Disclaimer Disclaimer Disclaimer Disclaimer Home           Home Home           Home About  site   About  site About  site   About  site About  me    About  me About  me    About  me Current reading   Current reading Current reading   Current reading CONTACT   CONTACT CONTACT   CONTACT Links                  Links Links                  Links Book shelves      Book shelves Book shelves      Book shelves
Blog Archive    Blog Archive Blog Archive    Blog Archive Recent posts  Recent posts Recent posts  Recent posts
Having a closer look at the novel ‘Nanny State Index’, and improving their analyses a little bit Today is a big day! Indeed, today is the day a novel metric has been launched – the Nanny State Index……..  I will let that sink in for a bit….(if you don’t know what the nanny state is, have a look here)  The website can be found here (http://nannystateindex.org/), and as you could probably have guessed  from the name of this novel index it comes with all the bells and whistles attached. Being constructed by  the ‘European Policy Information Center’ (EPICENTER), which aims to promote the principles of a free  society from a free-market perspective, and is an initiative of six think tanks including our UK’s very own  Institute of Economic Affairs, there is of course no confusion as to why this new index has been created.   Indeed, the countries in the index are shown in in a variety of colours and range from green (the freest  countries) to red; the least free countries. So unfortunately, having moved from one of the freest  countries in the EU, the Netherlands, I have moved to one of the least free I could have possibly moved to  – the UK. And since everybody knows that green is good and red is bad, it seems I have managed to self-  imprison my wild side. Crap….  Looking at this “novel” Index, I personally understand the inclusion of the tobacco and alcohol components  of the Index, these have after all a pretty big impact on disease risk. However, while I can also understand  the inclusion of the food component this is all relatively new and would not yet be expecting to have any  impact (if at all, but that is another debate), the inclusion of the e-cigarette component is a bit odd – or  maybe premature; the e-cigarette territory is all relatively new and inclusion at this moment would just  distort the Index itself if one is interested in its correlation with life expectancy. In fact, just by eye-  balling the table the e-cigarette component seems to have quite a big impact on the ranking.    And then there are of course some issues of who made this index, how numbers were assigned, etcetera,  but let’s leave that issue aside for a moment.  The ‘Nanny State Index’, or NSI, comes with an analysis page on which it is clearly demonstrated that NSI  score is in no way correlated to life expectancy. There are some other plots in which the alcohol  consumption NSI component is plotted against alcohol consumption (to which, in contrast to the  accompanying text, at face value it seems negatively correlated), and anti-smoking policies against  tobacco-smoking (no obvious correlation), but according to the site the big picture relates to the absence  of a correlation between ‘nanny stating’ and life expectancy. That’s all fair enough; one would imagine  national ‘nudging’ policies – or nanny stating in neoliberalist terminology – to have a small effect anyway;  the idea being that small effects on a national population scale have quite a large impact in total.  The plotted correlations however, are univariate ones only. I mentioned this previously when I reviewed  ‘The Spirit Delusion’ (here), but when one looks at relatively subtle effects without taking into account  other factors – potentially with a much bigger impact – it is easy to miss associations in all the stuff that is  going on ‘behind the scenes’. For example, it has been known for decades that the ‘Mediterranean  lifestyle’ is associated with higher life expectancy (for example <link>), while also investment in and the  actual quality of the healthcare system will have a big impact on the mortality patterns.   The good news is that we can easily do these, somewhat more sophisticated, analyses ourselves. The NSI is  available from their website and data on life expectancy at birth (as well as, out of interest, healthy life  expectancy at birth) and healthcare costs (as a percentage of GDP) can be freely downloaded. Of course,  we also know the geographical location of countries.   I also generated two new NSI which I believe to be more accurate; I did not include the e-cigarette  component since I don’t think that should be in there (see above), and so I have calculated NSI2, which  includes alcohol, tobacco and food policies, and NSI3, which only includes alcohol and tobacco policies.  And then there is one final caveat for what follows below. The NSI includes all 28 countries in the  European Union, but in my opinion Luxembourg should be excluded. It’s GDP per capita is much higher  than of other countries, it has a very peculiar economy, it’s very small and it has a relatively large  proportion of people only living here for a fairly limited time, for example. Note that in the following, no  statistically significant associations are observed when Luxembourg is included. This demonstrates (a) that  Luxembourg clearly is different and exclusion is justified, but (b) that conclusions, by me or others, based  on in/exclusion of one data point remain open to discussion. For the statistically-minded, P-values in that  case are in the order of 0.10-0.20 and model estimates are about the same as the results below, so  Luxembourg mainly just adds noise (i.e. it’s the only model residual >2). So it is up to you whether you  accept this or not, but future analyses are done with the 27 other countries, which in my opinion is the  correct approach. So just to check, if we use the dataset with the 27 countries and we calculate the association between NSI  and life expectancy (at birth) there is indeed nothing there (model parameter ~ 0.02; P-value = 0.69;  adjusted R-squared ~ 0%). So far so good, that is what the ‘nanny state index’ people found as well.   Now let’s get rid of some unnecessary noise in the data that we can explain. By adding whether this is a  Mediterranean country (or not) as well as the healthcare expenditure (as % of GDP) to the model (I pointed  out above why that is a good idea), the model suddenly explains 68% of the differences between countries;  we are getting somewhere! The effect of nanny-stating has doubled with a model parameter of 0.04 -  meaning that for each additional index point, whatever that means, life expectancy at birth increases by  0.04, or 2 weeks – and the corresponding P-value is about 0.20.  So that is quite interesting. Now let’s get rid of this weird e-cigarette component in the NSI; the one that does not really make sense.   This model now explains 72% of the differences between countries, because we removed some irrelevant  noise, and we now see a statistically significant (P~0.02) increase of 0.09 (4.5 weeks) per additional point  of the NSI (alcohol, tobacco + food policies). Indeed, once we conduct some better analyses than the ones  presented at www.nannystateindex.org website, it seems that ‘The Nanny State’ – or more politely,  national public health policy – does have a beneficial effect on life expectancy…  In fact, if one removed the food component as well, which I am not sure should be in there just yet, the  model fit further improves (75%) and it seems that each NSI (alcohol + tobacco policy) point increase  increases average life expectancy by almost 6 weeks.   More important than just life expectancy, in my opinion, is healthy life expectancy since this refers to  your expected number of years without disease. If we put this metric in the model we find that for every  NSI ((alcohol + tobacco policy) point increase, we extend our healthy life (on average) by 4.7 weeks (P-  value~0.04; R-squared ~ 66%).  Now there are many ifs and buts with the analyse above (see, for example, the figure above), but these  are essentially the same ones as with the original analyses on www.nannystateindex.org. What I have  added here are just some straightforward refinements, and I am sure better ones with additional  important variables (preferably with more data, but well, there are only 28 countries in the EU) can be  done.   Nonetheless, I really am quite happy with the above; especially given that the actual policies we are  talking about here do not forbid anything and residents in any of the red countries in the list are just as  free as people in the other EU countries to buy and consume what they want. In fact, these are just forms  of – fairly unintrusively – nudging people towards healthier lifestyles.   And as we see above, there is more evidence that this makes people live longer, and live longer healthier,  than that these policies don’t do anything at all. This, interestingly, fits in nicely with evidence for  reducing health inequities and health inequalities (review of recent book here) and against neoliberalist,  free-market fundamentalist, policies (review of recent book here).   So thank you very much Nanny State Index people, this new index is excellent! 
Back   Back Back   Back