Notes on concern about global warming as measured by the FIU/Florida Poll
Ten times since 1993 the FIU/Florida Poll has included a tracking question on global warming: How concerned are you about the problems of global warming and loss of the ozone layer -- are you very concerned, somewhat concerned, not very concerned, or have you not heard much about this problem?
"Loss of the ozone layer" is included in the wording since it was a major public concern in 1993, and has been kept to retain consistency of wording. The question context in the poll in terms of questions preceding and following this one is also kept constant.
Over the time period response to the question has been fairly constant, with around 30% reporting that they are very concerned. In the most recent poll done in the summer of 2006 that number rose to 44% very concerned. Over all ten years 32.5% of the people who answered the question were very concerned, 35.9% somewhat concerned, 22.6% not very concerned, and 8.7% said they had not heard much about the problem. A total of 12,207 people answered the question, and 78 were asked it but did not give an answer.
The table and chart below show the percentages for each year:
1993 |
1996 |
1997 |
1998 |
2000 |
2001 |
2002 |
2003 |
2004 |
2006 |
|
Very concerned |
31% |
28% |
32% |
31% |
32% |
32% |
32% |
28% |
35% |
44% |
Somewhat concerned |
35% |
38% |
36% |
35% |
36% |
39% |
36% |
39% |
35% |
30% |
Not very concerned |
23% |
24% |
24% |
25% |
23% |
22% |
22% |
24% |
22% |
21% |
Haven't heard about it |
10% |
10% |
9% |
9% |
9% |
7% |
10% |
9% |
9% |
6% |
Relation to demographics
Some results fit with the conventional wisdom, others do not. These tables show percentages of respondents who say they are very concerned. Appendix 2 has charts for all the responses.
Percent very concerned by education
grade school |
some high school |
high school grad |
some college |
college grad |
graduate school |
32% |
31% |
31% |
33% |
34% |
35% |
Percent very concerned by household income
under $10,000 |
$10,000 - $20,000 |
$20,000 - $30,000 |
$30,000 - $50,000 |
over $50,000 |
34% |
36% |
34% |
31% |
30% |
Percent very concerned by other demographics
men |
women |
non-Black |
Black |
non-Hispanic |
Hispanic |
Democrat, Independent, or other |
Republican |
31% |
35% |
32% |
38% |
31% |
43% |
38% |
23% |
Percent very concerned by region of
North |
West-Central |
Southeast |
29% |
30% |
39% |
Multivariate Analysis
A problem with tables like the above is that different demographics overlap in their effects. For example, wealthier people are more likely to have more education. One way to deal with this is to use a multivariate analysis that uses all the demographics and shows the unique contribution of each one. In this case the procedure used is logistic regression. In the procedure the change in “odds” of saying “very concerned” is estimated. An odds of two means doubling the likelihood of saying “very concerned”.
It is important to note that if two variables are correlated with each other, the significance of the unique contribution of each in the regression analysis decreases. For example, the results above show that both African-Americans and lower income people are more likely to be concerned about global warming. However, in the multivariate analysis showing only the unique (non-overlapping) effect of each, neither is significant.
Variable |
Change in odds of being very concerned as result of variable being true (1.0 = no effect) |
Statistical significance = confidence that another survey like this one would produce a similar result |
increase in one year of age |
no significant effect |
|
being female |
increases odds of being “very concerned” by 1.15 |
significant |
being Black as opposed to white non-Hispanic |
no significant effect |
|
being Hispanic as opposed to white non-Hispanic |
increases odds of being “very concerned” by 1.61 |
significant |
additional year of education |
increases odds of being “very concerned” by 1.06 |
significant |
additional $10K of annual income |
no significant effect |
|
being Republican |
decreases odds of being “very concerned” by .48 |
significant |
living in southeast
|
increases odds of being “very concerned” by 1.27 |
significant |
Looking at changes over time
Change in odds of being very concerned as result of variable being true (1.0 = no effect) |
|||
variable |
1997-2000 |
2001-2003 |
2004-2006 |
being female |
1.26 |
Not significant |
Not significant |
additional year of education |
1.04 |
1.04 |
1.08 |
being Hispanic as opposed to white non-Hispanic |
1.73 |
1.58 |
1.34 |
being Republican |
0.57 |
0.47 |
0.37 |
living in southeast
|
1.16 |
1.37 |
1.39 |
Two stories of interest are increasing polarization over time and fairly constant concern of Hispanics.
Hispanic effect may be questionnaire wording or cultural Question wording may show some effect of ozone layer being included.
How concerned are you about the problems of global warming and loss of the ozone layer -- are you very concerned, somewhat concerned, or not very concerned, or have you not heard much about this problem?
¿Cuán preocupado a está usted con los problemas
Probably is result of Hispanics not buying in to the polarization, responding to the issue on its own terms.
Note the low amount of variance
explained (R2 ). One reason is that in
Appendix 1 logistic regression
. logistic gwconc yage female xblack xhispanic xyred inc10 gop soeast [pweight = wgt], robust
Logistic regression Number of obs = 7452
Wald chi2(8) = 282.65
Prob > chi2 = 0.0000
Log pseudolikelihood = -4550.0325 Pseudo R2 = 0.0333
------------------------------------------------------------------------------
| Robust
gwconc | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
yage | 1.001796 .0015253 1.18 0.239 .9988111 1.00479
female | 1.147679 .0606268 2.61 0.009 1.034797 1.272875
xblack | 1.149108 .0971737 1.64 0.100 .9735973 1.356258
xhispanic | 1.613307 .1172822 6.58 0.000 1.399064 1.860359
xyred | 1.064105 .0126202 5.24 0.000 1.039655 1.08913
inc10 | .982817 .0091369 -1.86 0.062 .9650712 1.000889
gop | .4869967 .0294724 -11.89 0.000 .4325262 .5483269
soeast | 1.275521 .0724246 4.29 0.000 1.141185 1.425671
------------------------------------------------------------------------------
. logistic gwconc female xhispanic xyred gop soeast [pweight = wgt] if yearcat==1, robust
Logistic regression Number of obs = 2880
Wald chi2(5) = 82.00
Prob > chi2 = 0.0000
Log pseudolikelihood = -1749.9347 Pseudo R2 = 0.0242
------------------------------------------------------------------------------
| Robust
gwconc | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | 1.256307 .1045083 2.74 0.006 1.067301 1.478783
xhispanic | 1.739705 .1952153 4.93 0.000 1.396243 2.167657
xyred | 1.037074 .0178286 2.12 0.034 1.002712 1.072613
gop | .5712301 .0532153 -6.01 0.000 .4758981 .6856591
soeast | 1.162902 .1053928 1.67 0.096 .9736423 1.38895
------------------------------------------------------------------------------
. logistic gwconc female xhispanic xyred gop soeast [pweight = wgt] if yearcat==2, robust
Logistic regression Number of obs = 2877
Wald chi2(5) = 104.61
Prob > chi2 = 0.0000
Log pseudolikelihood = -1715.8104 Pseudo R2 = 0.0326
------------------------------------------------------------------------------
| Robust
gwconc | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | 1.075459 .091206 0.86 0.391 .9107647 1.269934
xhispanic | 1.585523 .1741904 4.20 0.000 1.278372 1.966474
xyred | 1.036819 .0180613 2.08 0.038 1.002017 1.07283
gop | .4695693 .045537 -7.80 0.000 .3882876 .567866
soeast | 1.371528 .122735 3.53 0.000 1.150886 1.634469
------------------------------------------------------------------------------
. logistic gwconc female xhispanic xyred gop soeast [pweight = wgt] if yearcat==3, robust
Logistic regression Number of obs = 1736
Wald chi2(5) = 102.46
Prob > chi2 = 0.0000
Log pseudolikelihood = -1092.8245 Pseudo R2 = 0.0518
------------------------------------------------------------------------------
| Robust
gwconc | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | 1.193915 .1315194 1.61 0.108 .9620699 1.481631
xhispanic | 1.341519 .1993328 1.98 0.048 1.00258 1.795042
xyred | 1.083899 .0252573 3.46 0.001 1.035509 1.13455
gop | .3676042 .0448317 -8.21 0.000 .2894482 .4668636
soeast | 1.388519 .1630524 2.80 0.005 1.103052 1.747864
Appendix 2 – Charts
