Genetically modified

 
What Demographic and Psychological Factors Predict GM Policy Support and Purchasing 
Behavior?
Multiple regression models predicting policy support 
I next conducted a series of multiple linear regression models in order to explain variation among 
the six policy preference variables described in the previous section. The objective of these 
analyses was to explore the relative importance of the knowledge, perceived risk, trust, and 
demographic variables described above in the extent to which they predict support for policies 
surrounding GM foods. Across all of the following regression analyses I examined correlations 
among my predictor variables in order to check for multi-collinearity; the correlations were below 
thresholds for concern. Table 7 displays the results of a series of multiple linear regression models 
Does not avoid 
Little effort 
Moderate effort 
Strong effort 
Policy Mean 
95% CI 
Mean 
95% CI 
Mean 
95% CI 
Mean 
95% CI 
1 
3.87 
3.69-4.05 
4.63 
4.47-4.78 
4.65 
4.39-4.92 
4.65 
4.21-5.10 
2 
3.53 
3.33-3.74 
4.48 
4.30-4.67 
4.57 
4.29-4.85 
4.65 
4.21-5.10 
3 
4.21 
4.05-4.34 
3.16 
2.92-3.40 
2.55 
2.24-2.86 
1.87 
1.45-2.29 
4 
3.92 
3.72-4.11 
3.05 
2.80-3.30 
2.12 
1.83-2.42 
1.70 
1.18-2.21 
5 
4.08 
3.93-4.24 
2.82 
2.60-3.04 
1.84 
1.61-2.06 
1.35 
1.10-1.60 
6 
2.04 
1.82-2.25 
2.65 
2.38-2.91 
3.61 
3.25-3.97 
4.26 
3.77-4.75 

---

25
predicting each of the six policy support measures, with the most supported policies on the left and 
the least supported policies on the right. All predictor variables were standardized such that a one-
unit increase corresponds to a one standard deviation increase in that variable. As a result, the 
Constant value for each model is the average level of support for that policy, in addition to being 
the y-intercept. “Risk” is an index of risk perceptions across five questions; raw scores were 
divided by the length of the scale (5 or 7) and then added together; higher values indicate higher 
risk perceptions. The variables “Genetics”, “Regulation”, and “Self-assessed” refer to those types 
of knowledge. “Genetics” is a composite of Qs 4 and 5 and “Regulation” is a composite of Qs. 2 
and 3 (shown below in Table 8); respondents either answered zero, one, or two of these questions 
correctly. “Ideology” is a composite of respondents’ political ideology across fiscal and social 
issues; higher values indicate more liberal views on social and fiscal issues. Coefficients are 
reported for statistically significant and statistically suggestive (p < 0.10) results. 

---

26
Table 7. Results of Multiple Linear Regression Models Predicting Support for Different Policies 
(1) 
(2) 
(3) 
(4) 
(5) 
(6) 
Label 
Mutagenesis 
Label 
GM foods 
Grow in 
Dev. World 
Gov’t 
Funding 
Grow in the 
US 
Ban 
GM 
foods 
Variable 
Risk
0.29** 
0.39** 
-1.03** 
-0.91** 
-1.15** 
0.81** 
(.000) 
(.000) 
(.000) 
(.000) 
(.000) 
(.000) 
Genetics
-0.19** 
-0.22** 
0.14* 
0.12** 
(.002) 
(.001) 
(.039) 
(.006) 
Regulation
-0.12^ 
-0.09* 
(.066) 
(.036) 
Self-assessed 
-0.19** 
-0.22** 
(.004) 
(.002) 
Importance
0.15* 
0.19* 
(.046) 
(.019) 
Liberal 
ideology 
0.14* 
(.028) 
T/F Q1 
Education 
Income 
Age 
Constant 
4.26** 
4.04** 
3.46** 
3.19** 
3.14** 
2.67** 
Observations 
267 
267 
267 
267 
267 
267 
R-squared 
0.23 
0.28 
0.60 
0.46 
0.77 
0.39 
Notes. Standardized regression coefficients are reported with their associated p values in parentheses. ^p < .10, *p < 
.05 **p < .01 
 
Table 8. True/False questions asked to respondents 
1. GM foods are sprayed with more pesticides than crops grown conventionally 
2. Foods that have been genetically modified can be labeled as USDA Organic 
3. Foods that have been created through mutation breeding can be labeled Organic 
4. All fresh produce contains genes that have been altered by humans 
5. Genetic modification of food crops alters fewer genes than conventional breeding
In general, no sociodemographic information reliably predicted support for any of the 
policies, controlling for the other variables. Over and above sociodemographic information, 

---

27
perceived and actual knowledge personal importance of GMs, and risk perceptions reliably 
predicted support or opposition to all six policies, and all estimates are highly significant. For 
growing GM foods in the developing world (3) and banning GM foods (6) risk perceptions was 
the only variable that predicted support, over and above other measured constructs. A one standard 
deviation increase in risk perceptions corresponds to a one-unit decrease in support of growing 
GM foods in the developing world and four-fifths of a unit increase in banning the growth of GM 
foods in the US.
The models predicting support for labeling of GM foods (2) and mutagenesis (1) performed 
very similarly to one another. Higher risk perceptions predicted more support for labeling and the 
effect was slightly larger for GM foods (adjusted r-square = 0.39 for GM vs 0.29 for mutagenesis). 
A lower knowledge of genetics predicted more support for both labeling policies, as did lower 
self-assessed knowledge. This suggests that those with more knowledge of genetics want labeling 
less, over and above how much they think they know about GM foods. Additionally, it suggests 
that those who think they know more about GM foods are also less supportive of labeling, 
regardless of how much they actually know. Finally, perceiving the issue of GM foods to be more 
important predicted higher support for labeling in general, and the effect was again slightly higher 
for GM foods than for mutagenesis (0.19 vs. 0.15). Given that, on average, it appears that 
respondents guessed on the mutation breeding question (3) it is plausible that respondents 
answered questions about mutagenesis similarly to questions about GM foods and, therefore, the 
models behaved similarly.
Three variables other than risk perceptions predicted support towards government funding 
of GM foods. Higher levels of genetics knowledge was associated with increasing support for 
government funding. In addition, a more liberal political ideology predicted support for 

---

28
government funding, controlling for all other variables. This finding is perhaps unsurprising given 
that liberals are viewed as more willing to support government funding in general. Finally, higher 
levels of knowledge of GM regulation was associated with lower levels of support for government 
funding, though the result is not statistically significant at the p < .05 threshold. Two of the same 
variables that predicted support for government funding also predicted support for growing GM 
crops in the United States. Again, higher levels of genetics knowledge were associated with 
increasing support for growing these crops and higher levels of knowledge of GM regulation 
predicted lower levels of support for this policy. This could be explained if having more knowledge 
of regulation was associated with restrictions to GM foods in general; however, it doesn’t have 
that association nor does more knowledge of regulation predict avoidance of GM foods, as 
described below.