean of the nonvegveg group’s animal items
consumption is very unlikely to be equal to the population mean.

Conclusions: Meat-eating vegetarians do eat less meat than the overall


population, by about a quarter.

5Note that this is the number of items reported that contain meat or fish, not a direct measure of
the quantity of meat or fish consumed. Again, this is excluding infants.
6

4. Another possibility is that the meat-eating vegetarians are misreporting


because of a translation issue . The NHANES data gives the language in which
the interviews were conducted—English or Spanish—and whether an interpreter
was used.

Diet % Spanish interview (n) % used interpreter (n)


nonvegveg 24.2% (53/219) 1.8% (4/219)
sayveg 17.1% (56/327) 1.53% (5/327)
Population 15.7% (2,552/16,240) 0.71% (116/16,240)

Conclusions: There is a higher proportion of non-English speakers in the


nonvegveg group than in the sayveg or overall population. All but three of the 56
respondents who said they were vegetarian and completed the interview in
Spanish reported eating some meat or fish. But, this is only a quarter of the self-
identified vegetarians who also report eating meat or fish so the meat-eating
vegetarians is not just a translation issue.
(I also looked at how many meat or fish products were reported by the
nonvegveg Spanish speakers—pretty similar to the other nonvegveg
respondents: average of 2.68 items.)

5. I then conducted a regression to see whether non-English language interviews


were predictive of being a meat-eating self-identified vegetarian (nonvegveg). I
included the variables that were significant last time, plus age (i.e. gender,
education, age, income). I didn’t include “other race” or “born outside USA in
non-Spanish speaking country”, since the effect of those variables should be
independent of the effect of speaking Spanish in one’s interviews.

Regressor Coefficient (s.e.)


Female 0.0054 (0.0019) **
College grad (HH reference person) 0.0037 (0.0028)
Age (years) 0.000048 (0.000042)
Household income (thousands) -0.000071 (0.000035)*
Spanish language interview 0.0092 (0.0034) **

* significant at 5% level
**significant at 1% level
7

Conclusions: Take these with a grain of salt, given the size of the meat-eating
self-identified vegetarian group in the sample (219). Most interesting, given their
significance and effect size are Spanish language and female.

6. So, I broke down the nonvegveg and trueveg groups by gender6:

Diet Male Female


nonvegveg 31% (86/219) 69% (133/219)
Trueveg 34% (39/108) 66% (69/108)
sayveg 32% (125/327) 68% (202/327)
Population 47.6% (7,611/15,495) 52.4% (7,884/15,495)

Conclusions: More females than males are self-identified vegetarians and true
vegetarians. The discrepancy between saying that one is vegetarian and
reporting no meat consumption is not more common in one gender than another.

7. Having described the demographics of self-identified vegetarians previously,


it seemed interesting to see whether other dietary groups were similar or
different. So, I calculated population estimates for the demographics of some of
the groups described in (2) above using the survey weights from NHANES.
Results follow in tables. I’ve made graphics to illustrate, too (see attached jpegs).

Cautionary note: We are dealing with quite small sample sizes for some of these
(e.g. true vegetarians are only 108 out of 16,241 respondents). So, take these
population estimates as very rough. I’ve put in the 95% confidence intervals for
gender to give an idea of the range of possible variation.

Gender
Proportion of… Male 95% C.I. Female 95% C.I.
Sayveg 32% 24 – 42% 68% 58 – 76%
Trueveg 34% 24 – 47% 66% 53 – 76%
Nonvegveg 31% 21 – 42% 69% 58 – 79%
Nomeat 31% 25 – 39% 69% 61 – 75%
Population7 48% 47 – 49% 52% 51 – 53%

6 All these estimates are for the non-infant population (i.e. age >=1).
8

Conclusions: The gender effect is strong and consistent across vegetarians and
non-meat eaters.

Race/ethnicity
Proportion Mexican- Other NH-White NH-Black Other
of… American Hispanic
Sayveg 7.8% 5.7% 59.7% 7.5% 19.2%
Trueveg 4.3% 2.2% 70.9% 2.7% 19.9%
Nonvegveg 10.4% 8.1% 51.8% 11.0% 18.8%
Nomeat 6.8% 4.1% 71.3% 5.0% 12.8%
Population 9.9% 5.5% 66.7% 12.1% 5.8%

Conclusions: (1) True vegetarians are almost entirely non-Hispanic white or


“other” race/ethnicity (91% of true vegetarians versus 72% of the population). (2)
The proportions who do not report eating meat more closely match the
population as a whole but are still very skewed non-Hispanic white and “other”
(84% versus 72%).

Country of birth
Proportion USA Mexico Other Other non-
of… Spanish Spanish
Sayveg 75.0% 2.8% 3.1% 19.0%
Trueveg 79.9% 0.2% 0.4% 19.5%
Nonvegveg 71.6% 4.7% 5.1% 18.6%
Nomeat 82.4% 2.9% 0.8% 13.9%
Population 86.5% 4.1% 3.1% 6.2%

Conclusions: (1) Virtually no true vegetarians come from Spanish-speaking


countries. (2) Very few people from Spanish-speaking countries even report not
eating meat.

7Note that these are population estimates using the NHANES dataset. They may differ from
population estimates using other datasets. For example, the US Census Bureau estimates that 49%
of the population is male and 51% female.
9

Education of adult respondents


Proportion Less than 9th-11th High Some College
of… 9th grade grade School college graduate
Sayveg 6.9% 5.8% 13.9% 32.2% 41.2%
Trueveg 2.1% 0.3% 6.5% 26.4% 64.8%
Nonvegveg 10.5% 9.8% 19.3% 36.5% 24.0%
Nomeat 5.9% 13.3% 14.3% 21.9% 44.6%
Population 6.1% 12.8% 23.6% 29.8% 27.5%

Conclusions: (1) True adult vegetarians almost all went to college. (2) Not eating
meat is more closely aligned with the education levels of the population as a
whole, though it still skews educated.

Education of household reference person


Proportion Less than 9th-11th High Some College
of… 9th grade grade School college graduate
Sayveg 5.1% 5.5% 20.4% 31.7% 36.5%
Trueveg 0.4% 0.7% 3.7% 37.9% 57.2%
Nonvegveg 8.5% 8.9% 32.4% 27.2% 21.7%
Nomeat 6.2% 4.9% 18.5% 27.3% 42.8%
Population 6.6% 12.2% 23.5% 28.9% 28.6%

Conclusions: When we include all the children as well as the adults and use the
education level of the household reference person (i.e. an adult in the household)
the education differential just described is even more pronounced. True
vegetarians overwhelmingly come from college-educated households.

Summary conclusions
• About half of all self-identified vegetarians regularly eat meat or fish.
• There are few pescetarians.
• The gender effect is strong—more women than men say they are
vegetarians, are vegetarians, and have meat-free days.
• “True vegetarians” overwhelmingly come from educated households and
are not Black or Hispanic.
• A large proportion of people have days when they eat neither eat nor fish.
This cuts across race/ethnicity and education.
• People born in Spanish-speaking countries very rarely have meat-free
days.

You might also like