Saturday, June 18, 2011

The relationship between hunger and petroleum consumption-Part 1

In part 9 of my previous series, I estimated that of the 25 barrels of petroleum consumed per person per year (b/py) in the USA, at most 1.7 b/py was needed to support the present food production system. 

In part 10, based on the analysis of population growth and per capita petroleum consumption trends for some selected countries and regions, I down-graded my estimate of the per capita petroleum needed to sustain the food production system, and therefore prevent starvation and population decline, from 1.7 b/py to between 1.2 and 0.2 b/py.  For the purposes of predicting future population trends, I took 1.1 b/py as my best-estimate of the minimum per capita consumption level needed to sustain our current petroleum-driven food production system.

One of the problems I have had when trying to relate population growth/decline to rates of petroleum consumption is that at some point, a country or region gets food-aid from other countries or regions to prevent what would otherwise be famine and death. This is the way, I think, that some countries, like Ethiopia, can continue to have very low levels of petroleum consumption (e.g., <0.2 b/py) but still have population growth at 3% per year. 

Is this an insane, unsustainable arrangement?  Why, yes, I think so, but that is the arrangement that presently exists for many of the poorest countries of the world.

In this article I have used a softer measure of a country's food production system’s ability to sustain its population: hunger. 

If a population is hungry, then starvation and death, and probably social unrest, may not be too far behind. 

Of course, at present, foreign food aid will step in and try to mitigate the situation, and therefore, complicate any possible relationship between population growth and petroleum consumption, which is what I have been trying to quantify. 

Here’s my thinking: the effects of food-aid would not have as large an effect on hunger as it would on population decline.  That is, foreign food aid might prevent a net population decline but not fully eliminate hunger.  If this is the case, then maybe we can still see a relationship between hunger and petroleum consumption.  Essentially, hunger is serving as a softer proxy for population decline due to starvation.

Here’s my underlying hypothesis: if the amount of per capita petroleum consumption is important to a country's ability to sustain a petroleum-driven food production system, then the lower the per capita petroleum consumption, the higher the level of hunger in that country. 

Okay, let's see what we can find out.

The Global Hunger Index (GHI)

The International Food Policy Research Institute has developed the GHI as a means to quantify hunger:

The GHI incorporates three interlinked hunger-related indicators ...
1. The proportion of undernourished as a percentage of the population (reflecting the share of the population with insufficient dietary energy intake);
2. The prevalence of underweight in children under the age of five (indicating the proportion of children suffering from low weight for their age); and
3. The mortality rate of children under the age of five (partially reflecting the fatal synergy between inadequate dietary intake and unhealthy environments).
The index ranks countries on a 100-point scale, with 0 being the best score (no hunger) and 100 being the worst, although neither of these extremes is reached in practice. Values less than 5.0 reflect low hunger, values between 5.0 and 9.9 reflect moderate hunger, values between 10.0 and 19.9 indicate a serious problem, values between 20.0 and 29.9 are alarming, and values of 30.0 or higher are extremely alarming.

The International Food Policy Research Institute has made their GHI for 1990 and 2010 available in spread sheet form, which is what I have used for this analysis. 

In practice, any country scoring below 5 is not given a specific score and the scores of developed countries are typically not reported.  Presumably, only the countries at risk of hunger are study, although as you will see there are plenty of countries having a GHI score of < 5, or, a “low hunger” index.

The published GHI data consists of 99 countries for 1990 and 122 countries for 2010.  The International Food Policy Research Institute provided partial data for several MENA countries (e.g., Afghanistan, Bahrain, Bhutan, Iraq, Oman, Qater, Somalia), and Papua New Guinea, but no hunger index value, and therefore, I did not consider these countries.

The increase in the number of countries is mostly from former Soviet Union countries and some European countries that are presented in the 2010 data set, but were not in the 1990 data set. 

Per capita petroleum consumption

For the countries for which I have a GHI value, I calculated per capita petroleum consumption using the petroleum consumption data reported by the EIA (EIA Countries) and the population data reported by the US census bureau (US census bureau’s international database) for 1990 and 2009. 

These are two resources that likely will not be available in the near future, so let’s enjoy them while we can.

The raw data and preliminary impressions

Figure 1 presents a plot of GHI versus per capita petroleum consumption for all 122 countries for which a GHI value in 2010 was reported by the International Food Policy Research Institute.

To make a plot, for all countries having a GHI of less than 5, I assigned a GHI of 2.5.  This helps differentiate from a few countries that have a GHI of about 5 or slightly higher. 

For reference, I also show per capita petroleum consumption values for the USA as an asterisk (*), with an assumed GHI of 0, which is probably not true, but as I noted above International Food Policy Research Institute does not provide a GHI for the USA or many other developed countries.

The trend line in Figure 1 corresponds to a power equation also shown in the figure.  I don’t ascribe any particular meaning to this power curve or its best-fit parameter values—but it does help show the general trend in the relationship between GHI and per capita petroleum consumption.

Figure 2 shows the analogous plot for GHI versus per capita petroleum consumption for all 99 countries reported by the International Food Policy Research Institute.

There is a pretty obvious trend for the hunger index to be higher for those countries with the lower per capita consumption levels.  This trend applies to both the 2010 and 1990 data sets. 

Using the trend line for the 2010 data set, the GHI transitions from the low to moderate hunger category (GHI > 5) when per capita petroleum consumption drops below about 4.5 b/py.  No doubt the value of this transition point is influenced by my assignment of all GHI values of less than 5 being equal to 2.5.  Additionally there are a very broad range of per capita petroleum consumption values for countries in the low “hunger index” category. 

However, I think that the values of the higher GHI categories would be less influenced by this and the per capita petroleum consumption values are in a much tighter range, although there are still a few outliers.  The transition from the moderate to serious hunger category (GHI > 10) is reached at about 1.2 b/py, and the transition from serious to alarming (GHI >20) is reached at about 0.32 b/py.

Similar results are obtained for the 1990 data set, as summarized in the table below:

Per capita petroleum consumption at transition point between hunger categories
Hunger category transition
per cap consumption (1990)
per cap consumption (2010)
low => moderate
moderate  => serious
serious => alarming
alarming => extremely alarming

The trend line suggests that GHI is much higher in those countries where per capita petroleum consumption drops below 1-2 b/py.  This is pretty consistent with my previous estimates discussed at the beginning of this article.  Once again, I don’t think there is a population decline in any of the 122 countries featured in the 2010 data set—but I suspect that countries in the serious or higher hunger categories would have a population decline, but for receiving food aid.

Are these results statistically significant?  How does this data break down according to different regions of the world?  There are a few prominent outlying data points in Figures 1 and 2—what countries are those?  For the countries for which we have data for both time periods—did the hunger index change from 1990 to 2010 in a manner that is consistent with the expected change in per capita consumption? Questions, questions, questions.

Join me next time when I discuss these issues and several other interesting facets of this analysis. 


  1. Hi Crash Watcher, interesting info and lots of good work. If you haven't seen it, have a look at there is a heap of stats that can be plotted over time. its interactive and provides some great food for thought.

  2. your Blog information is so Good.

    The relationship between hunger and petroleum consumption-Part 1

  3. Hi Anonymous, I took a look at and it looks interesting, although the petroluem production/consumption data is a bit dated.

  4. I am not sure where you are going with this analysis but on order to construct a relationship between GHI and b/py (based on aggregate data for different countries) you would eventually need to consider the variance in b/py across the each country’s population. Two countries with the same average b/py might be expected to have much different GHI values depending on the distribution (or availability) of petroleum across its population. Compare, for example, a country with a tight or narrow (low variance) distribution – where most of the population consumes about the same amount of petroleum-based energy – to one that has a uniform or perhaps a bimodal distribution (say 30% with high b/py and 70% with a near-zero rate; most of the 70% might be expected to be at high risk for starvation and infant mortality). I am assuming that your b/py data are being derived by dividing total fuel consumption by the number of people reported to be living in the country. As a result, the validity of your results might improve by eliminating some countries based on suspected differences in distributions. Not an easy problem to solve.

    1. Hi sixtiessman,

      I think that you mean "where WERE you going with this" since this post is over a year ago.

      If you didn't figure it out from the context of the subsequent posts in the series, I was looking for broad evidence of a relationship between per capita petroleum consumption and petroleum-driven food production, and in particular, at what low per cap consumption level such food production systems start to fall apart. My subsequent posts did show significant differences, so I am pretty satisfied with the results.


Your comments, questions and suggestions are welcome! However, comments with cursing or ad hominem attacks will be removed.