Health effects of various power sources

Consumption per capita vs health outcomes -- cancer rates (cancer deaths as % of all deaths), cancer mortality (deaths per capita per annum) and life expectancies (male and females separately).

Consumption per capita vs cancer rates (i.e. deaths as % of all deaths)

We see that both coal and hydro power are associated with a decrease in cancer rates across the OECD countries. For each 1000 Kt/mn capita of coal consumption, there's an associated decline of 4.6 points in the cancer death rates. Similarly, for each 100,000 GWh/mn cap of hydro power consumption there's an associated decline of around 10 points of cancer death rate. Each of these associations is about the 95% confidence level, and we can see from the associated r2's that the single connection "explains" at least 1/4 of the variations seen across the dataset. (This is relatively high in epidemiological terms ;-). But we also see there is no significant connection between nuclear power consumption and cancer death rates. I.e. the confidence is less than 60% and very low explanation power (i.e. r2 < 1%).

Summary: coal and/or hydro power seem related to lowered cancer rates; nuclear power seems related to no change in cancer death rates. We might speculate that OECD countries substituting nuclear power for other forms of energy would have an associated (relative) increase in cancer death rates.

Consumption pc vs cancer mortality (i.e. deaths per 1000 capita)

The most significant relationship seem to link hydro power consumption with decreased cancer mortality. I.e. for each additional 1000 GWh/mn cap of hydro consumption there's an associated reduction of around 3.1 cancer deaths per 1000 capita. This relationship has a borderline "good" confidence of around 90%, and it's explanation power shows about 10% of the variation in cancer mortality across the dataset is explained by this single relationship. However, we see the relationships between the other variates and cancer mortality are significantly weaker. We'd be generally loath to accept that nuclear power, e.g., had any significant link with cancer mortality (either increased or decreased) since it's confidence is less than 80%.

Consumption pc vs overall mortality (deaths per 1000 capita, pa)

Here we see a situation that's the reverse of the above. I.e. there is a significant and powerful relationship between nuclear power consumption and overall deaths rates, but there is no significant relationship between coal or hydro power consumption and annual death rates. Note that nuclear power consumption seems to "explain" about 20% of the variation in death rates across the dataset. For each 1000 GWh/mn cap in additional nuclear power consumption there's an associated increase of around .24 death per 1000 capita per annum.

Consumption pc vs life expectancy

We note there is no significant relationship between power consumption and life expectancy. At the very weak level of around 78% we find that increased coal consumption may be associated with increased male life expectancy. It isn't clear from this study what might explain even this weak relationship, but it's possibly due to a "healthy worker effect" associated with male-dominated industries (e.g. coal mining, steel production, etc).

Consumption (gross) vs cancer rates

*** Models that have a Spearman significance of better than 5% are marked thus. I.e. the ordering of data by dep and indep variates is "similar" and we're required to reject an hypothesis of no relationship between them. In those models with +ve betas, those countries with the largest value of indep variate generally correspond with those with the largest dep variate; similarly for the smallest value for the variates. In those models with -ve betas a higher indep variate generally corresponds with a lower dep variate, and vice versa.

Consumption (gross) vs cancer mortality (deaths per 1000 cap, pa)

We see that all relationships are very significant. And we also see the Coal and Nuclear models have extremely high explanation power (i.e. the single relationships explain almost all variation seen in the datasets). And each relationship shows a positive relationship -- in each case, an increase in gross consumption is related to an increase in cancer mortality (deaths per 1000 capita, pa). For each 1000 Kt of coal consumption pa there's an associated increase of 6.7 cancer deaths per 1000 cap; for each 1000 GWh of hydro consumption there's an increase of 1.2 cancer deaths per 1000 cap; and for each 1000 GWh of nuclear power consumption there's an increase of around .7 cancer deaths per 1000 capita. While the "hydro" variate has high significance from a T-test on the OLS \beta, it's interesting to note it has only "low" explanation power as measured by the r2 (i.e. 20% vs >= 85%). It's also the only model of the 3 to fail a Spearman test at 5% CI.

Consumption (gross) vs overall mortality (deaths per 1000 cap, pa)

Here, gross hydro consumption (in GWh per annum) is significantly related to reduced mortality, whereas coal and nuclear power are "weakly" related to reduced mortality. Note also, the explanation power of the "hydro" variate is significantly better than the others. For each 1 mn GWh in hydro consumption across the dataset, there's an associated reduction of around 5.4 deaths per 1000 cap, per annum. For each 1 mn GWh of nuclear power consumption there's (weakly) a reduction of around 1.6 deaths per 1000 cap, pa; and for each 1 mn Kt of coal consumption pa, there's a reduction of around 16 deaths per 1000 cap, pa (normally any population experiences from 100 to 200 deaths per 1000 cap, pa).

Consumption (gross) vs life expectancy (male, female in years)

We note the relationship between life expectancy and consumption is significant in the case of nuclear power, and "weak" in the case of hydro power. (It's also interesting to note that the "hydro" variate is significant at better than 5% in a non-parametric Spearman test; the other 2 models fail). All other relationships we'd classify as "non-extant". In addition, the nuclear power consumption data "explains" from 30 to 40% of the life expectancy data. But note -- the relationship is NEGATIVE. For each additional 1 mn GWh of nuclear power consumption in the OECD countries of the dataset there's a REDUCTION of around 2.8 years of life expectancy for men, and around 2.9 for women. If we accept the 80% CI for the hydro model, we'd accept for each 1 mn GWh of hydro power production in OECD countries there's an increase of around 3.1 years of life for females (and no apparent effect on male life expectancy).

Summary

We therefore see there is some evidence that coal power is associated with decreased cancer rates, and increased cancer mortality. Hydro power is associated with decreased cancer rates, increased cancer mortality, and decreased overall death rates. Nuclear power is associated with decreased cancer rates, increased cancer mortality, decreased life expectancies, and increased overall death rates pa.

Obviously none of the relationships is completely or obviously causal.

I.e. we can't say that "simply" a higher level of (gross) hydro power consumption causes a decrease in the death rate; it's equally likely (a priori) that those countries that invest in higher levels of hydro power (Norway, NZ, Switzerland) are also those that, for various reasons, have lower overall deaths rates (Italy, Japan, Turkey).

Similarly, we can't conclude more per-capita nuclear power consumption results in an increase in the death rate pa or lowers the life expectancy of a population. It is just as likely, a priori, that those OECD countries that simply have higher death rates (Sweden, UK) and/or shorter life expectancy (US, UK) -- for some reason -- are those most likely to implement higher levels of nuclear power consumption (Sweden, France) than others (Italy, Netherlands).

(Of course, if there seems no relationship between the different national characterisations -- the limiting cases appearing in the datasets used here appear in parens, above -- don't seem to obtain then the assumption that "it just happens to be the case because the data is like that" is questionable, and we'd be forced to re-examine the possibility of a causal link).

Power consumption vs national wellness

Models

All 3 forms of power are associated with greater Human Development Indexes. I.e. the more developed nations are those that consumer more power, no matter what its source. It's interesting that, for a given change in power consumption, hydro power is associated with around 4 times more HDI then nuclear power. Hydro power is associated with "more development" than nuclear power. The HDI measures a combination of factors like literacy and life expectancy.

Of the 3 forms of power, only hydro is associated with GDP per capita -- an economic rationalist form of "national wellness". For each 1000 GWh of hydro consumption, there's an associated $US20 per capita of additional GDP.

Only nuclear power seems associated with unemployment. For each 1 mn GWh of additional nuclear power consumption there's an associated decrease of 2.8 points of unemployment. (The unemployment rate is the proportion of the workforce currently out of work).

However, only hydro power is associated with overall employment. The "participation rate" indicates the number of people in the so-called "economically active" part of the population -- usually the 15-65 yo's -- that are in the workforce, whether currently employed or not. For each 100,000 GWh of hydro consumption there's a corresponding increase of around 6.2 points of participation. It may be that hydro power has a significant "flow on" effect to other sectors of the economy and nuclear and coal power does not. (Note also the well-known relationship between workforce and unemployment. For the OECD countries each additional 1% of the 15-65's in the workforce corresponds to approx 0.1 points of increased unemployment).

Finally, all 3 forms of power have a relationship with life expectancy. While increased coal and nuclear power consumption are associated with reduced life expectancy, hydro power is associated with increased life expectancy. (Some of these associations are "weak"; but the association between nuclear power and decreased male life expectancy is "high", at 93% CI). The differences between patterns in male and female life expectancy are due to several factors. Two important ones are (a) the so-called "healthy worker effect" -- whereby persons employed tend to have a longer life expectancy than those not employed -- and (b) the difference between male and female attitudes to personal health; males pay generally pay less attention to symptoms and pay fewer visits to doctors than do females. So countries with more doctors but less employment (note: not the same as "less unemployment") would tend to have a +ve impact of female health and a negative impact on male health.

[As one indication of the kind of negative link possible:

27 Mar 1999 Cardiff. The Brit govt has agreed to pay as much as $A5.12 bn to former miners suffering from lung diseases. The largest industrial compensation case in Brit history may benefit as many as 100K former miners who contracted chronic bronchitis and emphysema from breathing coal dust. The miners took action against the govt and the nationalised coal industry 8 ya, and won the case in court in 1997.

ADVISORY: Email to these sites is filtered. Unsolicited email may be automajically re-directed to the relevant postmaster.