Introduction:
In the past decades, there has been an ongoing push to improve our natural environment across the globe. However, the level of success in such measures varies to a great degree. I suspect that in many respects, environmental concern can largely be considered a luxury of wealthier societies, because 1) they have already essentially solved basic living/survival issues and 2) they have established a stability in financial and competency resources to begin fixing the environment.
From a personal standpoint, I care about this, because as an American living abroad, I am somewhat ashamed of how little my country does in caring for the environment, despite our relative financial and knowledge resources to do so. On the other hand, I’ve lived in developing countries that also are failing to protect their environment adequately, but at least they have an excuse that they are still developing (e.g. economically weak, lacking infrastructure, political corruption, civil war…)
In the end, I will not explore this issue via country comparisons, but rather by exploring a tiny bit about American’s environmental opinion as compared to how stable their life is more at an individual level.
Data:
I am using the GSS 1972-2012 dataset, whereby over 57000 people over three decades have taken part in this expansive survey (with 114 variables), just with face-to-face interviews of 18+ years of age, mainly in the Chicago area. Clearly there are many generalization issues to contend with from such a relatively narrow sample group, and it’s likely not very valid to generalize these results to the whole US population, much less the world’s. And since this is data from an observational survey, it’s not possible to build up causalities either.
Nonetheless, I will look at respondents’ working status (WRKSTAT) as a variable which can perhaps partially explain their own personal opinions, as a response variable, about further national environmental protection funding (NATENVIR). Both variables are categorical, though NATENVIR is also ordinal.
Quoting/linking as follows from the GSS Codebook:
WRKSTAT - LABOR FORCE STATUS: Last week were you working full time, part time, going to school, keeping house, or what?
NATENVIR - IMPROVING & PROTECTING ENVIRONMENT: We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount.
Exploratory data analysis:
As can be seen below by the tables, the explanatory variable, WRKSTAT, has eight possible categories and the response variable, NATENVIR, has only three.
table(gss$wrkstat)
##
## Working Fulltime Working Parttime Temp Not Working Unempl, Laid Off
## 28207 5842 1213 1873
## Retired School Keeping House Other
## 7642 1751 9387 1132
table(gss$natenvir)
##
## Too Little About Right Too Much
## 19259 9539 2816
A quick look at the mosaic plot below, chosen because we’re comparing two categorical variables, shows surprisingly little variation in proportions. The largest variation appears to be between retired respondents and those in school. This suggests that, regarding my research question, that while there is some variety, perhaps working status is not the strongest indicator for how people feel about the environment.
mosaicplot(table(gss$wrkstat,gss$natenvir))
Inference:
H-null: There is no difference among any levels of working status for the proportion of people who feel there is “too little” national funding going towards environmenal protection.
H-alt: There is a difference among at least two levels of working status for the proportion of people who feel there is “too little” national funding going towards environmenal protection.
Since the sample is so large, there’s no need to do any boot-strapping or other simulations to generate data. Since we’re dealing with two multiple-levelled categorical variables, a confidence interval isn’t really appropriate, which is why a Chi-squared independence test is the obvious analysis to be done. In this case, a success has been defined as those who feel there’s “too little” funding and failures are either “about right” or “too much”.
source("http://bit.ly/dasi_inference")
inference(gss$natenvir, gss$wrkstat, est = "proportion", type = "ht", method = "theoretical", success = "too little", alternative = "greater")
## Response variable: categorical, Explanatory variable: categorical
## Chi-square test of independence
##
## Summary statistics:
## x
## y Working Fulltime Working Parttime Temp Not Working
## Too Little 10026 2020 434
## About Right 4303 906 186
## Too Much 1294 257 75
## Sum 15623 3183 695
## x
## y Unempl, Laid Off Retired School Keeping House Other Sum
## Too Little 695 1916 741 3088 336 19256
## About Right 287 1547 202 1954 151 9536
## Too Much 74 551 44 455 64 2814
## Sum 1056 4014 987 5497 551 31606
## H_0: Response and explanatory variable are independent.
## H_A: Response and explanatory variable are dependent.
## Check conditions: expected counts
## x
## y Working Fulltime Working Parttime Temp Not Working
## Too Little 9518.33 1939.25 423.43
## About Right 4713.69 960.36 209.69
## Too Much 1390.97 283.39 61.88
## x
## y Unempl, Laid Off Retired School Keeping House Other
## Too Little 643.37 2445.54 601.33 3349.05 335.70
## About Right 318.61 1211.08 297.79 1658.53 166.24
## Too Much 94.02 357.38 87.88 489.42 49.06
##
## Pearson's Chi-squared test
##
## data: y_table
## X-squared = 575.034, df = 14, p-value < 2.2e-16
With such a tiny p-value on the Chi-squared independence test, it’s clear that the null hypothesis should be rejected, and therefore there is some association between some levels of working status and those feeling the national government doesn’t fund environmental protection enough.
Looking at the mosaic plot reveals that full-time, part-time, temporarily unemployed, laid-off and other groups are nearly identical to each other; retired people show a definite trend towards less environmental funding and home-makers slightly so, while those in school clearly show the opposite case, being the most-inclined to increase environmental funding. It’s most likely these three groups which are making the p-value so small.
Conclusion:
In the end, it seems that despite the alternative hypothesis being supported, I suspect that my original theory about environmental concern being a luxury for those with stability (at least in employment) was not truly addressed within the data. From that theory, one would’ve expected full-time respondents to prefer more funding for the environment than laid-off people, but clearly this is not truly the case.
In fact, among the traditional employment groups (full-time vs. part-time vs. temporarily unemployed vs. laid-off) there’s hardly any noticeable difference. Actually, the largest discrepancy of all eight working status groups is between two other groups: retirees and students. This suggests that perhaps a better indicator to compare to environmental concern than working status, would be simply the respondents’ age, with younger people (e.g. students) feeling there’s “too little” and older people (e.g. retirees) feeling there’s “about right” or even “too much” funding.
This implies that future researchers interested in this subject ought to focus on age as a determinant to environmental concern, though other obvious variables which come to mind might be political affiliation. However, since I am still not entirely dissuaded from my theory about environmental concern being a luxury, another avenue to pursue might be analysis with personal income being an explanatory variable. Furthermore, I would also be interested to pursue these matters further with country-by-country comparisons, perhaps looking into people’s opinions as explained by their country’s GDP or Human Development Index.
Reference:
The R code to access the data is as follows:
Appendix:
The dataset is too large (n = 31606), so therefore I’m attaching here a small sample of 25 datapoints within the whole dataset:
4975 Working Parttime About Right
4976 Keeping House Too Little
4977 Working Fulltime Too Little
4978 Working Fulltime Too Little
4979 Working Fulltime About Right
4980 Retired Too Little
4981 Keeping House About Right
4982 Working Fulltime Too Little
4983 Keeping House Too Little
4984 Working Fulltime Too Little
4985 Working Fulltime About Right
4986 Temp Not Working Too Little
4987 Unempl, Laid Off About Right
4988 Working Fulltime About Right
4989 Working Fulltime About Right
4990 Keeping House About Right
4991 Unempl, Laid Off About Right
4992 Working Fulltime Too Little
4993 Working Fulltime About Right
4994 Temp Not Working Too Much
4995 Working Fulltime Too Little
4996 Working Fulltime About Right
4997 Working Fulltime Too Little
4998 Keeping House About Right
4999 Working Fulltime Too Much
5000 Working Fulltime Too Little