Month-of-Birth Patterns in the Northern and Southern Hemisphere
One of the first scholars to be interested in the effect of month of birth on longevity was Ellsworth Huntington. In his book “Seasonality” (1938) he described the relationship between the seasons of the year and social, psychological, and demographic phenomena.
On the basis of 10,890 persons in the Dictionary of American Biography
who died at an average age of 68.9 years, he found that those born in Feb-
ruary died at age 69.7, whereas those born in June at an average age of
67.8. He observed a secondary maximum in September and October. He
obtained similar results for 13,891 New Englanders from genealogical
memoirs of 80 families; 3,019 New Englanders and New Yorkers; 12,173
people from New York, New Jersey and Pennsylvania and 9,921 people
from Maryland, Virginia, North and South Carolina, a few from states
farther south or west; and 7,517 people in the British “Who Was Who”.
Huntington comes to the conclusion that, in all populations (with the ex-
ception of the British population), people born in February or March live
longer than those born in July or August. Huntington’s finding stimulated
extensive research in the area of psychology. Since the 1930s more than
200 studies have been conducted about the prevalence of schizophrenia
and month of birth (for a review see Torrey 1997). More recently, re-
searchers attempted to attribute the differences in the prevalence of schizo-
phrenia by month of birth to maternal virus infections such as measles or
flu during pregnancy. The results, however, are not conclusive.
Eysenck took up the topic of the impact of month of birth on life span in
his book “Astrology. Science or Superstition?” in 1982 (Eysenck & Nias
1982) and Miura edited a series of epidemiological studies in his book en-
titled “Seasonality” (Miura 1987). In the 1980s and 1990s quite a few epi-
demiological studies were conducted that found a relationship between
month of birth and specific diseases such as allergies, insulin-dependent
diabetes, congenital malformations, Parkinson’s and Alzheimer’s disease,
breast cancer, etc. (see Chapter 5). In all these studies month of birth is
used as an indicator for the environment early in life.
20 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
This chapter takes up the topic and provides evidence from population
data that month of birth and life span are indeed related. In contrast to ear-
lier research, this study is not based on sample data but on complete
population data for countries of the Northern and Southern Hemisphere.
Huntington conjectured that the month of birth is an indicator for envi-
ronmental factors that are linked to the seasons of the year. If this is true,
then the patterns of two geographically close populations should resemble
each other, and the pattern in the Northern Hemisphere should be mirrored
in the Southern Hemisphere. Furthermore, life spans of people who were
born in the Northern Hemisphere but who died in the Southern Hemi-
sphere should resemble the pattern of the Northern Hemisphere.
The first step, therefore, was to obtain data on the populations of Den-
mark, Austria, and Australia to test this conjecture. Statistics Denmark
provided longitudinal data based on the Danish population register that
follows every person living in Denmark from 1968 to the present. Statis-
tics Austria and Statistics Australia provided death certificates for all
deaths that occurred between 1988 and 1996 in Austria, and between 1993
and 1997 in Australia. In addition to these three countries, data were also
obtained for Hawaii. Hawaii is of particular interest for the study of differ-
ences in life span by month of birth because it is located close to the
equator. The data for Hawaii are based on US death records for the years
1989 to 1997; these records include place of birth.
The optimal data to test for differences in life span by season of birth are
longitudinal data. Birth cohorts born in a specific season are followed from
birth to death and life expectancy can be calculated using simple life-table
methods. Such data rarely exist however. The data that are closest to this
requirement are register data from the Scandinavian countries. The Danish
data used in this study consist of a mortality follow-up of all Danes who
were at least 50 years old on 1 April 1968. This is a total of 1,371,003
people, who were followed up to week 32 of 1998. The study excludes
1,994 people who were lost to the registry during the observation period.
Among those who are included in the study, 86% (1,176,383 individuals)
died before week 32 of 1998; 14% (192,626 individuals) were still alive at
the end of the follow-up.
Thus, for Denmark both the risk population and the number of deaths
are known, which means that it is possible to estimate remaining life ex-
pectancy at age 50 on the basis of life tables that were corrected for left
truncation. This was achieved by calculating occurrence and exposure ma-
trices that take into account an individual’s age on 1 April 1968. For ex-
ample, a person who was 70 at the beginning of the study and who died at
age 80 enters the exposures for ages 70 to 80 but is not included in the ex-
2 Month-of-Birth Patterns in the Northern and Southern Hemisphere 21
posures for ages 50 to 69. The central age-specific death rate is based on
the occurrence-exposure matrix. The corresponding life-table death rate is
derived by means of the Greville Method (Greville 1943).
Population registers do not exist for Austria, Hawaii, and Australia,
where only individual death records are available. Exact dates of birth and
death are known for a total of 681,677 Austrians who died between 1988
and 1996 and for 219,820 native-born Australians who died between 1993
and 1997 at ages 50+. 42,969 decedents of similar age were born in Hawaii
and died between 1989 and 1997. The population at risk, however, is un-
known, which means that life span by month of birth cannot be estimated
on the basis of simple life-table techniques. For Austria, Australia, and
Hawaii remaining life span at age 50 was therefore estimated by calculat-
ing the average of the exact ages at death.
Mean age at death is not equivalent to life expectancy when cohorts are
not extinct. Gavrilov and Gavrilova (2003) pointed out that mean age at
death is influenced by changes in the seasonal distribution of births. In
Northern Europe the number of births generally peaks in February and
March and reaches a minimum in December. Suppose that in younger co-
horts the seasonality in births has become smaller as compared to older
cohorts. This would imply that in younger populations there are proporti-
anally less people born in winter and more born in fall. In death data this
shift in the birth distribution will be reflected in a higher mean age at death
for those born in winter and a lower mean age at death for those born in
fall.
The usual procedure to account for the effect of possible shifts in the
seasonal distribution of births is to compare the seasonal birth distribution
at the time of birth of a cohort with the birth dates of the deceased or the
survivors of the birth cohort at a given age. Unfortunately, for Australia
and Hawaii this information is not available. For Austria the seasonal dis-
tribution of births is recorded in statistical yearbooks for the years 1881-
191.2 and it will be compared with the birth dates of the survivors in the
1981 census. The resulting pattern is compared with the month-of-birth
pattern on the basis of the mean age at death.
Furthermore, the Danish twin registry is drawn on for the calculation of
remaining life expectancy using the information about the population at
risk and the number of deaths. The month-of-birth pattern is then com-
pared with the pattern on the basis of the death counts alone.
In all these analyses the age range is generally restricted to ages 50 and
above so as to exclude the possibility that the differences in life span sim-
ply reflect differences in survival during the first part of life by month of
birth rather than at old age.
22 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
2.1 Life Span by Season of Birth
A similar relationship between month of birth and life span exists in both
of the Northern. Hemisphere countries. Adults born in the autumn (Octo-
ber—December) live longer than those born in the spring (April—June). The
difference in life span between the spring- and autumn-born is twice as
large in Austria (0.6 years) as in Denmark (0.3 years).
In Denmark remaining life expectancy at age 50 is 27.52 years. The av-
erage age at death is lowest among those born around week 18, and it
peaks at week 51. The life span of Danes born in specific months varies
periodically around the mean (Fig. 2.1.C). For those born in the second
quarter, life spans are 0.19 ± 0.05 years shorter than average; for those
born in the fourth quarter they are 0.12 ± 0.04 years longer than average.
This difference is statistically significant (Cox-Mantel statistic: p<0.001).
In Austria deaths occurred at an average age of 77.7. The mean life span
of people born in specific months of the year deviates from this average
(Fig. 2.1.A). The average age at death is lowest for those born around
week 20 and highest for those born around week 46. The deviation in
mean age at death is highly significant (Bonferroni test: p<0.001) for those
born in the second and the fourth quarters. The life spans of people born
between weeks 14 and 26 are 0.28 ± 0.03 years below average; life spans
of those born between weeks 40 and 52 are 0.32 ± 0.03 years above aver-
age.
The pattern in the Northern Hemisphere is mirrored in the Southern
Hemisphere. The mean age at death of people born in Australia in the sec-
ond quarter of the year is 78.0; those born in the fourth quarter die at a
mean age of 77.65. The difference of 0.35 years is statistically significant
(Bonferroni test: p<0.001) (Fig. 2.1.B).
Significant differences in the life span by month of birth exist for Ha-
waii (Fig. 2.1.D). The mean age at death of people born in Hawaii is 74.5.
The mean age at death of the March-born is 0.53 years above the average;
that of the October-born is 0.29 years below the average. The difference
between the two months is statistically significant at p=0.004 (Bonferroni
test).
Based on the results reported above one can conclude that the pattern in
the Southern Hemisphere is a reversed mirror image of the Northern
Hemisphere pattern. This can be clearly seen from the correlations of the
mont-of-birth patterns. The correlation of the deviations in life span by
month of birth between Austria and Denmark is 0.83 (Pearson correlation,
one sided test: p<0.0001). Between Austria and Australia it is -0.79
(p<0.001) and between Denmark and Australia -0.80 (p<0.001).
Week of Birth Week of Birth
Figure 2.1. Austria (A) and Australia (B): Mean ages at death by week of birth.
The fitted line is estimated by a cosine term with a period of 52 weeks.
Single cosinor analysis with a period of 52 weeks (or 12 months) shows
that in all four populations the differences in lifespan follow a seasonal
pattern. For each of the four populations, ages at death (Austria, Australia
and Hawaii) and remaining life expectancy at age 50 (Denmark) are fitted
24 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
Males Females
C Denmark
0.8
2 7 12 17 22 27 32 37 42 47 52Week of Birth -0.8 2 7 12 17 22 27 32 37 42 47 52 Week of Birth
D Hawaii
0.8
0.6
0.4
0.2
-0.0
-0.2
-0.4
-0.6
1 2 3 4 5 6 7 8 9 10 11 12 Month of Birth -0.8 1 2 3 4 5 6 7 8 9 10 11 12 Month of Birth
Figure 2.1. (continued) Denmark (C) and Hawaii (D): Mean ages at death by
month of birth. The fitted line is estimated by a cosine term with a period of 12
months (Hawaii) and 52 weeks (Denmark).
by a cosine term age= ao*cos(t-ad, where ao is the amplitude, al the acro-
phase (maximum), and t=week of birth /52 *271- (Hawaii: t=month of birth
/12*271). The parameters are estimated using the least squares estimation
procedure. Figures 2.1.A—D show the weekly means together with the fit-
2.1 Life Span by Season of Birth 25
New South Wales Victoria
1.6 1.6
1.2
0.8 0.8
0.4
T.
ca
w 0.0 0.0
>-
-0.4
-0.8 -0.8
-1.2
-1.6 -1.6
2 7 12 17 22 27 32 37 4247 52 2 7 12 17 22 27 32 37 42 4752
Week of Birth Week of Birth
Queensland South Australia
1.6 1.6
1.2 1.2
0.8 0.8
0.4 0.4
N 0.0 0.0
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
2 7 12 17 22 27 32 37 42 47 52 2 7 12 17 22 27 32 3742 47 52
Week of Birth Week of Birth
Western Australia Tasmania
1.6
1.2 4
3
0.8
2
U) 0.4 1
ao
>-
-0.4 -1
-2
-0.8
-3
-1.2 -4
-1.6 • -5
2 7 12 17 22 27 32 37 42 47 52 2 7 12 17 22 27 32 37 42 47 52
Week of Birth Week of Birth
Figure 2.2. Australia: Mean age at death by week of birth and region of birth.
The fitted line is estimated by a cosine term with a period of 52 weeks. The co-
sine term is significant for all regions with the exception of Queensland and Tas-
mania.
26 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
ted cosine terms. The fit of the sinusoidal functions is highly significant,
with p<0.001.
For Austria the data allow us to analyse causes of death for the years
1990-1997 (Table 2.1). There is a general tendency for people born in the
first half of the year, especially in the second quarter, to die at younger
ages than those born in the second half of the year, especially in the fourth
quarter. The differences between people born in the second and fourth
quarters are significant for both heart disease and cerebrovascular disease.
In the group of malignant neoplasms, significant differences exist for
stomach cancer and the residual group ‘other neoplasms’. Differences for
lung cancer, cancer of the urinary system, and diabetes mellitus are of bor-
derline significance (p=0.06). Significant differences exist for chronic res-
piratory diseases, pneumonia and influenza, digestive diseases, the residual
group ‘other natural causes of death’, and for violent causes of death. Al-
though the Austrian death data consist of more than 600,000 decedents, the
number of deaths from a particular cause of death can become quite small,
especially when the data are further distinguished by month of birth. Some
of the non-significant results may thus be caused by insufficient numbers
of observations. Chapter 5 reports the cause-specific results from the
analysis of about 16 million death certificates from the United States.
Since Australia is a whole continent, the analysis of the month-of-
pattern for different birth regions may provide evidence concerning the
underlying causal mechanisms. It appears that the reversal of the pattern
seen in the Northern Hemisphere exists in almost all Australian states and
territories: the mean age at death in the first half-year is generally higher
than in the second half-year (Table 2.2, Fig. 2.2).
The peak-to-trough difference between those born in the second and the
fourth quarter is significant for New South Wales (0.39 years), Victoria
(0.89 years), and South Australia (0.61 years). It is of borderline signifi-
cance (p<0.1) for Western Australia (0.49 years). The largest difference
(p=0.157), which is however not significant, exists for Tasmania, with 1.6
years. The cosine analysis supports these results. The cosine functions are
generally highly significant with the exception of Tasmania and Queen-
sland. The regional differences in the pattern of life span by month of birth
are not significant (ANOVA F-test for the interaction effect between
month of birth and region: p=0.585).
The most interesting result of the regional analysis is that the peak-to-
trough difference for Queensland is not significant. The non-significant re-
sult is not due to small numbers of observations. There are more observa-
tions for Queensland than, for example, Victoria, where the difference is
highly significant. Figure 2.3 reveals that in Queensland, like in the other
Table 2.1. Difference in mean age at death for people born in a specific season from average age at death by major causes of death;
Austria 1990-1997.
Causes of death (ICD Number)
Season of birth
Number of
deaths
Winter
(Jan-Mar)
Spring
(Apr-Jun)
Summer
(Jul-Sep)
Autumn
(Oct-Dec)
p*
Heart disease (390-429, 440-458)
-0.08
-0.30
0.06
0.33
0.00
257,167
CVD (430-438)
-0.01
-0.23
-0.03
0.27
0.00
80,367
Malignant neoplasms
Breast (174)
0.12
-0.25
0.00
0.14
0.53
12,201
Uterus & female genital organs (179-184)
0.05
-0.08
0.10
-0.07
0.92
9,530
Prostate & male genital organs (185-187)
-0.18
-0.05
0.01
0.24
0.40
9,401
Urinary system (188-189)
-0.09
0.07
-0.39
0.42
0.06
8,508
Haemoblastoses (201-208)
-0.19
-0.07
-0.02
0.28
0.44
9,147
Lungs (162)
-0.14
-0.18
0.06
0.26
0.06
24,178
Stomach (151)
-0.10
-0.44
0.03
0.50
0.00
12,689
Intestines (152-154)
-0.02
-0.12
0.12
0.02
0.70
21,233
Other
-0.07
-0.24
-0.01
0.33
0.00
39,359
Respiratory system (460-479, 488-519)
0.04
-0.18
-0.16
0.33
0.05
18,018
Pneumonia and influenza (480-487)
-0.01
-0.10
-0.35
0.45
0.02
10,807
Digestive system (520-579)
0.23
-0.20
-0.28
0.24
0.01
27,384
Diabetes mellitus (250)
-0.06
-0.33
0.19
0.20
0.06
13,537
Other natural
-0.26
-0.22
-0.08
0.58
0.00
27,197
Violent (E800-E999)
-0.04
-0.41
-0.01
0.49
0.00
21,998
Bold figures: maximum and minimum difference from average age at death
* p value: Anova F-Test for all seasons
1111-UT jo uoseas ueds oPI
Table 2.2. Mean age at death by state of birth and quarter of birth, amplitude, and peak estimated by cosinor analysis based on
Australian death records for the years 1993 to 1997.
Quarter of Birth
Max. Dif+
Cosinor Analysis
Deaths
1
2
3
4
Amplitude
Maximum
Week
New South Wales
77.1
77.2
77.0
76.9
0.39**
0.23
19
126,308
0.14-0.31
16-22
Victoria
77.1
78.2
78.1
77.3
0.89**
0.38
22
14,130
0.13-0.96
16-27
Queensland
77.4
77.7
77.5
77.6
0.31
n.s.
n.s.
20,686
South Australia
78.3
78.5
78.2
77.9
0.61**
0.33
20
32,260
0.17-0.49
16-24
Western Australia
77.5
77.4
77.2
77.0
0.49**
0.24
16
23,395
0.05-0.43
9-22
Tasmania
78.6
79.5
78.3
77.9
1.60
n.s.
n.s.
1,843
**
Bonferroni test: p<0.01, * Bonferroni test: p<0.1, bold figures indicate the minimum and the maximum mean age at death
+
Maximum difference in mean age at death between quarters
n.s.
cosinor term not significant
2 Month-of-Birth-Patterns in the Northern and Southern Hemisphere
2.1 Life Span by Season of Birth 29
Australian states, mean
0.6 age at death is highest for
those born in May or
0 4
1 t
0.2 ever, for those born in
-0 0 4 / January and February.
The two major cities of
0.2 /1 Queensland are Brisbane
(27.39 South Latitude,
153.12 East Longitude)
-0 6 and Tovvnsville (19.25
1 2 3 4 5 6 7 8 9 10 11 12
Month of Birth South Latitude, 146.77
East Longitude). Queen-
Hawaii – Queensland
sland is geographically
almost a min-or image of
Figure 2.3. Queensland and Hawaii: Deviation in Hawaii, whose capital,
life span for people born in a specific month from Honolulu, is situated at
the average remaining life span at age 50. 21:18 North Latitude and
157:51 West Longitude.
Comparing the month-of-birth patterns for these two countries one finds
that they are reversed: in Hawaii mean age at death is highest for those
born during the first four months, in Queensland it is lowest; from May on
mean age at death is generally lower in Hawaii and generally higher in
Queensland. The correlation between Hawaii and Queensland is -0.594
and significant at p=0.021.
Comparing the seasons of the year in both Hawaii and Queensland,
those born at the beginning of the winter season, which is characterized by
lower temperatures, live longer. In tropical northern Australia the wet sea-
son corresponds with summer and lasts from November through April. The
dry season corresponds with winter and lasts from May through October.
In Hawaii the wet season coincides with moderate temperatures and lasts
from October to May, while the hot and dry summer lasts from June to
September.
The Australian data distinguish between native-born Australians of for-
eign heritage and aborigines and contains the death records of 2,254 Abo-
rigines and 218,279 native-born non-indigenous Australians. These figures
exclude records with birth dates 1 January and 1 July because of a heaping
of birth dates on these two days. It is highly likely that unknown birth
dates were assigned randomly to one of these two dates. The mean age at
death of aborigines is almost 10 years lower (67.9 years) than that of the
non-indigenous Australian population (77.8 years). The week-of-birth
pattern is similar (Fig. 2.4), but the amplitude is much larger among abo-
30 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
rigines (1.17 years, 95% CI: 0.40-1.94). The cosine function peaks in week
20 (95 % CI: 15-25). The amplitude among the native-born non-
indigenous Australian population is 0.24 years (95% CI: 0.18-0.30), mean
age at death peaks for decedents born in week 19 (95% CI:17-22). The fit
of the cosine functions is highly significant at p<=0.001.
Aborigins Native-Born Australians
0.8
0.6
0.4
U)0.2
.41111ni►, 1 1. a) -0.0
I 7’VMM ›–0.2
-0.4
-0.6
-0.8
2 7 12 17 22 27 32 37 42 47 52 2 7 12 17 22 27 32 37 42 47 52
Week of Birth Week of Birth
Figure 2.4. Aborigines and native-born Australians of foreign heritage: Devia-
tion in mean age at death for people born in a specific week from the average
age at death (ages 50+) estimated by a cosine function with a period of 52
weeks.
2.2 A Different Approach
If people born in a specific month experience a higher mortality risk than
others, the distribution of birth dates of the total population changes with
age (Vaupel & Yashin 1985). This can be easily tested on the basis of
population censuses. A simple tabulation of the Austrian 1981 census
gives the proportion of those born in the first and fourth quarters for five-
year age-groups. A lower mean age at death for people born in the second
and third quarters implies that, with advancing age, the proportion of peo-
ple born in the first and fourth quarters of the year will increase. This is
exactly what is to be observed in the Austrian 1981 census, where the pro-
portion increases by almost 5 percentage points from 49.7 per cent at ages
50-54 to 54.5 per cent at ages 95+. The change in the proportion is statisti-
cally significant at p=0.0001 (x2 test).
2.3 Bias in the Month-of-Birth Pattern 31
This change could also be due, however, to a change in the seasonal
distribution of births. For example, it has been shown that, in Austria, the
seasonal fluctuations in the number of births changes between the periods
1881-1912 and 1947-1959 (Doblhammer et al. 2000). A better approach,
therefore, is to compare the seasonal distribution of births of a cohort with
the seasonal distribution of birth dates among survivors fifty or more years
later. For Austria the seasonal distribution of births is available for almost
all years between 1881 and 1912. The statistical yearbooks, which were
published annually by the Central Bureau of Statistics of the Austro-
Hungarian Empire, contain the number of births by month of the year. It is
therefore possible to compare the seasonal birth distribution of the years
1881 to 1911 in the German-speaking regions of the Austro-Hungarian
Empire with the distribution of the survivors in the 1981 Austrian census.
At the time of the census the survivors were between the ages of 70 and
100 (Table 2.3). The distribution of birth dates is clearly different at the
time of birth and in the 1981 census. For all birth cohorts the proportion of
the fall-born increases with age. The increase is 1.1 percentage point for
the youngest cohort (1907-1911) and 4.4 percentage points for the oldest
cohort (1881-1886).
Table 2.3. Percent of winter births at the time of birth and in the Austrian 1981
census for different birth cohorts.
Birth cohort Age of the co- % winter births % winter birth % Difference
hort in the at the time of dates in the
1981 census birth 1981 census
1881-1886 95-100 50.1 54.5 4.4
1887-1891 90-94 50.2 53.8 3.6
1892-1896 85-89 49.9 53.3 3.3
1897-1901 80-84 49.8 52.4 2.6
1902-1906 75-79 49.9 51.4 1.5
1907-1911 70-74 49.7 50.8 1.1
2.3 Bias in the Month-of-Birth Pattern
In the case of extinct cohorts, mean age at death is similar to life expec-
tancy estimated by the life-table method. The data that exist for Austria,
32 2 Month-of-Birth Patterns in the Northern and Southern Hemisphere
Australia, and Hawaii, however, do not come from extinct cohorts and
mean age at death is therefore a biased estimate of life expectancy.
One unobserved factor that causes bias in the month-of-birth pattern is a
change in the seasonal distribution of births over time. In Northern Europe
the number of births usually peaks in February and March, declines there-
after, reaches a secondary peak in September and a trough in December. In
the United States the pattern is reversed with a trough in spring and a peak
in late summer and fall (Lam & Miron 1996). The basic seasonal pattern
has not changed over time but in some countries such as the United States
the fluctuations became smaller between 1947 and 1976 (Seiver 1985), in
other countries such as Austria they became larger (Doblhammer et al.
2000).
Take the case of Austria, where information about the monthly number
of births exists for the years 1881 to 1912 and 1945 onwards. No informa-
tion is available starting from World War I, which led to the collapse of the
Austrian Hungarian Monarchy and the destruction of most administrative
structures, until the end of World War II.
The seaonal pattern develops as follows: the peak in February decreases
from 8.83% for birth cohorts 1881-1890 to 8.73% for birth cohorts 1891-
1900. It remains stable for birth cohort 1901-1912 and considerably in-
creases (9.15%) for birth cohorts 1947 to 1959. The intermediary Septem-
ber peak sharply increases between 1881-1990 (8.08%) and 1891-1900
(8.33%) and remains stable thereafter, while the December trough remains
almost unchanged between 1881-1890 and 1947-1959. These changes in
the seasonal distribution of births result in proportionally more younger
decedents among the February-born, and mean age at death will therefore
underestimate true life expectancy.
In the following an attempt is presented for assessing the bias by using
the information about the seasonal birth distribution for the years available.
In the Austrian death records of the years 1988 to 1996 decedents aged 50
and above were born between 1881 and 1946. The frequency distribution
of birth years is: 1881-1890: 0.14%, 1891-1900: 7.2%, 1901-1912: 44.3%,
1913-1922: 25.4%, 1923-1932: 16% and 1933-1946: 7.1%. To calculate
the number of expected decedents by month of birth, one needs the aver-
age seasonal birth distribution for the time period 1881-1946. This average
is derived by weighting the seasonal birth distributions for the respective
time period with the according frequencies of birth years observed in the
death records. Since no seasonal birth distribution is available for the time
periods 1913-1922, 1923-1932 and 1933-1946, it is assumed that the sea-
sonal distribution of the period 1901-1912 holds true for the period 1913-
1922, while from 1923 onwards the seasonal distribution of births follows
the pattern observed in 1947-1959. The distribution of birth months in the
2.3 Bias in the Month-of-Birth Pattern 33
death records is then compared with the distribution of birth months in the
average seasonal birth distribution. A similar calculation is performed for
all decedents aged 80+. In the death records they were born between 1881
and 1916 and the seasonal distribution of the number of births is known
with the exception of the last 4 years. Again a weighted seasonal birth dis-
tribution is calculated and compared with the distribution of birth months
in the death records.
Figure 2.5 shows the month-of-birth pattern based on mean age at death,
together with the percentage deviation of the birth months in the death rec-
ords for ages 50+ and 80+ as compared to the average seasonal birth dis-
tribution for the time period 1881-1946 (ages 50+) and 1881-1916 (ages
80+). For example, the value -0.06 indicates that for ages 80+ the propor-
tion of the February-born is 0.06 percentage points lower in the death rec-
ords than in the corresponding seasonal birth distribution of the years
1881-1916.
Comparing the two trajectories in Figure 2.5.A it appears that the pattern
based on mean age at death is shifted to the left for those born in the first
half-year. In other words, excess mortality, particularly of those born in the
first three months, is overestimated. This is especially true for ages 50+.
For ages 80+ the patterns based on mean age at death and on the frequency
distribution of birth dates are close, and no serious bias is introduced by
using mean age at death as a measure of life expectancy.
The seasonal birth distribution for the period 1881-1912 is based on the
U)
a)
>-
0.4
0.3
0.2
0.1
-0.0
-0.1
-0.2
-0.3
-0.4
A
4
3
2
1
0
-1
-2
-3
-4
0.15
0.10
0.05
0.00
-0.05
-0.101
I
1 2 3
4 5
6 7
8 9
101112
Month of Birth
8
6
4
2
0
/\
– -2
■ /
– -4
– -6
Percentage Points
1 2 3 4 5 6 7 8
9 1011 12 -8
Month of Birth
(obs/exp)-1 ages 50+
— (obs/exp)-1 ages 80+
Mean age at death ages 50+
— — Mean age at death ages 80+
Figure 2.5. Deviation of mean ages at death from average mean age at death for
ages 50+ (A) and 80+ (B) and percent deviation of the observed seasonal birth dis-
tribution for ages 50+ and 80+ from the expected distributions. Austrian death re-
cords 1988-1996 and seasonal distribution of births for the years 1881-1912 and
1947-1959.
https://www.demogr.mpg.de/books/drm/002/2.pdf
