Monthly Archives: March 2021

Deaths in Perspective

It is many years since I watched TV and I have never been a big newspaper reader. It is many months since I listened to the radio, and I certainly do not read the news feeds that are thrown at me every time I look at my mobile or PC. Despite that, there are some news items that have such an insistent presence on the interweb that they are forced onto my attention.

Don’t worry – this will be a data driven post, as usual. But first I must set the context without which this post would be unnecessary.

There are cynical people who will exploit heinous crimes, perhaps the murder of an individual or perhaps the homicide of a class of people, to progress a political or personal agenda. This is highly reprehensible.

There is another type of person who is driven by emotional reasoning and who will easily be manipulated by this kind of exploitation. What may result is a widespread reaction which lacks any balance, judgment or fairness and is actually motivated by pre-existing prejudice. The latest event is just the spark that sets off a conflagration whose true cause is the endemic prejudice that has been long in the building. Consequently, this second type of person is also reprehensible, but with the mitigation that they may have been misled.

The misleading of this second class of person is greatly assisted by a characteristic from which many who are inclined to emotional reasoning suffer: innumeracy. Innumeracy is usually understood to mean merely a poor grasp of elementary arithmetic. But actually the condition is more serious, and may be a cognitive impairment for which the sufferer is blameless. The condition causes such people to have no real feel for the significance of statistics which differ by being large, small or extremely small.

The result may be a wildly exaggerated fear, unjustified by reality, and this fear may be deliberately promoted by the wicked cynical manipulators.

It is more than six years since I last attempted to put death statistics into proper perspective as a counter to the issue discussed above. I do so again here, via the Table below. The Table relates to England and Wales and shows the number of people dying from selected causes in a recent year. For each of these figures I have found the size of the relevant demographic, i.e., the total population of the same sex and age range, which allows the death data to be expressed as a percentage of the corresponding demographic population. The purpose is to give a feel for probability (or prevalence, if you prefer).

For in-utero deaths, neonatal deaths, infant deaths and abortions, the corresponding population defining the denominator is the sum of the number of live births and the number of deaths/destructions in-utero.

For the deaths of rough sleepers the corresponding population defining the denominator is the number of rough sleepers.  For non-resident parents the denominator is the size of the sample to which the death statistic relates.

Notes describing in more detail how the data were derived, and the sources used, follow the Table.

All death data are “per year” and hence the percentage figures, interpreted as probabilities of dying of that cause, are also probabilities per year.

Hence, the percentages give the probability of dying per year of that cause for a randomly chosen person or foetus from the sub-population in question.  

As always with probabilities, one has to be careful about interpretation. For example, for deaths due to alcohol or drugs I have used the total population of England and Wales as the denominator. In reality, the risk from these causes is (obviously) hugely concentrated upon those who abuse these substances (and whose individual probability of dying as a result will be far larger than the average). Nevertheless, for a person chosen at random from the whole population, the probabilities are correct.

Another important interpretational issue is the age at death. Without wishing to be callous or to “kill granny”, the death of an 80-year-old is not the same as the death of a much younger person who might have expected many more decades of healthy life. In this respect, deaths from (actually, with) Covid-19 are not comparable with most of the other causes listed, as the average age at death with Covid is actually rather greater than the overall average age at death. The same applies to the bulk of deaths due to diseases (natural causes). Of greater relevance for comparison purposes, therefore, are early deaths from diseases. Here I refer you to The Empathy Gap, Table 3.3, which breaks down early death, defined as death before age 45, into its various causes. In the Table below I give only the totality of early deaths.

The Table is in order of increasing probability.

One of the points being made, of course, is that the probability of homicide, of all kinds, is small compared with other causes of death. The probability of homicide is greatest for adult men. The probability of homicide for a boy is comparable with, perhaps somewhat larger than, the probability of homicide for a woman (even when domestic and non-domestic homicides are summed).

Domestic homicides are the only category where deaths of females exceed that of males. In all other categories (where the numbers of deaths are far larger) deaths of males outnumber those of females.

Across all causes, early deaths of men (before age 45) exceed early deaths of women by 78%.

Premature deaths of men (before age 75) exceed premature deaths of women by 43%.

CauseNumber of deathsDemographic PopulationPercentage
Domestic homicide (men)2823,800,0000.00012%
Non-domestic homicide (women)8524,500,0000.00035%
Domestic homicide (women)9124,500,0000.00037%
Homicide of girls286,450,0000.00043%
Homicide of boys536,450,0000.00082%
Non-domestic homicide (men)44523,800,0000.00187%
Accidents (female)98630,300,0000.00325%
Drugs (female)1,42530,300,0000.00470%
Suicide (female)1,38826,600,0000.00522%
Accidents (male)2,53129,500,0000.00858%
Alcohol (female)2,62230,300,0000.00865%
Drugs (male)2,96829,500,0000.01006%
Suicide (male)4,30326,000,0000.01655%
Alcohol (male)4,99829,500,0000.01694%
All early death (<45 years), female6,09817,000,0000.03587%
All early death (<45 years), male10,82916,600,0000.06523%
Women 25 to 55, all causes0.11600%
Men 25 to 55, all causes0.19100%
Covid-19 (female)60,94030,300,0000.20112%
Neonatal death (<28 days)1,771867,0620.20425%
Covid-19 (male)72,32229,500,0000.24516%
All infant deaths (<1 year)2,387867,0620.27530%
Still births / death in-utero2,689867,0620.31013%
Non-resident parents in CMS1,189362,2450.32800%
All diseases (both sexes)579,65359,800,0000.96932%
Rough sleepers (women)914,9501.83838%
Rough sleepers (men)68728,0502.44920%
Table: Probabilities of dying per year for a randomly chosen individual from the demographics indicated

Population: I have used death data from a mix of recent years, from 2016/17 to 2019/20, but for simplicity I have used time-invariant populations of England & Wales as follows,

  • All: 29.5M men, 30.3M women.
  • 16 and above: 23.8M men, 24.5M women
  • Under 18s: 6.45M boys, 6.45M girls
  • 10 and above: 26.0M males, 26.6M females.
  • Under 45: 16.6M men, 17.0M women.

Homicides: Domestic homicide data are taken from [1], using the average over the three years April 2016 to March 2019. However this source appears misleading for non-domestic homicides, for which I have used [2]. Here I have averaged over the four years 2016/17 to 2019/20. This source breaks down data into age ranges. Note that the terms “men” and “women” in this context relate to homicides of those aged 16 and above. In the Table, “boys” and “girls” refers to those under 18 (so there is some double accounting with the adult range). As [2] gives data in the age range 16-24, I have used 2/9 of these data to represent the age range 16-17, which is crude but adequate for the present purposes.

Suicide: Data from [3] relates to those aged 10 and above (2019).

Covid-19: I have used the official statistics based on “death within 28 days of a positive test”. This is, at best, death with Covid, not necessarily of Covid. Hence these data are upper bounds. Ref.[4] gives data to end-February 2021, which is a reasonable cut-off as it represents just 12 months of Covid data, the epidemic within the UK taking off in March 2020.

Alcohol: I have used data from 2016, [5], after which the definition changed. There is potential ambiguity here as many cancers are epidemiologically related to alcohol consumption but, being non-deterministic, will not be included here. The data here relates only to cases which can be unambiguously related to alcohol.

Drug Poisoning: Data from [6] for 2019.

Accidents: These data include all accidents: road accidents, accidents at work or elsewhere, accidental deaths due to falls, fire, drowning, noxious substances, etc. I have been a bit lazy here and used data from The Empathy Gap, Table 3.1, which is for 2016 and actually relates only to deaths before age 75. So these data are likely to be under-estimates.

Rough Sleepers: The number of deaths of rough sleepers in 2019 has been taken from [7]. The size of the population of rough sleepers is subject to serious ambiguity, as explained at length in The Empathy Gap chapter 16. Sources like [8] and [9] only attempt an estimate of how many people are sleeping rough on a specific night. Even this is misleading because it fails to account for people in night shelters on that night, who might be back on the street the next night. More importantly, though, the rough sleeping population is not static. There is a huge difference between the number of people sleeping rough on a given night and the number of people who have slept rough at some time during a given year. Data from CHAIN in London suggest the latter is 7 or 8 times larger than the former. On this basis, in The Empathy Gap I estimated that around 36,000 people had slept rough in England at least once in 2017/18. Some rough corroboration of a figure of this magnitude comes from the “Everyone In” project in 2020, to get rough sleepers off the streets during the Covid pandemic, which supported 33,000 people to do just that, Ref. [9]. In short, I have used 33,000 as the relevant population. Based on [8], 85% of rough sleepers are men. Note that [7] indicates 88.3% of deaths of rough sleepers were men in the same year, 2019. Consequently it would seem that male rough sleepers are somewhat more likely to die than female rough sleepers. (37% of rough sleeper deaths are attributable to drugs, and 14.4% to suicide).

Abortion data taken from [10] for 2019 (all time high), 207,384.

Births & Infant Death: Taken from [11]: live births 656,989; still births (including deaths in-utero) from week 24 onwards 2,689; neonatal deaths (i.e., <28 days) 1,771; all infant deaths (<1 year) 2,387. Denominator for this category, and abortion, 656,989 + 2,689 + 207,384 = 867,062.

Early Death: I define this as deaths from all causes before age 45, see The Empathy Gap, Table 3.3 (2016).

Non-Resident Parents paying into CMS: Data here is taken from my earlier post Deaths Whilst Paying Child Maintenance, such non-resident parents (overwhelmingly fathers) have an elevated death rate compared with men or women of comparable age, i.e., mostly 25 to 55. The latter are also included in the Table for comparison. Note that I have converted the data in Table 1 of the earlier post to a yearly figure by dividing by 2.75.