Let’s be clear, lower tax means lower prosperity, worse public services, worse health and worse education.

The evidence is in the data. The graph below (updated 2017-06-08 following suggestions in the comments) shows* outcome* measured using the UN’s inequality-adjusted human development index (IHDI) as a function of *input* measured using tax as a percentage of national income (GDP) using data collated by the Heritage Foundation. IHDI is a number that measures income, life expectancy and years in education.* The higher the score the better. Top of the IHDI league table (see here) is Norway (which despite its oil wealth also has high taxes), bottom is the Central Africa Republic. The UK (in colour) is currently 13th, better than the US, but not as good as Germany, Australia or any of the Scandinavian countries. Included are all countries where data is available which includes more than 95% of global population.

The line assumes a diminishing return with increasing tax. However the exact form of this line is a bit arbitrary as we do not know what kind of correlation to expect. We could argue that countries above the line are getting relatively good value for money whereas those below the line could do better although there is a large country-by-country variation that depends on individual circumstances, such as Norway has oil as well as high taxes.

A simple linear fit gives a Pearson correlation coefficient *r* = 0.77. This tells us that prosperity outcome is roughly proportional to government input – the size of the governments spend and tax circuit. The more you spend and tax, the better the quality of life, the better your life expectancy and the better your education.

The conclusion is clear. A vote for a tax party of low tax is a vote for a worse life, worse health and worse education. On Thursday, why not vote for prosperity instead?

* IHDI also includes a correction for inequality but in most cases there is not much difference between the HDI and IHDI.

Hi Charles

Nice graphic and strong correlation

I seem to remember the circle size is the log of the country’s population.

One slight quibble – the version of the IHDI table I get when following the link (table 3) has the US in joint 10th place with Canada and the UK at 16th place (2015 data) which does not agree with your text.

Probably worth checking?

Circle area is proportional to population.

The table at

http://hdr.undp.org/en/composite/IHDI

is HDI ranked. IHDI is the 2nd column. UK rises from 16th to 13th as 4 ‘countries’ (Singapore, Hong Kong, NZ and Liechtenstein) are missing in the IHDI dataset and Finland jumps above UK as it is more equal.

Charles

thanks for the clarification. Some colour coding might be quite nice; China India and the US are fairly easy to identify but excellent work.

The UK is now in colour.

I wonder if I should conclude that a sovereign currency issuer tends to benefit when more of its currency is issued? And when more of it is taxed it just reduces inflation and encourages what the Tories would call deficit minimisation?

Hi. Interesting blog. Does it seem realistic that there is a linear relationship between the two variables? Just intuitively it would seem that at low tax/GDP levels, the improvement in HDI might have greater impact (gradient) but this would flatten out as tax as a percent of GDP headed toward high levels (in the sense if tax was say 100% of GDP you would have probably exceeded a point of diminishing returns). Would a polynomial fit the data better?

Plus, while I hesitate to come over as “awkward” on my first (and therefore possibly last post) you did say on the Tax Research blog (http://www.taxresearch.org.uk/Blog/2017/06/05/37411/) that “On the Pearson correlation coefficient – reducing a data set to a single number is always dangerous! In Physics we do not use it!”. While I agree that the basic thrust of your point may well be correct, isn’t there a risk that this just replicates the sort of “fluffy” analysis that tends to plague most economics?

I’ll get my coat …

No need to get your coat, these are good points. I agree at some point the curve must saturate or turn over and a polynomial may be a better choice than a straight line, then there are other factors such as should I weight the points according to population. Basically, I have a lot of freedom as to how I choose to fit the data and we can argue about whether the fit is useful at all – I did it because people wanted to know the r coefficient. The data is not fluffy (assuming we trust the econometricians) and my advice is always to look at the data and make up your own mind.

PS. I might try your suggestion of a polynomial fit just to see what happens.

I think it was Jonny van Neumann who said

“With four parameters I can fit an elephant, and with five I can make him wiggle his trunk”. I think I might have been one of the people who suggested thercoefficient or at least some metric to see how statistically significant the correlation was. I think David’s point re saturation is a good one and I would probably fit an expression such ask(1-exp(t/tau)) – where t is the “x” axis. This is equivalent to charging an RC circuit and measuring the voltage across the capacitor. I would have a lot more confidence in ana priorimodel based approach rather than a polynomial.Good point. I will try to fit a saturation curve. I will probably let the origin float as well.

Figure updated now with saturating fit.

Charles Thanks.

Letting the origin float is sensible as that is done by default with a best line fit. I liked the flag (Union Jack) graphic. It would be great also to see flags for our near neighbours; Scandinavia, Ireland, the Netherlands, France and Germany, but I suspect that the Irish and Danish flag for example might be too small to render properly.