A Lockdown Birthday Review: Changes in UK Wellbeing over the Pandemic

One year ago, headlines warned about the possibility of lockdown: “Life on Hold for Three Months” gasped one. A bucketful of hindsight later we now know what wishful thinking that was, having slogged through a year where the pandemic transformed our daily lives beyond recognition. This lockdown birthday serves as a useful marker to reflect on what we’ve been through. For one, it’s important to recognize that some people have inevitably been hit harder by the pandemic than others. Just as there are considerable differences in mortality rates, so will there be differences in wellbeing. Increased economic uncertainty was never going to worry the wealthy as much as the poor, and restricted social contact wasn’t going to make shacked-up couples feel as isolated as singletons. So how did the pandemic actually impact happiness in the UK?

While the media has tried to shine a light on the challenges felt by specific social groups, this has been based on anecdotes and speculation, leaving reports about the pandemic feeling paradoxical at times. We heard both about the terrible social isolation, but also the remarkable sense of solidarity and community; from those yearning for a hug to those discovering their neighbours had names on the new Whatsapp street group for the first time. So empirical questions remain: How did people’s sense of loneliness actually change during lockdown? Similarly, the enormous financial instability driven by the deepest recession in living memory was contrasted with bulging personal bank accounts as people’s spending on travel, dining-out and commuting plummeted. So again, how did people’s feelings of financial wellbeing change over the year? In this post I attempt to answer some of these questions visually, presenting statistical models describing how wellbeing (across different dimensions) has varied over time and across different social groups.

Method
I investigate these questions using high-quality data from Understanding Society (UK Household Longitudinal Study), which has surveyed the same group of people continuously throughout the pandemic. The number of people in each analysis I present varies, but most are based on ~110,000 survey responses from ~33,000 people. The models are cubic growth curve models, and the code used to generate the models are provided here. All analyses control for basic demographics and regions of the UK.

We are able to observe seven time-points. The first is from a survey wave conducted before the pandemic (time 0) which can be considered a baseline comparison, mostly from 2019. Subsequent waves were in 2020, in April (1), May (2), June (3), July (4), September (5) and then November (6). Data for January 2021 will be released soon and I will update the analyses then, it will be interesting to follow these patterns into lockdown 3. For those who need a refresh, a reminder of when the UK was in lockdown can be found here.

I consider the impact of lockdown on three measures of wellbeing: general subjective wellbeing (i.e., happiness), financial wellbeing, and loneliness. General wellbeing is measured by combining 12 questions from the General Health Questionnaire (GHQ). Including questions such as “Have you recently been feeling reasonably happy, all things considered?” and “Have you recently been feeling unhappy or depressed?”. Financial wellbeing was measured by the question: “How well would you say you yourself are managing financially these days? Would you say you are...Living comfortably; Doing alright; Just about getting by; Finding it quite difficult; Finding it very difficult.” And loneliness was measured with a question “In the last 4 weeks, how often did you feel lonely?” Answers were: Hardly ever or never; Some of the time; Often. The measures are imperfect but they are some of the few measured repeatedly throughout the pandemic. To improve comparability across the three measures, in all analyses wellbeing scores are standardized so that the average is zero and the units are in standard deviations.

Question 1: How did wellbeing change over the year?

The first question is also the simplest. How did wellbeing change over the course of the year?

• General wellbeing (happiness) followed the rhythm of the lockdowns. It declined during lockdown 1, rose when restrictions lifted and declined again during lockdown 2.

• To ensure this is accurate, we can also compare this against alternative measures in the survey. In particular a question about their satisfaction with their life overall, where we see a very similar pattern of change using this alternative measure of wellbeing.

• Perhaps surprisingly to some, we can see that financial wellbeing increased (on average) during the first lockdown. More people said they were living financially comfortable lives over this period than either before or after. This is likely due to the reduction in spending that took place during these months. However, financial wellbeing returned to pre-lockdown levels by the end of the year.

• Paradoxically, people reported feeling less lonely over the early part of the pandemic. But this increased during the second lockdown towards the end of the year, rising higher than pre-pandemic levels. This is also likely to be driven by the weather, which allowed for more in-person socialising during the warmer summer months.

Question 2: Did younger or older people take a bigger hit to their wellbeing?

• We can see younger people see a bigger change to their happiness during periods of lockdown, we see the same pattern in their financial wellbeing and sense of loneliness. It is worth highlighting that while there has been a wide focus on the loneliness of older people, those aged 20 experienced considerably greater loneliness (both during and before) the pandemic than those of older ages.

• Younger people appeal to experience the greatest uplift in financial wellbeing during the first lockdown, and this may reflect the greater share of their income spent on goods and services that were closed during the first lockdown (e.g., shops, entertainment, dining out).

• The models presented below treat age continuously. I am just choosing to plot the estimated level of wellbeing for people aged 20, 40 and 70 to provide a comparison.

Question 3: Did women suffer more than men?

• Women appear to show stronger responses to periods of lockdown in their wellbeing compared to men. Falling during times of restrictions and rising again when the restrictions are lifted.

• Loneliness in particular shows a different pattern over the year than men. While men seemed to experience slightly less loneliness during the first lockdown compared to their pre-pandemic state, women felt greater loneliness. Both groups experienced an uptick in loneliness during the second lockdown in November.

Question 4: Did any gender differences vary by age group?

• Looking at gender differences or age differences alone may obscure important differences in how gender and life stage lead to differential responses to the pandemic. In this analysis, rather than treating age as continuous, the analysis creates three equally sized buckets of the population: those aged 15-44, 45-60 and 61+.

• Younger women see by far the largest change over the year, following the pattern of lockdown., particularly in terms of their general wellbeing (happiness) and their feelings of loneliness. We again see men experiencing a neutral or even positive response to the lockdown period compared to women.

Question 5: Did single people suffer more than couples? [Part 1]

• These models split the sample into those living with a partner and those not living with a partner. I present an analysis after this looking at the same question with household size.

• While general wellbeing is not statistically distinguishable pre-pandemic (time 0), differences grow after the onset of the pandemic. However, both groups seem to benefit and suffer to similar degrees during periods of lockdowns.

• Those not living with a partner feel considerably greater loneliness (including before the pandemic started). Though they appear to feel less lonely over the course of the pandemic. Both groups feel lonelier during the second lockdown.

Question 5: Did single people suffer more than couples? [Part 2]

• Living with a partner is not the only household formation that is likely to protect against loneliness. We repeat the same analyses looking at household size.

• We again see that those living alone experience the lowest general wellbeing, but they follow the same pattern as other types of household over the year.

• Those living alone experienced much greater loneliness, and this seemed to change more as lockdown was imposed or lifted compared to households of other sizes.

Question 6: Were some Regions of the UK more Impacted than Others?

• Finally, we looked at how wellbeing changed by region. This time we are using an alternative measure of general wellbeing (satisfaction with life), as it shows an interesting pattern for London; which was similar in wellbeing to other regions before the pandemic, but felt less happy over the course of the year. Perhaps this is due to greater overcrowding, smaller housing units and people not having access to gardens.

• We also see that while Londoners on average had lower financial wellbeing, likely due to the comparably higher cost of living, changes in financial wellbeing followed a similar pattern across regions over time.

Conclusion

• The pandemic has not impacted everyone equally and so it’s important to try and understand the wellbeing effects on social groups using data rather than only anecdotes and stories.

• The results are simply descriptions of correlations over time. It is not possible to make any claims that the lockdown itself caused any of these changes in wellbeing.

• Younger women seem to show the most dramatic declines in wellbeing under lockdown, especially the impact on their feelings of social connection. Given they are a group who have an extremely low risk of suffering health effects from the virus, their contribution to keeping the virus under control should be recognised.

• As the world continues to battle the virus, more lockdowns are likely to be imposed. And findings like these might help to allocate resources or estimate the impact across social groups.

• Interestingly, not all patterns fit a simplistic narrative of the year. While general happiness declined, feelings of financial wellbeing and social connection rose during the lockdown.

Is a (money) problem shared, a problem halved?

Communication about money is a social, cultural and psychological taboo. But severe consequences can result when people refrain from discussing money. People need to talk about their finances in order to relieve stress and access advice. ‘A problem shared is a problem halved' is a proverbial saying expressing the idea that when experiencing difficulties, it is beneficial to talk to others about them. I test this hypothesis in the context of money problems. Using survey data from the UK provided by the FCA/Money Advice Service. I use two cross-sectional surveys, one from 2015 and one from 2018, to understand the relationship between financial distress, talking about money and subjective wellbeing.

2015 FCA Data

Wellbeing
Overall, how satisfied are you with your life nowadays?
[0 = Not at all satisfied, 10 = Completely satisfied]

Financial Distress
How satisfied are you with your overall financial circumstances?
[0 = Not at all satisfied, 10 = Completely satisfied]
To what extent do you feel that keeping up with your bills and credit commit
[1 = It is not a burden at all, 3 = It is a heavy burden]

Money Talk
Do you discuss your household finances openly with any of the following people?
1. Friends
2. My partner/spouse
3. Parents/Family
4. My children
5. My colleagues
6. Members of my local community
7. Other, please specify
8. I prefer not to talk about my finances with any of these people

We can plot the marginal effect of talking about money on swl across financial distress.

2018 FCA Data

Wellbeing
Overall, how satisfied are you with your life nowadays?
[0 = Not at all satisfied, 10 = Completely satisfied]

Financial Distress
How satisfied are you with your overall financial circumstances?
[0 = Not at all satisfied, 10 = Completely satisfied]
To what extent do you feel that keeping up with your bills and credit commit
[1 = It is not a burden at all, 3 = It is a heavy burden]

Money Talk
Talk openly about household finances with…partner/ spouse?
[1 = Strongly Disagree, 5 = Strongly Agree]
Talk openly about household finances with…other family members/ friend?
[1 = Strongly Disagree, 5 = Strongly Agree]

Conclusion

• When people are not financially distressed (e.g. they have high financial wellbeing) talking about money doesn't have an impact on their overall wellbeing

• However, when people are suffering financially, then talking about money makes a significant difference to their overall happiness.

• Thus, after controlling for demographics and financial variables, our results are consistent with the idea that talking about money could improve the happiness of those suffering from financial diistress.

But are your interactions actually linear?

The default approach in social science research is to use models which assume a linear relationship between one’s predictor variables and dependent variable. This is due in part to the attractive qualities of OLS linear regression as a modelling technique; particularly its ease of use and interpretability.

Now when it comes to looking at interactions between our independent variables, OLS continues to be the standard approach used. If the relationship between our IV and DV is indeed perfectly linear, then OLS remains a sensible approach. But what about when the relationship is not linear? And how common are truly linear relationships in social science research, anyway?

Nonparametric regression

Nonparametric regression is an alternative modelling approach which is helpful when we are uncertain about the shape of the relationship between our IVs and the DV. When we estimate a standard OLS linear regression, we assume that the functional form for the mean of the outcome is a linear combination of the specified covariates. Nonparametric regression is similar to a standard OLS, but it makes no such assumption.

Instead, nonparametric regression gives us the correct relationship between the outcome and the covariates regardless of their true shape. Using a nonparametric regression is a safe alternative to use in almost all circumstances, because even if the relationship between our IV and DV is truly linear, nonparametric regression results are still consistent (just less efficient…and slow). A further advantage is that we can use the same approach regardless of whether our DV is continuous, ordinal, count or any other level of measurement.

Although nonparametric regression is a way to obtain estimates that are robust to mistakes in specifying the functional form, this robustness comes at a cost. You need many observations and much more time to compute the estimates. Due to the computational intensity, the time it takes to run increases with the number of covariates (see the ‘curse of dimensionality’). So this approach is most useful when you are interested in the relationship between a small number of variables. As computing power increases, and social scientists increasingly run analyses on cloud computing rather than local machines, then the attractiveness of this approach is also likely to increase over the coming years.

Example: Number of Lifetime Sexual Partners

To demonstrate the difference between standard OLS regression interaction effects compared to nonparametric regression, we use data from the National Survey of Sexual Attitudes and Lifestyles (access via UK Data Archive: https://beta.ukdataservice.ac.uk/datacatalogue/series/series?id=2000036).

One of the questions in the survey is the number of sexual partners respondents have had in their lifetime. We might reasonably expect that one important predictor of lifetime sexual partners is a persons age; generally older people will have had more than younger people. We might also expect that the degree to which a person is religious will also be correlated with their number of sexual partners. We might also expect there to be an interaction between these two variables. [FYI: This is me trying to find a fun example, rather than to test a deeply held theory].

In the code below, I open the dataset from the download link above. For simplicity, we will be looking only at males in the survey. I next create a variable for age (in years) and a variable for religious conviction. This is the question: “Importance of Religion and Religious Beliefs Now” (1 = Not Important at All; 4 = Very Important).

Below, we can see the distribution of the total number of partners. We winsorize the variable below to better illustrate the distribution, while analyzing the raw variable in the analyses to come. The number of sexual partners ranges from 0 to 3306 in the raw data, while the top 1% of responses have been constrained to 150 in the below graph.

Next, we can see religious importance broken down. With the majority of people saying that religion is not important to them at all (for international readers, the UK is a very non-religious place).

Finally, before we run any models, let’s plot the raw data to see what this looks like. As the below scatter plot shows, there are some extreme outliers, which are likely to have an influential role on our models. In particular, those who report very large numbers of sexual partners (i.e. 1000+) also typically report that religion is not important to them at all.

When we repeat this using a winsorized version of sexual partners (like the histogram above) to limit the extreme outliers, this provides more information on the distribution of responses but this is very messy and difficult to make-out any discernible pattern.

OLS Linear Regression.

We run a boring old OLS linear regression, and include an interaction between age and religious importance. We treat religious importance as a factor variable (basically treating religiosity as four groups), and age as a continuous variable. The interaction results suggest there is a significant difference between some of the religiosity groups in their number of sexual partners across the age spectrum.

To understand the shape of this interaction, we now plot the OLS regression across levels of age and religious importance.

We use the margins command in stata to plot these results.

margins, at(age=(16(5)71) religion_not_important = (1 2 3 4))
marginsplot, recast(line) recastci(rarea) graphregion(color(white))

The results tell a fairly clear story: those who claim religion is not important to them at all report a high number of lifetime sexual partners. This relationship is strongest the older a person is.

Negative Binomial Regression

Now you might be thinking - “Hey there! Number of sexual partners is a count variable, so employing an OLS model it is a poor comparison against this fancy nonparametric regression. Instead we should use a model designed to deal with count variables.” And you’d be right. Next let’s use a negative binomial regression (similar to a Poisson regression model, but for over-dispersed count data).

When we plot the expected number of sexual partners from the negative binomial regression, we see a similar pattern to that found in the OLS regression, except now the relationship between the religious and sexual partners is non-linear.

Nonparametric Regression

We next model the data as a nonparametric series regression. There are two options here: ‘npregress series’ or ‘npregress kernel’. Stata does a good job of explaining the differences between these approaches in an accessible way here, so I’m going to leave a deeper description of this to another time.

What is clear from the above code and output though is the simplicity of the command. It is simply:

npregress series DV IV1 IV2

Notice that including the interaction in the model is not necessary in nonparametric regression models.

We next use the margins command and marginsplot, just as we would in a normal regression model:

margins religion_not_important, at(age=(16(5)71))
marginsplot, recast(line) recastci(rarea) graphregion(color(white))

Now we see a very different picture emerge. If we focus on young people (e.g., under 30) first, the model predicts that those who were more religious (the blue line) behaved differently compared with other respondents. For those in the most religious group, they had fewer sexual partners when in their late teens and twenties.

The relationship between age and sexual partners is also interesting in later life. With age showing a curvilinear relationship on sexual partners at higher levels of age. Those in the most religious group also show this rise and fall. Keep in mind that this is simply a cross-section of data, and so age is also capturing cohort effects (people who are 60 or 70 now grew up with very different sexual norms than those today, and this may contribute to differences in number of sexual partners).

I do not want to over-interpret the graph, considering that a huge swathe of important variables (e.g., relationship status, values, etc) are not accounted for in the model. My point is simply that the pattern of results is very different using a nonparametric regression, and because we are not imposing linearity on the relationship, the relationship described over is likely to be a better representation of the true relationship found in the world.

Where in England has the highest rate of bankruptcies?

Using official data from the UK Government, I map the UK based on Local Authority Districts, to visualise which areas had the highest rate of insolvencies. This is measured at a rate of per 10,000 of population. 

The highest rate was in Torbay, with 619 insolvencies and a rate of 58.2 per 10,000 of population (0.88%). The lowest was in Wandsworth, with 302 insolvencies and a rate of 12.2 (0.12%).

The interactive map below, created in Tableau, allows you to explore the data and compare different areas.

I’m from Brighton originally, a city on the south coast of England. Brighton had 597 insolvencies, a rate of 27.2 (0.27%), so was somewhere in the middle of the pack.

Which areas in UK have highest and lowest earnings?

Using data from the Annual Survey of Hours and Earnings (2017), I have built a map of weekly earnings across the UK. In the interactive map below you can scroll over specific areas. 

When looking at earnings by where employees live, those in Kensington and Chelsea had the highest median gross weekly income (£862), while those in West Somerset had the lowest (£407).

The map was built in Tableau.

Which occupations have seen largest rise and fall in earnings?

Occupations vary in how much they pay, but they also vary in the rate of which their pay changes. The disrupting force of technology, as well as government pay restrictions during austerity, mean some occupations are paying their workforce less than a year ago.

I plot government data below from 2015 on the percentage change in earnings over the previous year.

The data below suggests that the worst-performing group was air travel assistants, publicans and sheet metal workers. Surprisingly, higher skilled occupations are also on the losing list, including Vets. In contrast, some of the occupations with the highest increases were cooks and those working with glass.

Should we choose the best paid or the happiest career?

Using government data, I plot occupations on both their average subjective wellbeing and median earnings. You can scroll over the points to see which occupation each point represents. The happiest occupation is a clergymen. The most well paid jobs are pilots and doctors. Except for the extreme outlier of clergymen, the graph suggests that the best paid jobs also tend to employ the happiest workers. 

Can "Cash on Hand" Buy Happiness?

While financial advisors would surely disapprove, keeping cash on hand may have hedonic benefits. 

In a paper from 2016 (see here). We show that 'cash on hand'—the balance of one’s checking and savings accounts— may be a better predictor of life satisfaction than income. In a field study using 585 U.K. bank customers, we paired individual survey responses on life satisfaction with account data held by the bank, including the full account balances for each respondent. Individuals with higher liquid wealth were found to have more positive perceptions of their financial well-being, which, in turn, predicted higher life satisfaction, suggesting that liquid wealth is indirectly associated with life satisfaction.

This effect persisted after accounting for multiple controls, including investments, total spending, and indebtedness (which predicted financial well-being) and demographics (which predicted life satisfaction).

Our results suggest that having readily accessible sources of cash is of unique importance to life satisfaction, above and beyond raw earnings, investments, or indebtedness. Therefore, to improve the well-being of citizens, policymakers should focus not just on boosting incomes but also on increasing people’s immediate access to money.

View fullsize

Figure 1. Liquid Wealth and Satisfaction with Life.

View fullsize

Figure 2. Liquid Wealth and Financial distress

What drives savings behaviour? Different psychological traits influence the 'established' and the 'striving' differently. →

Research often assumes that the influence of psychological characteristics on savings behaviour is the same across demographic or socio-economic groups. Yet, it is also possible that psychological characteristics influence an individual’s propensity to save differently based on life-cycle stage, gender, education level, or income – factors which themselves also influence savings behavior.

In a paper published earlier this year, we use a technique called a finite mixture model. We apply this approach to a representative sample of UK households (n = 3382) and identify two different groups of people in the UK: 'striving' and 'established' households. We find that the relationship between psychological characteristics and savings behavior differs across these two classes, demonstrating that  different psychological characteristics – such as self-control – will be more or less influential on savings behavior depending on an individual’s environment and life-cycle stage

View fullsize

Ever wondered how much Oxford business undergrads make after 5 years? Me neither. But now we know it's £160k.

The UK Government has started collecting a new, interesting data set. The longitudinal education outcomes (LEO) data matches tax and benefits data to university graduation data, to see which groups of students earn the most (and least) after graduating. The dataset is publicly available here.

I was playing with the 2017 LOE data, looking at business school students. I was surprised by how dramatic the Oxbridge-skew is in earnings, especially after 5 years.

You can see that while the median income across all business graduates is about £28,000. It is £160,000 for Oxford grads in business. 

UCL, where I work, has the 10th highest median earnings. It is impressive how well Bath undergrads do. Maybe it's the placement year many of them take? But still interesting if that carries-over to earnings years later.

View fullsize

[The dataset is noisy and flawed in various ways. So don't read into this graph too definitatively].