To tackle major challenges – from today’s cost of living crisis to enduring inequalities – data are essential. But it is important to bear in mind that the data we select – their strengths, limitations and timeliness – will influence the policy choices that we collectively make.
Newsletter from 10 June 2022
‘Photographs are a way of imprisoning reality’, wrote Susan Sontag in her 1977 collection of essays, On Photography. Photos capture a moment in time, presenting a portion of the world trimmed neatly into a single frame. The information omitted from the image vests great power in the artist behind the lens.
What does this observation imply for economists? Working with data, like taking pictures, requires us to make choices and accept limitations. Just as a photographer has to decide what is in the frame and what is not, data scientists, economists and policy-makers must choose what questions to ask and which data to collect, analyse and present.
Like photos, data become less relevant to the present day from the moment they are collected. Every day, thousands of people die and thousands more are born. As a result, the national census is out of date when it is only a day old.
Recognising the strengths and limitations of data was the subject of a talk by Hannah Fry (University College London) at Bristol Data Week. Hannah’s talk – hosted by the University of Bristol’s Jean Golding Institute – explored a collection of her favourite data stories, focusing particularly on cases of where things were not as they first appeared.
One example was the 1989 Challenger space shuttle mission, a disastrous episode for the National Aeronautics and Space Administration (NASA). The story goes that the American space agency was deciding whether to launch on a scheduled date. Conditions were set to be very cold, at around 40°F (about 4°C). Thinking this could be risky, the team’s engineers presented their superiors with the spacecraft’s performance data from various launch tests, using a chart that plotted system failures against air temperature.
During the tests, the shuttle had failed seven out of 24 times, at varying temperatures. To illustrate this, the engineers plotted the number of incidents on the y-axis against the air temperature on the x-axis. The points showed no clear trend beyond a vague u-shape, with failures seeming unrelated to weather conditions (see Figure 1).
Figure 1: Number of incidents of damage by temperature
Source: Fry, 2021
The decision went to a vote and almost everyone on the NASA team chose to go ahead with the launch. No one looked at the chart and asked what had happened to the 17 missing data points – the results from the tests that did not end in system failure.
Figure 2 shows the updated chart. As soon as the missing data (highlighted in blue) are added, the picture becomes clear. Flying in cold temperatures is extremely risky compared with temperatures above 65°F (around 18°C). In fact, the shuttle had never taken off without the systems failing at temperatures below this point.
Figure 2: Number of incidents of damage by temperature (all data)
Source: Fry, 2021
Launched in dangerously cold conditions, the Challenger’s O-rings leaked, causing a joint in the solid rocket boosters to fail. As a result, less than two minutes after taking off, the shuttle broke apart, killing all seven crew members.
This example, first used by Edward Tufte (Yale University and author of The Visual Display of Quantitative Information), clearly illustrates the similarity between data science and photography. The NASA team’s decision rested on flawed data, caused by a failure to recognise the limitations of presenting information in a particular way.
By only focusing on the system failures and not looking at the successful tests, the team were blind to the reality of the situation. It was what was out of frame that mattered for the safety of the launch.
One solution to these dangerous blind spots is getting people to cross-reference and verify the data in question. This requires work to be open source and replicable. Our Data Hub – as well as the charts used in all of our articles – is built on this philosophy. Anyone can download and check the numbers, view the code used to construct the chart and find links to the original data source.
Wealth and health
On Monday, we posted a piece by Ricky Kanabar (University of Bath) on the level of wealth inequality in the UK and, in particular, the causes of the UK’s racial wealth gap. Citing data from the Office for National Statistics (ONS), Ricky shows that at £243,700, the average level of total household net wealth holdings among Pakistani and Bangladeshi households is £201,500 lower than that of white British households.
One of the main drivers of differing wealth in the UK is home ownership. The article explains that according to survey data, 74% of British Indians report owning their home (either outright or with a mortgage). This is 11% higher than the average level across all ethnic groups. The equivalent figure stands at only 20% and 40%, respectively, among the black African and black Caribbean groups.
Looking at social housing, the pattern is reversed. Only 7% of Indians report living in social housing, compared with 44% and 40% for black African and black Caribbean groups, respectively.
On Thursday, we posted a follow-up piece by the same author, looking beyond the underlying causes of the racial wealth gap and focusing on the impact of Covid-19. In this article, Ricky explores how the pandemic brought the vast differences in wealth among different groups in the UK into sharp focus.
Figure 3: Total net household wealth by ethnic group relative to the white British group
Source: ONS, 2022
Note: Estimated difference in total net household wealth after controlling for age, sex, education level, socio-economic classification of the head of household, housing tenure and household composition; April 2018-March 2020 prices.
In particular, the surge in house prices caused by the pandemic has meant that groups with a higher level of home ownership have seen their wealth swell, while those without bricks and mortar assets fell further behind. Others have argued that quantitative easing has had the same effect.
On top of this, ethnic minorities were more likely to work in sectors most affected by lockdowns, such as shops and restaurants. This meant that they were at greater risk of seeing their incomes fall due either to being furloughed or having to work fewer hours. And due to the nature of their work, people belonging to these groups also faced higher risk of infection, and thus hospitalisation or death.
Given the composition of ethnic minority households before the pandemic, it is likely that differences in household wealth have widened during the crisis, Ricky concludes. This is partly a result of the way that the crisis has affected both labour and financial markets, with jobs in contact-intensive areas being slashed, while the values of houses and other assets have soared.
The crisis has revealed structural weaknesses and inequalities that were present long before the first coronavirus case was detected. As the worst of the pandemic subsides and the country faces new economic challenges like the cost of living crisis, this gulf in wealth is likely to play a part in who is hit hardest by rising prices.
