Economic uncertainty and cardiovascular disease mortality

Previous studies have found a link between economic conditions, such as recessions and unemployment, and cardiovascular disease as well as other health outcomes. More recent research argues that economic uncertainty—independently of unemployment—can affect health outcomes. Using data from England and Wales, we study the association between fluctuations in economic uncertainty and cardio-vascular disease mortality in the short term for the period 2001–2019. Controlling for several economic indicators (including unemployment), we find that economic uncertainty alone is strongly associated with deaths attributed to diseases of the circulatory system, ischemic heart disease and cerebrovascular disease. Our findings highlight the short-term link between economic conditions and cardiovascular health and reveal yet another health outcome that is associated with uncertainty.


| Mechanisms and objectives
Why would we expect economic uncertainty to affect cardiovascular disease mortality?The mechanism is believed to be via direct and indirect effects of psychological stress (Browning et al., 2006;Modrek & Cullen, 2013).In particular, we can think of two distinct pathways: a direct effect of stress on triggering acute cardiovascular disease events, as well as an indirect effect of stress on maladaptive coping behaviors, which may hasten the onset of cardiovascular diseases.By analyzing the link between economic uncertainty and cardiovascular disease deaths, we capture the former.The biological processes behind the direct triggering effect has been extensively studied and mainly include acute autonomic dysregulation, neuroendocrine activation and hemostatic and inflammatory responses among others (Bhattacharyya & Steptoe, 2007).Contrary to the impact of smoking, poor diet or lack of physical exercise, which damage the cardiovascular system on a scale of months to years, these mechanisms are relatively short-term and transitory.
There is an important distinction with regards to uncertainty as a determinant of health compared to the impact of unemployment or recessions on health for two main reasons.First, uncertainty does not always translate into changes in one's finances.While some people may work harder in anticipation of future financial turbulence, or may spend less and save more as a precautionary motive in response to uncertainty, others' behavior may remain unchanged until there are any actual changes in their finances.In the absence of increased unemployment, the opportunity cost of time (which relates to actual job loss), and health as an input into production (relating to hazardous working conditions), which are two of the main mechanisms on how recessions affect health that were put forward by Ruhm (2000) (Ruhm, 2000) might apply to fewer people.In any case, the mechanisms might differ to some extent compared to actual job loss, but both can change living conditions and behaviors.
Second, while economic uncertainty sometimes increases during recessions, the two often demonstrate different patterns.A typical example is the United Kingdom in recent years, where the economic uncertainty index reached very high levels after the Brexit referendum (Baker et al., 2021), while unemployment remained very low and the economy was growing (ONS, 2021).A similar situation was observed following the 2016 US Presidential elections (Baker et al., 2021;CBO, 2021).The impact of uncertainty on health may also unfold over a much shorter time scale (e.g., days) compared to those associated with economic fluctuations (typically over months or years) (Vandoros et al., 2018(Vandoros et al., , 2019a)).
Against this background, the objective of this paper is to examine the relationship between economic uncertainty and cardiovascular disease mortality, using data from England and Wales.This study contributes to the literature in a number of ways.First, it examines the impact of uncertainty on cardiovascular disease mortality, rather than the impact of actual adverse events (job loss, income loss) that have already emerged.To our knowledge, this is the first study on this topic.Second, it studies the short-term association between economic uncertainty and cardiovascular outcomes.Third, it provides evidence on yet another impact of uncertainty, in addition to others that have previously been put forward in the literature, such as suicides, car crashes and subjective well-being.It is worth mentioning that our study controls for unemployment, to rule out the possibility of any effect being driven by real economic outcomes instead of uncertainty.
Using variation in weekly data over the period 2001-2019, we found a strong positive correlation between economic uncertainty and mortality relating to strokes and heart attacks in England and Wales.Our results are robust to a number of different approaches and specifications, robustness checks, sensitivity analyses and placebo tests.
The remainder of the paper is organized as follows: Section 2 provides a background on the link between economic conditions and health; Section 3 presents data and methods; Section 4 presents the empirical results; and Section 5 concludes.
A contrasting body of evidence argues that mortality is countercyclical (Gerdtham & Johannesson, 2005;McInerney & Mellor, 2012).Apart from increased mortality, some studies also report a deterioration of physical health or self-reported health (Jofre-Bonet et al., 2018;Svensson, 2007;Vandoros et al., 2013).Another body of evidence suggests that the impact of recessions might not be that relevant for health any longer (Ruhm, 2015)-which is in line with earlier articles (Böckerman & Ilmakunnas, 2009).This suggests that the relationship between the economy and health may differ by period and setting.
There are a number of studies that have focused on the association between economic recessions and cardiovascular health (Asgeirsdottir et al., 2014;Neumayer, 2004;Ruhm, 2000Ruhm, , 2015;;Tapia Granados, 2005).This is not surprising.Cardiovascular disease is the world's leading cause of disability and death, being responsible for 17.9 million deaths in 2019 and representing 32% of global mortality (WHO, 2021).According to recent estimates, their prevalence almost doubled from 271 to 523 million people from 1990 to 2019, while trends in disability and years of life lost are also rapidly increasing (Roth et al., 2020).
Using cross-country data and unemployment rate as a proxy for economic conditions, several studies have shown that cardiovascular mortality and deaths from ischemic heart disease decrease during recessions (Neumayer, 2004;Ruhm, 2000;Tapia Granados & Ionides, 2017).A study based on data from European countries recently demonstrated an inverted-U association between GDP per capita and cardiovascular deaths (Spiteri & von Brockdorff, 2019), whereas a time-series analysis of high-income countries concluded that cardiovascular deaths are not responsive to economic conditions (Dadgar & Norström, 2020).Apart from mortality, existing empirical work has also examined how economic fluctuations impact incidence of cardiovascular disease and resulting hospitalisations.Large-scale economic declines have been found to increase the probability of cardiovascular events (Birgisdóttir et al., 2020;Torbica et al., 2015) and hospital admissions due to acute myocardial infarction (Katz et al., 2016).In addition, the impact of recessions on cardiovascular events has been found to be heterogeneous, as workers who experienced job loss demonstrated higher cardiovascular disease risk, as opposed to those who did not lose their job (Noelke & Avendano, 2015).
Going beyond economic recessions and unemployment, a strand of the literature has focused on the potential impact of stock market performance on cardiovascular health, revealing contradictory evidence.Using data from Asia and the US, several studies show that stock market volatility are associated with increased deaths from coronary heart disease (Ma et al., 2011), higher cardiovascular mortality (Lin et al., 2013) and greater incidence of acute myocardial infarction and stroke (Chen et al., 2012;Fiuzat et al., 2010).Contrary to the findings of these studies, another body of evidence did not find a link between stock market decline and admissions due to stroke and myocardial infarction (Yap et al., 2016) or cardiac deaths (Schwartz et al., 2012).
Last, apart from the actual economic situation and events that have already emerged, overall uncertainty and expectations about future economic activity may also influence cardiovascular health through various channels, such as job insecurity and increased psychosocial stress.The literature has shown that job insecurity, a form of economic uncertainty, can have a negative impact on health (Burgard et al., 2009;Caroli & Godard, 2016;Ferrie et al., 2005;Siegel et al., 2003).Such effects are not limited only to the individual whose job is at risk, but also to other household members (Bünnings et al., 2017).
Recent evidence also shows that economic uncertainty has detrimental impact on several outcomes, such as suicide rates (Antonakakis & Gupta, 2017;Vandoros et al., 2019a;Vandoros & Kawachi, 2021), antidepressant prescription rates (Vandoros et al., 2019b), and newborn health (Kyriopoulos et al., 2019).Road traffic accidents also appear to demonstrate short term spikes due to increased economic uncertainty (Vandoros et al., 2018) or stock market fluctuations (Giulietti et al., 2020), which might be a result of distraction, sleep deprivation or frustration.However, the potential link between economic uncertainty and cardiovascular health has not been empirically examined, despite the burden of disease associated with diseases of the circulatory system.

| Data
We used weekly mortality data for England and Wales spanning over the period 2001-2019 from the ONS, focusing on deaths from diseases of the circulatory system.To obtain data for cause-specific mortality, we relied on the International Classification of Diseases, developed by the World Health Organization (WHO, 2019).This tool systematically consolidates underlying causes of deaths and maps mortality statistics allowing for international and intranational comparability over time.In this study, we examined deaths from diseases of the circulatory system (ICD-10 code: I00-I99) and also studied cause-specific mortality from ischemic heart disease (ICD-10 code: I20-I25) and cerebrovascular disease (ICD-10 code: I60-I69).
To capture economic uncertainty, we used a weekly measure of Economic Policy Uncertainty for the UK, which was calculated based on daily data for Economic Policy Uncertainty Index (EPUI).The EPU index is compiled from textual analysis of digital archives (Baker et al., 2016), covering about 650 national and local newspapers across the UK.It reflects the coverage frequency of articles that include at least one keyword from each of the following term sets: (a) economic or economy, (b) uncertain or uncertainty, (c) spending, deficit, regulation, budget, tax, policy, Bank of England.We took the weekly average of this daily index in order to match deaths that were only available weekly.Table S1 in the Online Supplementary Material shows how the weekly average compares to the daily index.Their average is practically identical (daily: 275.4059; weekly: 275.577; but the daily index is more volatile (standard deviation -daily: 182.5388; weekly: 144.0267).We also obtained data for monthly unemployment rates and population aged more than 16 from the ONS Official Labor Market Statistics.Data on monthly GDP index and Consumer Price Index (CPIH) for the UK were also derived from the ONS. Figure S1 in the Online Supplementary Material shows weekly trends in the uncertainty index and deaths due to cerebrovascular disease, disease of the circulatory system, and ischemic heart disease.

| Empirical strategy
We test the extent to which economic uncertainty is associated with cardiovascular deaths.In doing so, we estimate the following regression model: where  log(Dt) is our main outcome, capturing the natural logarithm of weekly deaths. EPUIt is the weekly average of the EPUI.To standardize this measure, we have divided it by the standard deviation.We are mainly interested in coefficient   which measures the association between economic uncertainty and cardiovascular deaths. Xt is a vector of regressors that includes unemployment rate, consumer price index, GDP growth and population size.We also control for month fixed effects (FE)  mt and year FE  yt .Robust standard errors are reported throughout.After performing an Augmented Dickey-Fuller test, we rejected the null hypothesis of unit roots for both economic uncertainty and our outcomes of interest.
Next, we estimate a model to test potential non-linearity in the association between economic uncertainty and cardiovascular deaths.In doing so, we broke EPUI into terciles and controlled for it through dummy variables for each tercile.After omitting the middle tercile (reference category), we estimated the following equation: where  2), we split the distribution of economic uncertainty into three equal groups (i.e., terciles), each of which includes 330 observations.
After estimating the baseline equation, we perform several robustness checks and sensitivity analyses.First, we run Equation (1) without adjusting for the standard deviation and using the natural logarithm of EPUI.Second, we estimate models after removing extreme values.Third, we test the robustness of our findings by including linear and quadratic time trends in our specification.Fourth, we run the baseline specification after including a lagged dependent variable as a regressor (Giulietti et al., 2020).Fifth, we check the extent to which our main findings are robust to different model specifications, such as Poisson and Negative Binomial regression.Sixth, we perform an analysis using a dependent variable that captures the difference between the observed number of deaths and the rolling 1-month and 1-year average of deaths respectively.

| Baseline estimates
Table 1 presents the summary statistics of the main variables of interest.During the period examined, the weekly average number of cardiovascular deaths is almost 3,083, with a standard deviation of 648.The weekly average number of deaths from cerebrovascular and ischemic heart disease is 815 and 1424 respectively.
Table 2 presents our baseline regression findings.Based on the Augmented Dickey-Fuller test, we reject the hypothesis that our time series have unit roots (Table S2 in the Online Supplementary Material).We begin by showing the estimates of a regression model, in which we have only controlled for economic uncertainty and month and year FE.Columns 1-3 present the findings for the total number of cardiovascular deaths, as well as for deaths from cerebrovascular and ischemic heart disease.In Columns 4-6, we have included additional regressors (unemployment rate, CPIH, GDP and population size), as discussed in Methods section.Economic uncertainty is independently associated with cardiovascular mortality after controlling for several economic indicators, including unemployment, GDP growth and inflation.A one standard deviation increase in EPUI increases the number of cardiovascular deaths by 1.1%.The coefficient of interest is positive and statistically significant in all columns, with its magnitude being similar even after including additional controls.The size of the coefficient is similar regardless the outcome, demonstrating that economic uncertainty has a similar impact on deaths from cerebrovascular and ischemic heart disease.
We next estimate Equation (2) to examine potential non-linearity in the association between economic uncertainty and cardiovascular deaths.In the models presented in Table 3, we control for economic uncertainty through dummies for each tercile of the distribution, allowing for the top and bottom group to exhibit a differential impact on deaths.We find that only uncertainty in the top tercile matters for cardiovascular deaths.These findings demonstrate that high levels of economic uncertainty are particularly detrimental for cardiovascular health, contrary to those in the bottom tercile.The link is therefore asymmetric and stronger in periods of high uncertainty relative to those of moderate uncertainty.Our findings remain robust after the inclusion of several regressors.

| Leads and lags
Going beyond the contemporaneous links, we explore a potential relationship between lagged or future economic uncertainty and cardiovascular mortality.The former would show the duration of the association; while the latter serves as a placebo test.We first examine whether economic uncertainty up to 3 weeks earlier is associated with mortality (Table 4, Columns 1-3) and find some evidence in support of this hypothesis for the first week.Quantitatively, the estimate of the coefficient of the lagged economic uncertainty (Model 3 in Table 4) is lower than that of the contemporaneous impact.This is suggesting of a triggering effect of economic uncertainty on cardiovascular disease mortality.Apparently, our baseline model shows that the main impact on cardiovascular mortality appears in the same week, this additional analysis reveals a residual impact that occurs the week after.There is no association between uncertainty and mortality two or 3 weeks later, which might be a sign of adaptation.(or that the window of triggering effects has passed).On the contrary -and as expected-we do not report evidence of a relationship between future economic uncertainty and current cardiovascular deaths.This finding is encouraging as the models presented in Columns 4-6 essentially serve as placebo tests.

| Additional specifications and robustness checks
Our results are robust to different specifications and approaches.In Table 5, we estimated a model similar to Equation (1) after adding a time trend, apart from month and year FE.Including a time trend allows us to capture the general trajectory of cardiovascular deaths over time.As shown in Columns 1-3, the relationship between economic uncertainty and mortality holds after adding the time trend, with the size of the coefficient being unchanged.We also explore whether the inclusion of the lagged dependent variable as regressor affects the estimates.According to the regression models presented Columns 4-6, the estimates of the coefficient of interest are still significant and similar in terms of magnitude.Table S3 in the Supplementary Material T A B L E 4 Leads and lags of economic uncertainty (circulatory system diseases).
( presents some additional specifications, in which we control for quadratic time trend (Columns 1-3) and time trend and lagged outcomes (Columns 4-6).Our results once again are similar to those of the baseline model and hold the same interpretation.
We also use the differences between the weekly number of deaths and the rolling average of previous periods (i.e., one month and 1 year) as dependent variables.This approach allows us to examine the potential impact of economic uncertainty on the changes in cardiovascular deaths over different time periods.The results presented in Table S4 further confirm our baseline.Our results are also robust to different scaling approaches.In Table S5, we estimate models similar to Equation 1, in which we have either controlled for the logarithm of EPUI without any adjustment or we have scaled the logarithm of economic uncertainty by the time-series standard deviation.Our results remain robust and statistically significant across all outcomes we examined.In addition, we explore the extent to which our results are driven by outliers in the distribution of cardiovascular deaths.After excluding the bottom and top 1% and 5%, we show that our findings remain consistent and quantitatively similar to the baseline results.More details are presented in Table S6 in the Supplementary Material.In Table S7, we estimate a model similar to that presented in Equation (1).Instead of using the natural logarithm of cardiovascular deaths as dependent variable, we regress the actual number of deaths (in levels) against the same set of regressors.After employing Poisson and Negative Binomial models, we confirm that our findings are insensitive to different regression modeling approaches, while the coefficients remain significant and similar in size.Pay days in the UK often fall toward the end of the month, so we also controlled for week FE as an additional robustness check.Our results, presented in Table S8, are similar to the baseline estimates in terms of significance and size of the coefficient of interest.As an additional check for the same reason, Columns 1-4 in Table S9 show the regression estimates after excluding the first, second, third and fourth week of month, respectively.The positive link between economic uncertainty and deaths persists.Finally, the weekly average of the uncertainty index might have a smoothing effect on the highly volatile daily index.While the average takes into account the overall weekly exposure to uncertainty, it might be that the peak of uncertainty is more important.We thus examined whether the weekly peak, range or median of the uncertainty index are associated with weekly deaths.In the additional regressions, that are presented in Tables S10, S11 and S12 in the Online Supplementary Material, the coefficients of the peak and range are statistically insignificant, while that of the median is positive and significant, as in the baseline model.This suggests that the overall weekly exposure to uncertainty is more relevant for cardiovascular disease deaths than its weekly maximum value.

| DISCUSSION
Using data from England and Wales we found that economic uncertainty is positively and significantly associated with cardiovascular disease mortality, after controlling for a number of other economic indicators.Our findings apply to all three different types of mortality: diseases of the circulatory system; ischemic heart disease and cerebrovascular disease.One standard deviation increase in economic uncertainty is associated with a 1.1% increase in deaths from circulatory system diseases.Given that the uncertainty index is volatile, and multi-fold increases in the index are not uncommon, the magnitude of the association between uncertainty and mortality can be sizable.We also show that this link is non-linear and stronger during periods of high economic uncertainty.Contrary to the standard economic indicators, the economic uncertainty index is very volatile and reported at the weekly level, and thus captures an immediate or short-term association between the economy and health.This is particularly important when examining cardiovascular health.In cardiovascular epidemiology, there are two types of risk factors: those that "trigger" acute cardiovascular events -such as acute emotional distress-and the ones associated with long-term progression of atherosclerosis and other conditions.In this context, there are some plausible explanations for our findings.One possible explanation might be that increased uncertainty may act as a trigger, which has been found to increase the risk of cardiovascular events, even within minutes to hours after exposure to various stressors (Leor et al., 1996;Mefford et al., 2020).To add to that, our findings shows that only 1-week lagged and contemporaneous levels of economic uncertainty matter for cardiovascular disease deaths.This largely suggests that economic uncertainty acts as a trigger rather than the sole factor responsible for cardiovascular disease mortality.The biological processes and explanations behind this triggering effect under emotional stress have been widely examined (Bhattacharyya & Steptoe, 2007).There are alternative or complementary explanations for our findings.The triggering mechanism can explain an increase in the number of cardiovascular events, but our paper studies mortality rather than cardiovascular events or hospitalisations.Other explanations could be that individuals might delay seeking care when they experience a cardiovascular event, face increased ambulance response times and barriers to accessing care or receive poorer quality of care -thus worsening the chances of survival (Krumholz et al., 2009;Pell et al., 2001;Scholz et al., 2018).For example, uncertainty may cause increased demand for emergency care or ambulance services due to other conditions such as car accidents (Vandoros et al., 2018(Vandoros et al., , 2019a)), leading to limited capacity.All the factors mentioned above as part of a plausible mechanism may partially explain the findings simultaneously.
We subjected our results to a number of robustness checks.In particular, we tested for non-linearities; employed Poisson and Negative Binomial models and applied different model specifications in sensitivity analyses with step-wise inclusion of control variables or applying dynamic effects.Results persist when using alternative measures of uncertainty or excluding outliers.In a series of other tests, we examined whether past and future levels of economic uncertainty impact deaths.The association between uncertainty and mortality appears to be limited to the same and next week, with mortality two or 3 weeks later not being affected by earlier uncertainty.This might also be a sign of adaptation, in accordance with Bradford and Dolan (2010) and Wilson and Gilbert (2008) (Bradford & Dolan, 2010;Wilson & Gilbert, 2008).
These findings add to those of previous studies that have highlighted the association between economic conditions and cardiovascular morbidity and mortality (Birgisdóttir et al., 2020;Katz et al., 2016;Neumayer, 2004;Tapia Granados & Ionides, 2017).Our results also contribute to a growing body of literature that studies the impact of economic uncertainty on health.There appears to be robust and consistent evidence on a number of health outcomes, such as suicides (Antonakakis & Gupta, 2017;Vandoros et al., 2019a;Vandoros & Kawachi, 2021), newborn health (Kyriopoulos et al., 2019) and car crashes (Vandoros et al., 2018).Sudden events at the macro level causing uncertainty have also demonstrated an impact on health outcomes (Kavetsos et al., 2021;Metcalfe et al., 2011;Vandoros et al., 2019b).As these findings emerge, the role of uncertainty appears to be an important factor that consistently affects health outcomes.Future research can focus on additional outcomes that are yet to be explored.In addition, our results add to evidence on how exposure to population-based events and stressful circumstances trigger the onset cardiovascular events.Studies have shown that cardiovascular health deteriorates due to socio-political events, natural disasters or terrorist attacks (Mefford et al., 2020;Niiyama et al., 2014;Rosman et al., 2021;Shedd et al., 2004).In this context, our findings reveal that economic uncertainty may act, among others, as a trigger for cardiovascular events, but it is unlikely to be the sole cause of cardiovascular disease deaths.
This study is subject to limitations.We only have aggregate data on the number of deaths per week, so we could not control for individual characteristics that may contribute to cardiovascular mortality.Furthermore, we could not stratify by demographic or socioeconomic characteristics, in order to examine who is affected the most by fluctuations in economic uncertainty, as data were not available by socioeconomic or job status etc.There may be significant heterogeneity in the average effects that we have reported with respect to these characteristics.Importantly, we did not have data on cardiovascular events or hospitalisations, so we cannot be sure to what extent the mechanism acts via a triggering effect and to what extent it might be due to delays in seeking care or increased demand for healthcare.An additional limitation is that, as the number of deaths was reported weekly, we had to take the weekly average of the uncertainty index instead of using the daily measure, which possibly had a smoothing effect on the highly volatile daily index.
As cardiovascular disease is a leading cause of mortality, it is important to raise awareness among the population to take particular care especially in periods of increased uncertainty.Campaigns reminding people how to identify a stroke or heart attack and how to immediately react can intensify in such periods, and particular importance can be placed on patient adherence with regards to taking prescribed preventive treatment.In addition, hotlines can be set up to assist people experiencing acute distress during times of uncertainty.
Summary statistics.

Dependent variable Circulatory system diseases Cerebrovascular disease Ischemic heart disease Circulatory system diseases Cerebrovascular disease Ischemic heart disease
Controls include unemployment rate, CPIH, GDP and population size.Robust standard errors in parentheses.

Dependent variable Circulatory system diseases Cerebrovascular disease Ischemic heart disease Circulatory system diseases Cerebrovascular disease
Different model specifications.Controls include unemployment rate, CPIH, GDP and population size.EPUI denotes the index divided by the standard deviation.Robust standard errors in parentheses., 2023, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/hec.4678by Test, Wiley Online Library on [16/07/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License T A B L E 5Note: 10991050