Saturday, March 22, 2008

GDP Growth and Missing Energy

Reacting to my observation on the negative correlation between Oil Imports and GDP per capita, commenter UPn Student suggested that:
"you should chart out BTU-consumption versus GDP, not oil-importation versus GDP."
I think that suggestion made sense so i did some research. I divided the Final Energy Consumption* figures in my preceding post by the applicable population for a given year and mapped the result to GDP per capita** as follows:

Figure 1: GDP per capita and Energy Consumption per Capita
Click on image to enlarge

I then computed the correlation coefficients between per capita GDP and Energy Consumption during the three periods of expansion under Cory, FVR and GMA as well as the two periods of contraction that are sandwiched in between as shown in the diagram below.

Figure 2: Correlation Coefficients: GDP per Capita and Energy Consumption per Capita
Click on image to enlarge

For the periods of GDP expansion (1986 to 1989, 1993 to 1997 and 2002 to 2005), the above show similar patterns of correlation between per capita GDP and Energy Consumption as my previous computation which was based on Oil Imports. (As can be seen, for the three periods of expansion, the values of the correlation coefficients are practically identical.)

The first two post-EDSA expansions under Cory and FVR shows the expected positive correlation between per capita GDP and per capita Energy Consumption. By contrast, the current expansion under Gloria Arroyo shows a negative correlation. Even after factoring in the contributions of Malampaya's Natural Gas starting in 2002, the strangeness remains.

The next step would be to identify the 'missing' component and match that to the sector of the economy that seems to be consuming less energy than it normally does. Among the three components of Final Energy Consumption for the period 2002 to 2005, both Coal and Electricity has shown a positive correlation. In contrast, Petroleum has shown an almost perfect negative correlation:

Energy SourceCorrelation Coefficient
Petroleum- 0.99
Table 1: Correlation Coefficients: GDP per Capita and Energy Consumption by Source (2002 to 2005)

So the missing energy turns out to still be Petroleum (which is derived from Crude Oil).

Which sector of the economy then has been consuming less and less Petroleum per capita as per capita GDP increases? Of the three broad categories of power consuming sectors of the Philippine Economy, i.e. Industry***, Transport**** and Others***** (which includes the Residential and Commercial Sectors), all have shown negative correlations when it comes to Petroleum consumption and GDP per Capita:

Economic SectorCorrelation Coefficient
Industry- 0.84
Transport- 0.88
Others- 1.00
Table 2: Correlation Coefficients: GDP per Capita and Petroleum Consumption by Economic Sector (2002 to 2005)

It may again be argued that Petroleum is not the only source of energy of the above sectors so we need to consider the possibility of substitution of Petroleum by other sources (e.g. Electricity and Coal). If we take into account all those other sources of energy (i.e. Electricity and Coal) used by these sectors into the computation of the Correlation Coefficient, we get the following results:

Economic SectorCorrelation Coefficient
Industry + 0.83
Transport- 0.88
Others- 0.62
Table 3: Correlation Coefficients: GDP per Capita and Energy Consumption by Economic Sector (2002 to 2005)

The above shows that for the Industry Sector, substitution of Petroleum by other sources may offer a plausible explanation for the reduction in its usage. However, for the remaining sectors such as Transport and Others (includes Commercial and Residential), the question of missing energy is still unresolved. The alternative explanations i can think of are the following:
  1. We have become more efficient in consuming energy to the extent that energy consumed per person is actually decreasing even as incomes are increasing. This, however, is highly unlikely as it goes against the pattern of increasing energy consumption that goes with increasing incomes.
  2. Energy consumption is being understated. It may be that the smuggled oil is not being accounted for in the Energy Balance Tables.
  3. GDP growth has been overstated and corresponding per capita GDP has not really increased as much as has been reported. This is more consistent with the results of the 2003 to 2006 Family Income and Expenditure Survey (FIES).
I would welcome inputs and alternative explanations from the readers.

Update 7:00pm: In the comments section, Gabby asks how i determined whether a given sector substitutes between energy sources. For each sector, i computed the correlation coefficient using Petroleum Consumption (in the green cells) and did the same with Total Energy Consumption (in the yellow cells) as follows:

YearIndustry Sector
Correlation Coefficient-0.84+0.83
Table 4: Final Energy Consumption (in ktoe)
Industry Sector (2002 to 2005)

As seen above, with Petroleum alone, the Correlation Coefficient was negative (at -0.84) but if we include other energy sources i.e. Coal and Electricity, the value turns positive (at +0.83). This means that the increase in the alternative Energy Sources more than compensated for the relative decrease in Petroleum. The same cannot be seen for the rest of the sectors as shown in the following tables where the coefficients have remained in negative territory:

YearTransport Sector
Correlation Coefficient-0.88-0.88
Table 5: Final Energy Consumption (in ktoe)
Transport Sector (2002 to 2005)

YearOthers Sector
Correlation Coefficient-1.00-0.62
Table 6: Final Energy Consumption (in ktoe)
Others Sector (2002 to 2005)

Update Nov-16-2008: Economist-blogger Tyler Cowen of Marginalrevolution, takes a similar over-all approach to the above and applies it to China. He speculates that the year-on-year change in generated electrical power in China for the month of October may be a more reliable indicator than government-reported GDP statistics.

Update Jan-27-2009: In response to GabbyD's query in the comments section, here is the relative share of the Transportation Sector to total GDP (1990 to 2008).

YearTransportation & Communications Sector GDPTotal GDPRelative Share
Table 7:Relative Share of Transportation & Communications Sector to GDP (1990 to 2008)******

*Source: APEC Energy Database
**Source: NSCB for 1994 to 2006 and Virtual Philippines, Biz Pack, Statistics - GDP and Economic Indicators for 1986 to 1993.
***Industry Sector includes Iron and Steel, Chemical (incl. Petro-Chemical), Non Ferrous Metals, Non Metallic Mineral Products, Transportation Equipment, Machinery, Mining and Quarrying, Food, Beverages and Tobacco, Pulp, Paper and Printing, Wood and Wood Products, Construction, Textiles and Leather.
****Transport Sector includes International Civil Aviation, Domestic Air Transport, Road, Rail, Inland Waterways, Pipeline Transport.
*****Others Sector includes Commerce and Public Services, Residential, Agriculture and Fishing.
******Source: NEDA

Monday, March 17, 2008

GDP per Capita and Energy Consumption

Malampaya Project

In the preceding post, i pointed out the unusual negative correlation between GDP per capita and oil imports during the present growth episode (2001 to 2006). In the comments , fellow blogger Jon Limjap brought up the possibility that this may be due to the increased reliance on Natural Gas leading to a reduction in oil imports. The following chart shows the amount of energy consumed (in kilotons of oil equivalent or 'ktoe') according to power source*:

Figure 1: Final Consumption by Energy Source
Click on image to enlarge

One thing immediately noticeable is the sudden appearance starting in Year 2003 of 'Others'. (Nope, this is not because of Natural Gas from Malampaya.) The bulk of this category consists of biomass in the form of Fuel Wood & Woodwaste which accounts for over 80% of the total with the remainder made up of Bagasse, Charcoal and 'Other Biomass'. As explained in the APEC Energy Demand and Supply Outlook (2006):
In 2002, the Philippines’ residential energy consumption was dominated by biomass (from rice and coconut residues, and wood waste) at 58 percent. However due to the expanded access to electricity and the availability of commercial fuels, the share of biomass in total residential energy demand is projected to decrease from 58 percent in 2002 to 28 percent in 2030. Consequently, limited growth in biomass is expected at a modest rate of 0.2 percent due to the continued use in rural areas as a fuel for cooking.
Clearly, biomass as a source of fuel has already been in use even before 2002, although not accounted for in the Energy Balance Tables prior to that year. For consistency of comparisons, i will therefore disregard this component from the analysis and focus on Coal, Petroleum and Electricity and redraw the above chart as follows:

Figure 2: Final Consumption by Energy Source
(Coal, Petroleum & Electricity)

Click on image to enlarge

Natural Gas (along with Crude Oil, Geothermal, Hydroelectric and others) is used as an input to generate Electricity so any increase in contribution will be reflected in this component. However, it is readily apparent that, as a source of power, Electricity still constitutes a minority and is dwarfed by Petroleum. (The portion of Crude Oil that is used as input to Electricity is not reflected under Petroleum but is incorporated as part of Electricity. This prevents any double counting.)

The relative contributions of the various inputs to Electricity are shown in the diagram below.

Figure 3: Inputs to Electricity
Click on image to enlarge

So the answer to Jon's question above is that, as far as electricity is concerned, Natural Gas started contributing significantly to electricity generation in 2002 and along with coal has indeed replaced petroleum as an input to Electricity. However, it is also clear from the above that Electricity remains a small part of the Country's overall Energy Consumption. So the question remains on whether Energy Consumption has kept pace with GDP per capita growth. This would be the subject of my next blog entry.

*Source: APEC Energy Database

Monday, March 10, 2008

GDP Per Capita and Oil Imports: Strong, Weak and Strange Correlations

In his Manila Standard column, Peter Wallace pointed to the disconnect between GDP growth and oil imports:
"Imports were 6.6 percent less in 2007 than they were in 2006. Now in a healthy, growing economy that’s a most unlikely event. Within that oil imports fell 5.6 percent. Now that’s just impossible." [emphasis mine] "You can have some slowing if there’s a shift to alternative fuels, but in 2007 there wasn’t to any significant degree. Oil imports should be growing close to GDP growth, a bit slower but close, and not showing a contradicting trend as it did in 2007. So you’re left with only one logical alternative: smuggling increased substantially."
I share Wallace's misgivings because common sense leads me to believe that higher income leads to higher energy consumption and hence the need for more oil imports. Over at Manolo's however, fellow commenter Ca T questions the existence of such a correlation and she asks:
"...with regards correlation of oil imports to GDP growth, is there a study which shows there is high correlation? then what is the correlation coefficient..." - The Ca T, March 6th, 2008 at 3:27 am
I haven't come across such a study so i had to do my own research. First, i had to obtain data on the volume of crude oil imports. The chart below shows the volume of crude oil imports (in thousands of barrels) from 1986 to 2006.*

click on image to enlarge

I then divided the above by the applicable population for a given year and mapped the result to GDP per capita** as follows:

click on image to enlarge

The resulting correlation coefficient for the entire 21 year period (1986 to 2006) turns out to be weakly negative with a value of -0.35 (see top portion of following diagram). This means that a rising per capita GDP results in a lower volume of imports over time which is contrary to Wallace's claim as quoted above. This is quite different from the expected relationship and is different as well from the relationships found in the United States that shows a strong positive correlation between gdp per capita and oil consumption (estimated at +0.68).

A closer look at the data reveals a sharp difference in the relationship before and after 1997 (see bottom portion of above diagram). From 1986 to 1997, there is a strong positive correlation (at +0.76) between per capita GDP and oil imports similar to the one observed in the United States. Starting 1998 until 2006, however, the relationship becomes strongly negative (at -0.94). What accounts for this sudden reversal in the correlation after 1997? A further breakdown into periods of expansion and contraction sheds light on the anomaly. The diagram below shows the correlation coefficients between per capita GDP and oil imports during the three periods of expansion under Cory, FVR and GMA as well as the two periods of contraction that are sandwiched in between.

click on image to enlarge

As shown above, the periods of expansion during Cory and FVR's terms showed a strong positive correlation (+0.85 and +0.80 respectively) between per capita GDP and oil imports. Gloria Arroyo's expansion by contrast exhibits a strong negative correlation (at -0.87) between per capita GDP and oil imports. During periods of contraction, the earlier one (in the aftermath of Gringo's coup attempts and during the Power Crisis) shows a weak negative correlation (at -0.21) while the one during the Asian Financial crisis shows a strong negative correlation (at -0.92).

click on image to enlarge

The per year breakdown above explains the strength of the correlations. Starting 1997, the behavior has been almost completely dominated by an increase in GDP being accompanied by a decrease in oil imports. This has resulted in a situation where the high water mark in per capita GDP level during 2006 exists side by side with a per capita oil imports falling below the 1987 level. Strange (not to mention fishy).

*Source: Virtual Philippines Biz Pack, Statistics - Energy and Water Resources for 1986 to 2001 and NSCB for 2001 to 2006.
**Source: NSCB for 1994 to 2006 and Virtual Philippines, Biz Pack, Statistics - GDP and Economic Indicators for 1986 to 1993.