What the history of London’s air pollution can tell us about the future of today’s growing megacities

Our World in Data presents the empirical evidence on global development in entries dedicated to specific topics.
This blog post draws on data and research discussed in our entry on Air Pollution.

Cities in most high income countries have relatively low levels of local air pollution compared to cities in more rapidly developing ’emerging economies’. This, however, hasn’t always been the case.

National air pollution trends often follow the environmental kuznets curve (EKC). The EKC provides a hypothesis of the link between environmental degradation and economic development: in this case, air pollution initially worsens with the onset of industrial growth, but then peaks at a certain stage of economic development and from then on pollution levels begin to decline with increased development. Many high income nations are now at the late stage of this curve, with comparably low pollution levels. Meanwhile, developing nations span various stages of the growth-to-peak phase. I have previously written about this phenomenon in relation to sulphur dioxide (SO2) emissions here on Our World in Data.

If we take a historical look at pollution levels in London, for example, we see this EKC clearly. In the graph below, we have plotted the average levels of suspended particulate matter (SPM) in London’s air from 1700 to 2016. Suspended particulate matter (SPM) refers to fine solid or liquid particles which are suspended in Earth’s atmosphere (such as soot, smoke, dust and pollen). Exposure to SPM (especially very small particles, which can more easily infiltrate the respiratory system) has been strongly linked to negative cardiorespiratory health impacts, and even premature death. As we see, from 1700 on, London experienced a worsening of air pollution decade after decade. Over the course of two centuries the suspended particulate matter in London’s air doubled. But at the very end of the 19th century the concentration reached a peak and then began a steep decline so that today’s levels are almost 40-times lower than at that peak.

The data presented below has been kindly provided by Roger Fouquet, who has studied the topic of environmental quality, energy costs and economic development in great detail.

What explains this worsening and the subsequent improvement of London’s air quality?

The dominant contributor to London’s historic air pollution was coal burning. Throughout the 18th and 19th centuries, the coal industry in Great Britain expanded rapidly; driven not only by economic growth, but also by an expanding labour force and improved distribution networks (such as railways and waterways)., Increasing demand and falling coal prices (prices nearly halved between 1820 and 1850) led to a rapid increase in national coal consumption, rising from 20 million tonnes in 1820 to 160 million tonnes in 1900 (an eight-fold increase).

The decline in air pollution can be attributed to a complex mix of factors, including economic restructuring away from heavy industry, switching energy sources, and increased environmental regulation. There are thought to be three primary developments which led to this decline. Firstly, by the late 1800s, improved connectivity and commuter links allowed London’s population to spread further into surrounding suburban areas, inevitably leading to an overall reduction in population density. Even if such changes did not lead to a reduction in total emissions of pollutants, the dispersal and spreading of these population centres could have had some alleviation impact on concentrations in prime pollution hotspots.

Secondly, the United Kingdom introduced its Public Health Act for London in 1891. Under this new regulation, businesses in London which produced excessive smoke ran the risk of financial penalties if they did not adopt cleaner and more efficient energy practices, such as switching to less polluting (but more expensive) coal sources, and ensuring fires were adequately stoked. This put increasing pressure on businesses to shift towards better and cleaner industry practice.

The third potential source of this decline was a notable shift in heating and cooking sources from coal towards gas. Uptake of gas cookers rose sharply in Great Britain during the 1800 and 1900s. The Gas Light & Coke Company—which was the leading London supplier at the time—noted that in 1892 only 2 percent of residents had a gas cooker. By 1911, this had increased to 69 percent. In terms of air pollution impacts, gas is a much cleaner fuel relative to coal, meaning that such a large shift in heating and cooking sources may have contributed to the declining trend.

It’s difficult to fully capture just how polluted London’s air was throughout the 19th century. Throughout this period, London experienced frequent and severe fogs. Such fogs were often so dense that they halted railway journeys, interrupted general economic activities, and even contributed to London becoming a breeding ground for crime (crime rates rose sharply during these fog periods). London averaged 80 dense fog days per year, with some areas recording up to 180 in 1885.

Not only did air pollution incur a severe economic price, it also resulted in significant health costs. Air pollution deaths throughout this period rose steeply; in London, mortality from bronchitis increased from 25 deaths per 100,000 inhabitants in 1840 to 300 deaths per 100,000 in 1890. At its peak, 1-in-350 people died from bronchitis.

Although London was arguably one of the worst polluted cities during this time (and often referred to as the “Big Smoke”), many other industrial cities across Great Britain (and indeed across other nations) experienced similar air pollution problems. In the photograph below, we see pollution in Widnes, an industrial town close to Liverpool, in the late 19th century.

Air pollution in Widnes, late 19th century

London vs. today’s developing cities

From our first chart, we see that concentrations of suspended particulate matter (SPM) reached up to 623 micrograms per cubic metre. This figure will be meaningless to most without proper context. Let’s therefore compare historic London concentrations to those experienced in recent years in New Delhi—one of the world’s most polluted cities today.

In the first chart above, we can add SPM trends for Delhi from the later 1990’s to 2010 using the ‘add city’ button.

What we see is that concentrations in Delhi range from around 450 to 500 micrograms per cubic metre. This is undoubtedly high, but remains lower than peak concentrations in London during its rapid industrialization. It is wrong to assume that today’s major developing cities—such as Delhi, Beijing, Jakarta, Karachi—are experiencing unprecedented levels of air pollution. It’s likely that many of today’s high income cities have gone through similar periods of high (or higher) pollution levels. Perhaps what differentiates today’s transitioning cities is the population sizes which inhabit them; exposure to such pollution undoubtedly leads to high mortality figures in absolute terms.

If we see air pollution as an unfortunate by-product of economic and industrial development, an appropriate comparison would be based on levels of prosperity, rather than versus time. In the chart below we have plotted the same trends in SPM (on the y-axis) for London and Delhi, but now map these levels relative to gross domestic product (GDP) per capita (on the x-axis). These GDP per capita figures are adjusted for inflation and expressed in international dollars to reflect differences in living costs.

Interestingly, if we observe the evolution of these trends with time, we see that to achieve a given level of GDP per capita, Delhi’s air pollution levels have, and continue to, follow a similar pathway to that of London in the 19th century.

But the often-forgotten history of air pollution in today’s rich countries offers an important lesson about what is possible for world regions with lower levels of prosperity today. After air pollution worsens at the initial stages of development it declines at later stages and can reach historically low levels.

The key for Delhi and other transitioning cities will therefore be to continue shifting rightwards (increasing GDP per capita), but to try to peak below London’s 19th century pathway. If they can achieve this, then they will have succeeded in developing in a cleaner way than today’s high income cities.

There is a ‘happiness gap’ between East and West Germany

Our World in Data presents the empirical evidence on global development in entries dedicated to specific topics.
This blog post draws on data and research discussed in our entry on Happiness and Life Satisfaction.

In global surveys of happiness and life satisfaction, Germany usually ranks high. However, these national averages mask large inequalities. In the map below we focus on regional inequalities—specifically the gap in life satisfaction between West and East Germany.

The map below plots self-reported life satisfaction in Germany (using the 0-10 Cantril Ladder question), aggregating averages scores at the level of Federal States. The first thing that stands out is a clear divide between the East and the West, along the political division that existed before the reunification of Germany in 1990.
For example, the difference in levels between neighboring Schleswig-Holstein (in West Germany) and Mecklenburg-Vorpommern (in East Germany) are similar to the difference between Sweden and the US – a considerable contrast in self-reported life satisfaction.

Several academic studies have looked more closely at this ‘happiness gap’ in Germany using data from more detailed surveys, such as the German Socio-Economic Panel (e.g. Petrunyk and Pfeifer 2016). These studies provide two main insights:

First, the gap is partly driven by differences in household income and employment. But this is not the only aspect; even after controlling for socioeconomic and demographic differences, the East-West gap remains significant.

And second, the gap has been narrowing in recent years. In fact, the finding that the gap is narrowing is true both for the raw average differences, as well as for the ‘conditional differences’ (i.e. the differences that are estimated after controlling for socioeconomic and demographic characteristics). The following charts shows this.

Trends in life satisfaction for East and West Germany, 1992-2013

The observation that socioeconomic and demographic differences do not fully predict the observed East-West differences in self-reported happiness is related to a broader empirical phenomenon: Culture and history matter for self-reported life satisfaction—and in particular, ex-communist countries tend to have a lower subjective well-being than other countries with comparable levels of economic development. (More on this in our data-entry on Life Satisfaction.)

How to use Our World In Data visualizations in presentations

You can reproduce all interactive visualizations from Our World In Data in presentations and other documents, such as PowerPoint or LaTeX slides. This is possible because all our material is permissively licensed under CC-BY-SA.

Let us explain with a concrete example. Suppose you are a teacher and you want to use the following map on fertility for a lecture on population growth.

At the bottom right of the chart, you will find a little icon showing a hard disk with a downward arrow on top. This is a ‘download’ button.

If you click on it, you will see two options, as shown below. If you click on the first option, labeled “Save as .png”, you will download a snapshot of the chart in the ‘Portable Network Graphics’ format, which is supported by most word-processing and presentation-making programs (i.e. you can use the downloaded file in programs such as LyX, or Microsoft PowerPoint and Word).

If you click on the second option, labeled “Save as .svg” you will download a snapshot of the chart in the ‘Scalable Vector Graphics’ format, which is supported by most graphic design programs (i.e. you can use the downloaded file in programs such as Inkscape, or Adobe Illustrator).

That’s it! The only key point to bear in mind is that you can download different versions of the same chart, depending on what parameters are selected in the interactive visualization at the moment in which the corresponding snapshot is generated.

Here is a concrete example using the same visualization above. The following two static charts were generated by moving the time parameter in the bottom of the interactive chart: The first picture was downloaded when the dial in the interactive visualization was marking 1955, and the second was downloaded when it was marking 2015.

This option can be very helpful if you are interested in country-specific trends. Since our interactive visualizations allow you to switch between line-charts and maps, you can generate snapshots of country-specific trends simply by clicking on a country in the map.

Once again, in the example above, suppose you would like to include a slide in your lecture discussing the experience of South Korea. All you need to do is click South Korea in the map, and then select the ‘download’ option as described above. (Alternatively, you can click on the ‘Chart’ tab at the bottom of the visualization, then select South Korea from the dropdown menu, and finally select ‘download’.)

Recent improvements to interactive scatterplots

The interactive charts you see on Our World In Data are created using our own internal visualization app, OWID-Grapher. My colleague Aibek and I are always working to improve this tool—here is a summary of some of the recent feature additions.

Connected scatterplots showing change over time

In scatterplots where the timeline bar appears underneath, it is now possible to compare different points in time. Try pressing play on this chart to see a progression from the top left to lower right, representing the global development of both education and safer childhoods. You can hover over the lines to get more detailed information, or drag the endpoints of the timeline slider to compare different years.

Something to note about the grapher in general is that it is always possible to share an interactive chart as it currently appears. For example, if you set the timeline to show a particular year or range of years and then use the buttons in the lower right to link to or download the chart, the output will appear accordingly.

Selecting individual countries

It is now possible to toggle the selection of one or more countries in a scatterplot by clicking on them. This allows for charts that emphasize outliers or other places of interest. For example, this chart shows the surprising divergence of the United States from other developed countries in healthcare expenditure and health outcomes. Try selecting another country and pressing the play button to see how the two countries have changed relative to one another.

Highlighting the differences between world regions

The new legend on the right of these charts can be hovered over to view particular regions, or clicked to select them. This chart emphasizes that it is people in African countries who suffer most from the debilitating effects of communicable diseases.

Future work

Since the grapher generates these charts on the fly, if you’re reading this post later on they may appear quite different to how they did at the time of writing. This is one of the advantages of having a big visualization database—it means that as we improve the code, the changes will apply retroactively to any charts authored in the past, keeping them up to date and consistent.

As they are now, there are several areas where I think these scatterplots could still be improved. The interactive selection of countries is not always intuitive, and lacks an effective mobile design. The legend could synchronize itself with the “graying out” behavior. And the timeline year transition speed is sometimes slower than it should be. Aspects to work on in the coming days!

How much will it cost to mitigate climate change?

Our World in Data presents the empirical evidence on global development in entries dedicated to specific topics.
This explainer post draws on data and research discussed in our entry on CO2 and Other GHG Emissions.

United Nations (UN) member states signed the latest climate agreement, The Paris Agreement, in 2015; the agreement set an international target to limit average global warming to two degrees celsius above pre-industrial temperatures. An important question then arises: how do we achieve this, and how much will it cost to keep warming below the two degree target?

So, how do we estimate the economic cost of mitigating climate change? The most common way of measuring—and visualizing—the options and costs of reducing our greenhouse gas (GHG) emissions is to use the so-called ‘abatement cost curve’ (sometimes referred to as the ‘marginal abatement curve’). Here we will explore what these curves are and how we use them. But before proceeding, it’s important to note that these curves can be used at a range of levels to assess the best and most economic options we have for reducing GHG emissions. In particular, they can be constructed at global, regional, national or city levels. As such, a range of different abatement cost curves exist. In this post, we will be looking at a global curve developed by McKinsey & Company. Although we could have looked at a number of different examples, this analysis in particular is one of the most widely-referenced and accepted to date.

Here is a preview of what these abatement costs tell us: If we aggressively pursue all of the low-cost abatement opportunities currently available, the total global economic cost would be €200-350 billion per year by 2030. This is less than one percent of the forecasted global GDP in 2030.

What is an ‘abatement cost curve’?

An abatement cost curve measures two key variables, as shown on McKinsey’s chart below: abatement potential and the cost of abatement.

‘Abatement potential’ is the term we use to describe the magnitude of potential GHG reductions which could be technologically and economically feasible to achieve. We measure this in tonnes (or thousand/million/billion tonnes) of greenhouse gases (which is abbreviated as carbon dioxide equivalents, or CO2e). Note that our measure of CO2e includes all greenhouse gases, not just CO2. So, on the x-axis we have the abatement potential of our range of options for reducing our GHG emissions; here, each bar represents a specific technology or practice. The thicker the bar, the greater its potential for reducing emissions.

On the y-axis we have the abatement cost. This measures the cost of reducing our GHG emissions by one tonne by the year 2030, and in this case is given in € (i.e. € per tonne of CO2e saved). But it’s important to clarify here what we mean by the term ‘cost’. ‘Cost’ refers to the economic impact (which can be a loss or gain) of investing in a new technology rather than continuing with ‘business-as-usual’ technologies or policies. To do this, we first have to assume a ‘baseline’ of what we expect ‘business-as-usual’ policies and investments would be. This is done—for both costs and abatement potential—based on a combination of empirical evidence, energy models, and expert opinion. This can, of course, be challenging to do; the need to make long-term predictions/projections in this case is an important disadvantage to cost-abatement curves. Whilst not perfect, it does provide useful estimates of relative costs and abatement potential.

Let’s use an example to explain these costs. The majority of the world’s private vehicles are currently diesel or gasoline-powered. Our baseline or ‘business-as-usual’ scenario in this case would be that everyone continues driving vehicles run on fossil fuels, and that growth in the car market continues in line with market projections through to 2030. One option we might consider to reduce CO2 is to begin replacing these with electric-powered vehicles. Electric vehicles (at least currently) are typically more expensive than similar gasoline-powered vehicles, so buying an electric vehicle would require a notably higher investment than ‘business-as-usual’. However, with time, the running or operating costs of an electric vehicle may be lower than a traditional car (as a result of efficiency gains and lower cost of electricity relative to liquid fuel), so we will begin to get some economic return on our initial investment. This scenario—in which we need an up-front investment in new technology—typically means that the capital intensity of the opportunity is high; but spread across years of operation, the annual average cost falls.

Our potential mitigation options are ordered from left-to-right in terms of cost (getting progressively more expensive as we move to the right). The higher the bar, the greater the cost. McKinsey’s analysis defines an upper cost threshold of €60 per tonne avoided (i.e. the chart only shows alternatives that are ‘cheaper’ than €60 per tonne). This is the level they consider to be most economically attractive to investors. This figure is somewhat arbitrary, and as we will see later, the report has also estimated potential emissions savings from technologies above €60 per tonne separately.

Global greenhouse gas abatement cost curve

You will see that many options to the left-hand side of the curve have negative costs. Negative costs indicate options which would actually save money. These are typically related to energy efficiency or land management projects which would provide an economic return over the longer-term.

For example, replacing conventional lighting (continuing with this form of lighting would be defined as our ‘business-as-usual’ case) with energy efficient lighting, would more than pay back in economic terms from lower energy bills. Its ‘cost’ is therefore negative because we would actually save money. Globally, if we took advantage of all of our negative cost options, we see that we could cumulatively avoid emissions of 11-12 billion tonnes of CO2e.

At the other end of the scale we have the more expensive options for mitigation; these are typically related to the delivery of a low-carbon energy supply.

Abatement cost curves are used at a range of scales to aid investment decisions on GHG mitigation options. When making these decisions, both the abatement potential (width) and cost (height) of each option is important. For example, if we look at the lowest end of the scale we see that installing energy efficient lighting is the cheapest option (providing a high economic return through efficiency savings). However, its total abatement potential is very limited. If we were looking for an investment which could save a large quantity of GHGs, we might have to select an option which is more expensive, but has a greater abatement potential. The trade-off between cost and potential is an important one.

How low could our emissions go?

If we were to implement all of our options for reducing GHG emissions, how much CO2e would we save?

On the abatement cost curve above we see that cumulatively, the options costing up to €60 per tonne have a total abatement potential of 38 billion tonnes of CO2e per year (this is the total width along the horizontal axis). This total potential has been summarized in the chart below. On the y-axis we have our global GHG emissions in billion tonnes of CO2e, and on the x-axis we see this evolution with time through to 2030.

The top line we see in this chart is our ‘business-as-usual’ (BAU) pathway. This is our projection of how global emissions will increase if we don’t invest in technology and practices necessary to reduce emissions; this pathway is calculated based on expected population and economic growth projections. By 2030, with business-as-usual, we would expect our annual emissions to be 70 billion tonnes CO2e.

Below this line, we see our options for reducing emissions. Our <€60 per tonne abatement opportunities have been decomposed into energy efficiency, low-carbon energy supply and terrestrial carbon (forestry and agriculture) options. Combined, these give us our 38 billion tonnes of emissions avoided.

But we also have two additional opportunities where we could make GHG savings. The first lies in more expensive technical measures, costing  €60-100 per tonne avoided. Such opportunities are not unattainable, however, we might expect the range of cheaper options to have a significantly greater uptake before more expensive technologies are adopted. The second additional opportunity lies in social and behavioral changes which influence energy consumption and consumer choices towards lower-carbon options.

If we include these additional opportunities, our maximum technical abatement potential by 2030 totals 47 billion tonnes of CO2e per year. Our maximum global potential is therefore a 65-70% reduction relative to our current projected pathway.

Global greenhouse gas abatement potential

The total cost of global CO2 mitigation

How much would it cost in total to implement all of these options?

We can calculate this value using our abatement cost curve. The area of each of our bars represents the total cost for each technology or strategy. By adding all of these individual costs, we get the total economic cost. To get the approximated annual cost, we can divide this total by the number of years until our target date (in this case, the year 2030). This method is not strictly accurate, since it assumes that costs will be distributed equally across this timeframe; however, it gives us some relative approximation.

If we utilized all of our <€60 per tonne abatement opportunities to their full potential (which is an important assumption), McKinsey estimates the total global cost to be €200-350 billion per year by 2030. This is less than one percent of the forecasted global GDP in 2030.

This cost is however, made more complex by the timing of investment. This timing component is important for financing and capital investment. Our average annual cost is calculated as a balance of how much money we need to invest, and how much we get in return based on efficiency gains and reduced running costs relative to our current technologies. However, as we noted earlier, we often need an initial capital invesment to implement the technology; this initial capital can be high, but begins to pay back over time (resulting in a reduction in the average annual cost through time). The upfront capital investment needed is €530 billion per year by 2020 and €810 billion by 2030. Although these figures may seem substantial, many estimates project that the economic costs of not taking action to avert climate change would greatly exceed investments in mitigation opportunities.

Technical notes:

It should be noted that the figures presented in cost abatement curves carry significant uncertainty and interdependencies. The total abatement cost curve is developed based on the construction and combination of individual bars related to a particular technology or practice. The width and height (i.e. abatement potential and cost) is measured independently of other technologies and variables on the curve. However, this assumption is a large and important one: many technologies (for example, solar PV and battery storage) have strong interdependencies. The evolution of one will inevitably have some impact on the other.

Technology costing is also strongly dependent on scaling effects–unit costs often decrease as capacity increases. The individual costs represented on these charts are therefore calculated with the assumption that the development and scale-up of the respective technology is approached aggressively and full technical potential is achieved. Changes in technology prices through time are assumed based on technology-specific ‘learning curves’. Learning curves measure the reduction in cost for every doubling in a technology’s capacity. For example, the learning rate for solar PV has been 22% –this means that the cost per unit of power falls by 22% for every doubling of capacity. This analysis assumes that the learning rate for specific technologies remains consistent with past trends.

McKinsey&Company stress that their estimates therefore carry a high level of uncertainty. They are useful in providing an overview outlook on the total abatement potential (and whether this would be sufficient to meet global climate targets) and estimate the potential costs of doing so. Their most useful function is perhaps in allowing for relative comparisons of costs and potential between different technologies; these factors are key for decision-makers looking to investment in low-carbon opportunities. Although imperfect, they give a relative sense of magnitude of what could potentially be achieved and if the costs can be managed to do so.