Data Analysis: Emission share by Industry
This is a draft in progress. I will update periodically as I have time.
Follow me on a data analysis journey! I create one chart, puzzle at it, and then create a different chart that perhaps represents what I’m after a bit more closely.
Profit vs Emissions
Here’s a clustered column chart I created of share of profits vs total emissions (Scope 1–3). Data from Robert Höglund of Marginal Carbon. I originally chose stacked columns so the relationship between the two shares would be clear, but Joshua Lave suggested I try side-by-side (Thanks Joshua!). I’ve opted to go with side-by-side only for now. Analysis to follow.
Analysis
The charts show that some industries have an outsized profit share versus emissions while others are more balanced. Chemicals, Construction, Oil & Gas, and Utilities stand out in particular as producing a far greater share of economy-wide profits versus their share of emissions. Other industries like Materials, Conglomerates, and Retailing produce more emissions versus profits. A third group is quite balanced: Consumer Durables, Food, Drink, & Tobacco, and various smaller categories like Tech Hardware.
Why choose to compare profit share versus share of emissions? What’s to be gained from such a comparison? Profits, we might think, are a stand in for surplus value produced. Some industries are producing a lot of surplus value versus the emissions they are producing. In some cases, this may be due to inelastic demand — think Utilities or fuel, for instance. Whereas for others, it may be due to an intrinsic value-added nature of the industry — think Construction and Chemicals, which aren’t responsible for a large share of the emissions, but produce an enormous amount of value for the economy.
A few surprises:
· I would have thought transportation would be a thinner profit margin versus emissions. Their product is almost entirely emissions-generating and their profit margins are not high.
· Similarly with Oil & Gas: they make more money per emissions than I would have thought. My guess is that this is due to inelastic demand and other market-skewing forces.
· Utilities have the lion’s share of profits. I wouldn’t have guessed this initially. This is something worth investigation further in the data. Their emissions numbers aren’t surprising, but why do they have such a high share of profits?
· Why are media and banking so low when it comes to profits? That seems like a lacuna in the data. Something worth looking into further in the data.
There are some methodological questions that I didn’t address in making the graph. Isn’t materials part of the emissions for, say, Construction? The fact that Materials produces the lion’s share of emissions but Construction is depicted as producing relatively little may be inaccurate or even unfair. This is a complex question worth delving into in the future.
There are some questions I have about these data. Höglund has done great work generating the raw data, to be clear, but I need to look further at the raw data to get a clear sense of what we’re dealing with. This will help us answer the questions above.
Revenue vs Emissions
Here’s a chart from the same dataset comparing Scope 1–3 emissions with revenue instead of profit. Again both are shares of the totals. Some interesting differences here.
Analysis
Some key differences to note here:
· Whereas with profits, Chemicals, Construction, Oil & Gas, and Utilities showed a much higher share of profits vs emissions; when we compare revenue instead, chemicals now shows a much higher share of emissions vs revenue. The chemical industry is producing relatively less raw value vs the amount of emissions they are producing. But they have relatively higher profit margins, so they show an outsized profit share but a lower revenue share. Construction shows the same basic relationship as profits, but the values are closer. Oil & Gas look almost identical.
· Utilities are now closer to what I would’ve pretheoretically expected. A large portion of utility production and distribution produces emissions, so it stands to reason that they would have an outsized effect on emissions while producing a relatively cheap product that is a lower share of overall economic value.
· Materials is now the most dramatic standout on one side of the distribution. Let’s sort according to emissions:
We can see that Materials is the standout on the right side and has dramatically lower revenue vs other high emissions industries. Conglomerates and Utilities are twice the revenue and half or less of the emissions of materials. Consumer Durables has a comparable amount of emissions, but far higher revenue (a factor of ~7.5x).
As I mentioned earlier as a methodological concern: materials is bearing the brunt of the entire material economy’s emissions. It’s unclear how to disentangle the emissions of the construction industry and the consumer durable industry from the materials industry. Something to keep in mind. The method here was to go company-by-company and categorize according to industry (the categorizations were Höglund’s, I believe).
Let’s Sort the other way: according to revenue share:
Banking stands out as high revenue low emissions. Does this include the energy cost of crypto? Algorithmic trading? Unclear. I suppose those likely go under “Diversified Financials”. We’d have to dive deeper into the data collection methods to see what is happening here.
Insurance and Retailing raise no red flags for me. High profit, low emissions. Makes sense. Retailing, though, raises one question: does this include the emissions from freight and delivery? It should be included in scope 1–3, I would think.
Consumer Durables make sense, as well, a high profit and higher emission industry. IT Software and Services sees a somewhat surprising bump in emissions, one supposes from computing and a bit from traveling on-site IT repair folks.
It’s interesting to note that Hotels, Restaurants, and Leisure produces little revenue and has a higher share of emissions. Not terribly surprising. One wonders when the data covers in terms of years and if any of those numbers are skewed by the Pandemic. Again, a deeper dive into the data and collection methods is needed.
One interesting methodological note: when we sort one category, the other category’s outliers come into starker relief. Consumer Durables, Utilities, Conglomerates, Materials, and Chemicals stand out clearly. As do categories with far less emissions than revenue. Additionally, the “balanced” industries like Construction, Food, Transportation, etc. are clear in this sorted chart.
Conclusions
In retrospect, it’s a bit obvious, but my first lesson here is that revenue gives a better objective look at overall economic value creation than does profit. Share of overall profits is skewed in favor of high profit margin industries, but doesn’t reveal how much economic activity is produced by an industry. Revenue is a much better metric for this. As a result of the switch, the revenue vs emissions chart matches some expectations better.
A caution, though, is in order: economic value isn’t the only sort of value. Energy is enormously important to human life — so much so that its value can hardly be understated. But it is also relatively cheap. Sure, its production and distribution generates relatively little monetary value and so drives relatively little economic activity considered in itself. But having electricity available to us is absolutely essential to our everyday lives.
I’m left currently with a lot of methodological questions that are worth diving deeper into. In spite of these questions, though, we see some datapoints that don’t surprise us and others that do.
I’m left with an evaluative question: *should* we see balanced categories like construction and food? Is there something “fitting” or “right” about an industry producing revenues in proportion to their emissions? One supposes not, right? There is something perfectly sensible in the idea that banking would produce revenue (note: profits is a different question entirely) while producing very little emissions. They produce value without combusting fuels and expanding into carbon-sinks. That’s a sort of ideal, right? We want people producing value while producing very few emissions. Based on these quick reflections, I’m tempted to conclude that there is no moral or otherwise value in having a match between emissions and revenues. Materials isn’t an inherently “bad” industry because its emissions are much higher than its revenues. That is a result of commodity pricing and an inherently high-impact industry. Materials should, one hopes, lessen its emissions and fast, but that’s not because it isn’t making enough revenue to account for its emissions. That’s just because emissions are bad.
Continuing On the Journey…
Now I’m playing around with Tableau to learn its ins and outs. Here’s a chart of share of revenue vs share of scope 1–3 emissions by country instead of industry:
Some definite surprises here, which makes me think the data might be limited. China is the highest emitter currently (not in aggregate across history), but has very little emissions shown here. India has almost no emissions share according to this, but has 6.8% of emissions according to Our World in Data. Something doesn’t add up. Digging further into the data, I see that the data is individuated by company first and then those companies are sorted by geography. The data just don’t contain Indian companies. Saudi Arabia shows no emissions! A lacuna in the data that would need to be addressed to give a better overall picture of emissions and revenue by company.
None of this is to complain about the work Höglund has done, but the lesson for me here is to use simple charts to test the validity of your data. If it has surprising results, sometimes that’s a result of there being surprises in the data. But sometimes, that’s a result of incomplete data or data that needs to be recategorized. Some results are obviously not accurate descriptions of the world, so we can be assured that the data is incomplete or inaccurate. It’s up to the data analyst to review and update data as necessary and to ask good methodological questions as you go about how the data was reported, aggregated, and categorized. A lot of decisions can happen before you get your hands on the data.
These measures are, after all, hopelessly tangled up together so that it is difficult to tease them apart. Does Apple’s emissions include Foxconn’s? Isn’t Foxconn primarily a Chinese company? It’s a genuinely open question about how to parse this and where to place the responsibility for emissions that happen in China, but at the behest of a multi-national-but-primarily-American company like Apple. I won’t try to answer this question here. Big takeway: we’re all in this together and there’s no easy blame shifting to do because the data isn’t easily disentangled.