====================================================================== Rule #1: YOU MUST clip all extraneous text when replying to a message. ======================================================================
On 11/06/2010, at 1:09 AM, Ian Angus wrote: > DISSECTING THOSE ‘OVERPOPULATION’ NUMBERS: > PART TWO: THE PERILS OF PER CAPITA > http://climateandcapitalism.com/?p=2560 I've taken the liberty of putting a comment I made on the climate and capitalism site here as well, as I think it'd be of general interest, and as noted I'm interested in feedback. ----------------------- Ian, I’ve got a few points on the statistics of all this, which while a bit complicated I think will be helpful for the layperson (and those more knowledgeable about climate change and/or statistics might like to point out any errors). Firstly: distributions, such as of income, are indeed important to understand, rather than just a central tendency, but the latter is often important as well. With regard to this tendency, an important point is that a median is not the same as the average, or mean. The mean of a set of values comes from adding all the values up and dividing by how many values there are; the median is lining all the values up and picking the one in the middle. They are in fact the same in the familiar bell curve distributions which many measurements fall into, particularly natural ones, but not skewed ones, like income, and even more wealth, under capitalism, or the cancer survival rates in Gould's essay http://www.phoenix5.org/articles/GouldMessage.html This is important because means as a summery distort a skewed distribution a lot more than a median. If Gould had been given the mean survival rate he would have freaked out more as it would have been even less than the median 8 months he was told, because the bunching towards the bottom pulls the average below what the values for most individuals actually is, whereas the median indicates that half the people so far measured have lasted less than 8 months, and half more, stretching out to the 20 years Gould got and more. Because under capitalism there’s a relatively small range between even welfare recipients and well-paid workers compared to a few bourgeois squillionaires bunching the distribution well towards the top end, an average income figure will be well above what most people earn, while the median tends to be around a typical full time wage. So it can be a useful summary, though often a more detailed summary, like citing a range around the mean or median (as appropriate) or chopping up the distribution into bits such as deciles, is often more useful. Secondly: In preparing for a talk on population, consumerism and the environment for the Socialist Alliance in Melbourne in a couple of weeks, l’ve been thinking about valid ways of statistically showing actual relationships between environmental impact, population, “affluence” and “technology”. That’s because the (very partial) value of the I=PAT formula is that these factors are in some ways related, even if they don’t in and of themselves tell us the whole story, particularly about causation. What I’ve been mucking around with might prove useful I think in helping us understand the attraction of such formulae, because they relate to real if partial phenomena, and also help us put the case for the real explanations and solutions. What I’ve done is get together, for 12 (so far) countries across the size and wealth spectrums, and recorded annual CO2 emissions (a measure of I), population, GNP (a measure of A). I’m not really sure what the populationists means by T – how much tech? How advanced it is? When the iphone G4 is released does T go up a bit? I’ve got something to say about a valid measure of T below, but as I can’t think of anything easy to look up now I’ll ignore it. Which is lesson number one: when constructing a model the researcher, not God or the universe, decides what variables to include. Anyway I got my I, P and A numbers, and put them into the stats application. I got that to graph I vs P – and there’s definitely a moderate to strong linear positive relationship (for the stats heads, r=0.66). Countries with higher populations tend to produce more C02. If we do that for I vs A, we get a stronger positive relationship (r=0.80). Countries with higher GNPs tend to produce more C02. So you can see why many concerned punters will think, well it’s the population and the affluence causing this shit we’re in. If we want to see if we can make a more detailed mathematical model of how these variables relate together, with I as the dependent variable and the others as the independent variable (i.e. I = something to do with P and A) it actually makes no mathematical sense when we have the independent linear relationships mentioned above just to multiply P and A together. What we (or the computer) do is a technique called a regression, which works out a “line of best fit” of the form: I = a +bxP + cxA Where a,b and c are constants that my trusty Mac Mini works out. These don’t tell you directly about the different contributions of the two variables though, as their scales are completely different, but the program also works out “standardized coefficients” which convert variables to the same scale and so tell you about the proportional contributions. Here we get 0.51 for P, and 0.70 for A. That is, GNP makes about 150% of the contribution to CO2 production that population does in this model (stats heads will be interested that for the model r squared = 0.90, that is these variables account for 90% of the variation in I). So this exercise seems to be useful in providing evidence that, other things being equal, amount of stuff produced is of more concern than population in CO2 emissions (actually if there’s interactions between independent variables you should put in another terms that does multiply them, dxPxA to account for the interaction, but this seems to have little effect here so I’ve ignored it). I think we could add technology to our model in some way. While as mentioned a single measure of technology makes no sense to me, maybe there’s relatively straightforward measures of 2 variables: a “good T”, e.g. an index of how much energy is produced by turbines and solar panels, and a “bad T” e.g. an index of much energy is produced by oil and coal. I reckon these would make a contribution to the equation, with “good T” having a negative coefficient (more good stuff associated with less CO2), and a “bad T” having a positive one (more good stuff associated with more CO2). They’d have to be interactions between our A and these Ts (the effect of extra A would depend on the extent to which it was produced with good or bad T). This could also aid our arguments, by suggesting the importance of changing the tech c.f. population and affluence. But, the populationists might respond, well you’ve shown that population as such still has this big effect, so maybe we still should slash immigration or sterilize undesirables, or whatever their particular paneceas are. This though misunderstands two basic limitations of this sort of exercise. Firstly, the First Commandment of observational science, in which we observe a number of variables extracted from complex real world systems at one point in time, is: CORRELATION AIN’T CAUSATION. We’ve only shown association, and the pattern of causation might lie in underlying factors not so amenable to direct observation, and/or various interactions we haven’t uncovered (as opposed to experimental science, in which we have a lot more control over a limited number of variables and can plausibly show mathematical models of causation). Secondly, a related limitation is that these associations are only valid from time the data was observed. Regression can be used for “prediction”, but only in the sense we can “predict”, from our equation, the CO2 emissions of a country we didn’t use for the analysis, at the same point in time that we obtained the measurements we did use. A pointless exercise, as we could just look the actual number up. The point of the regression in this and many cases is to get an idea of the relative contribution of the different factors. I.e. taking these two limitations together, it’s not valid to say, hey let’s slash population by a certain amount by stopping all immigration next year and presto CO2 will fall by b times that amount. This policy have all sorts of effects, such as a probably catastrophic economic collapse and a rise in xenophobia, a situation which would make developing greener tech and enacting progressive social change harder. In short, we can’t actually prove anything through observation of complex real world systems, only build a case through varied analyses and sources of data. I think we can integrate the above into our arguments along these lines: * Yes we acknowledge the evidence shows that, all else being equal, more people and more stuff is associated with CO2. * But how do we address the issue of population, and of how stuff is produced? * Immigration restrictions and population control are politically perilous, and the former does nothing on the global level. The historical evidence shows that the effective way to reduce population growth is to increase living standards and women’s rights globally. But capitalism is a block to this. * Greener tech will ameliorate or reverse the effects of more stuff, even current evidence shows (I reasonably assumed above) and developing the tech will actually require more of many kinds of stuff (education, infrastructure). Again though capitalism is in the way of developing and generalizing this tech rapidly enough. * Therefore we need socialism (and the struggle for reforms along the way there). ________________________________________________ Send list submissions to: Marxism@lists.econ.utah.edu Set your options at: http://lists.econ.utah.edu/mailman/options/marxism/archive%40mail-archive.com