top of page

A Difference-in-Difference Analysis of Climate Change

Has the 1973 oil crisis saved the planet?


The oil crisis in 1973 occurred due to a political tension, in the aftermath of the war between Arab nations and Israel. The oil embargo was targeted to the nations that supported Israel during the war, which included the economic superpowers such as the U.S., U.K., and Japan. The world price of oil skyrocketed, which had devastating and lasting effects on the world economy that depended heavily on oil. A prolonged period of stagnation followed, where a high rate of inflation is accompanied by a low rate of economic growth: see, for example, this link.


However, there is a claim that the crisis in fact saved the planet: see this article. As a result of a higher price of oil, the world has cut down its consumption, by using less oil and adopting a range of technologies that would result in a higher energy efficiency. Efforts also have been made to find cheaper and cleaner alternatives to oil. As a result, the CO2 emission of the major industrial countries such as the U.S. have permanently changed its course.


In this post, using the difference-in-difference method, I quantify the impact of the 1973 oil crisis on CO2 emission. The U.S. and Saudi Arabia are selected as the representatives of the treatment group and control group respectively. The U.S. is the world’s biggest economy heavily impacted by the oil embargo, while Saudi Arabia were one of the OPEC nations at the other end of the crisis. The annual data set of CO2 emission and oil production (OPEC) from 1960 to 2021 is analyzed. The data, its sources, and R code can be found here.



 

Descriptive Analysis


Figure 1 above presents the time plots of CO2 emissions (in million tons) of the U.S. and Saudi Arabia, along with OPEC oil production (barrels/day). The oil production declines in 1973 and becomes stable until 1980, and then shows a big decline until the mid-1980’s. From then on, oil production shows a steady increase to 2021. The CO2 emission of the U.S. follows a similar pattern. Most strikingly, it was increasing at an annual rate of about 4% from 1960 to 1973, but it shows a negative or a much lower rate of increase from 1974. The annual average growth rate of CO2 emission of the U.S. from 1974 to 2021 is 0.15%. In contrast, the Saudi Arabian CO2 emission has been growing at the average rate of 13.3%, showing only brief periods of negative growth.


 

Difference-in-Difference Analysis

In keeping with the difference-in-difference (DID) analysis, we define the following dummy or indicator variables:


d2 = 1 for years on and after 1973, d2 = 0 for years before 1973;

dT = 1 for the U.S. (Treatment Group), dT = 0 for the Saudi Arabia (Control Group);

DID = d2 × dT: interaction between d2 and dT.


The regression equation for the DID analysis is written as below, where Y is the (pooled) CO2 emissions for the U.S. and Saudi Arabia and X is the oil production (all variables in natural logarithm: see Appendix below for details):


The equation (1) above can be decomposed into different parts, depending on the values of the dummy variables, as below:

  • d2 = 0 (before 1973); dT = 0 (Saudi Arabia)

  • d2 = 1 (after 1973); dT = 0 (Saudi Arabia)

  • d2 = 0 (before 1973); dT = 1 (U.S.)

  • d2 = 1 (after 1973); dT = 1 (U.S.)


The above decompositions can be summarized in the table below:


The last row of Table 1 shows the differences in each group, before and after the crisis. Taking the difference of the differences between the U.S. and Saudi Arabia, we have δ3 + δ4X.

The focus of the DID analysis is the parameter values of δ3 and δ4, which represent the difference in the response of Y to X between U.S. and Saudi Arabia, after the oil crisis in 1973. In particular, the key parameter of interest is δ4, which is the change in the elasticity of U.S. CO2 emission with respect to oil production after the crisis (because Y and X are the data transformed to the natural logarithmic scale: see Appendix below). The elasticity is a unit-free measure of linear association: it measures how much percentage of Y increases or decreases, in response to 1% increase of X.


 

Estimation Results

The equation (1) is estimated using the ordinary least-squares method with the following results:

Table 2 above presents the results for two different sample periods:

  • Using the data to 1989, the elasticity estimated of CO2 emission with respect to oil production is found to be 1.002 for both U.S. and Saudi Arabia before 1973, which means that 1% increase of oil production leads to nearly 1% increase of CO2 emission. After 1973, the elasticity for the U.S. is 0.05, reduced by 0.956 (the estimated value of δ4 is -0.956).

  • Using the whole data set, the elasticity estimate of CO2 emission with respect to oil production is found to be 1.458 for both U.S. and Saudi Arabia before 1973. After 1973, the elasticity for the U.S. is 0.236, reduced by 1.222 (the estimated value of δ4 is -1.222).

  • For the results from both sample periods, the reduction of the U.S. CO2 emission elasticity with respect to oil production is substantial, with a more dramatic effect immediately after the 1973 crisis.

  • The results imply that the U.S. response of CO2 emission to oil production has substantially reduced after 1973, in comparison with that of Saudi Arabia.

  • Both estimation results show strong explanatory power of the model, indicates by the R² values higher than 0.95.

This analysis provides a clear evidence that the 1973 oil shock may have saved the planet for many generations to come, by substantially reducing the CO2 emission for the country like the U.S.. One can also easily visualize the projection of the U.S. CO2 emission after 1973 in Figure 1, if it continued increasing at the annual average rate of 4% or more.


 

Conclusion


Has the 1973 oil crisis saved the planet? The answer may well be YES!


In this post, I have adopted the difference-in-difference analysis to measure the impact of 1973 oil shock on CO2 emission. Consistent with the conjecture made by this article, the estimation results provide clear and strong evidence that the crisis has substantially reduced the U.S. CO2 emission in comparison with that of Saudi Arabia.


The world may have suffered economically with a period of high inflation and a large-scale unemployment, but that may have been compensated by a cleaner and more liveable planet that could otherwise have been a disaster of a much bigger proportion.


 

Let Y = log(C) and X = log(O), where log is natural logarithmic function and C and O are the variables of interest, such as CO2 emission and oil production. The slope coefficient of the regression of Y against X measures ΔYX (change of Y in its units, in response to a unit change in X), where ΔX denotes the change of X.


Then, it can be shown that


ΔY = ΔC/C follows from the property of the natural logarithmic function, where ΔC/C is the proportional change of C or % change of C (when multiplied by 100). Hence, η is by definition the elasticity of C with respect to O, representing the percentage change of C in response to the 1% change of O.


Comments


bottom of page