NOT FOR QUOTATION WITHOUT PERMISSION OF THE AUTHORS
MORTAWTY
IN
ITALY:CONTOURS
OF A CXNTURY OF EVOLUTIONGraziella Caselli James Vaupel Anatoli Yashin
April 1 9 8 5 C P - 3 5 - 2 4
Dr. Graziella Caselli is from the Department of Demographic Science, University of Rome, Italy.
Collaborative Rzpers report work which has not been performed solely a t t h e International Institute for Applied Systems Analysis and which has received only Limited review. Views o r opinions expressed herein do not necessarily represent those of the Institute, its National Member Organizations, or other organizations supporting the work.
INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS 2361 Laxenburg, Austria
Contour maps of l t a l i a n male a n d female mortality r a t e s from a g e 0 t o 79 a n d from 1870 t o 1979 graphically display p e r s i s t e n t global a n d p r o m i n e n t local p a t t e r n s of mortality, simultaneously over a g e , by period, a n d f o r c o h o r t s . The maps give demographers visual a c c e s s t o previously recognized f e a t u r e s of t h e evolution of ltalian mortality a s well a s focusing a t t e n t i o n on s o m e n e g l e c t e d f e a t u r e s . Use of c o n t o u r maps to display v a r i o u s kinds of mortality d a t a , includ- ing mortality comparisons, may h e l p demographers b e t t e r u n d e r s t a n d t h e social a n d biological d e t e r m i n a n t s of mortality.
MORTALJTY IN JTALY:
CONTOURS OF A CENTURY OF EYDLUTION
Graziella Caselli, James Vaupel a n d Anatoli Yashin
INTRODUCTION
Contour maps provide a n intelligible a n d graphically striking means of summarizing large a r r a y s of dem.ographic data t h a t are s t r u c t u r e d by age and time. Here we use contour maps t o depict both t h e general development a n d t h e prominent local details of age-specific male and female mortality r a t e s for Italy from 1870 t o 1979. Contour maps permit demographers to visualize mor- tality surfaces from a different perspective t h a n t h e usual graphs of mortality r a t e s a t selected ages over time or a t selected times over age. This change in vantage point can highlight previously obscure p a t t e r n s in t h e interaction of
age, period, . a n d cohort effects, a n d t h u s stimulate a deeper understanding of t h e evolution of mortality over age a n d time and i t s possible social a n d biologi- cal determinants. For example, t h e map of Italian male mortality shown in Fig- u r e 1 reveals a s t r i h n g b u t unremarked p a t t e r n of high mortality for t h e cohorts who were in t h e i r early twenties during World War 1.
CONTOUR MAF' IdETHODS
Contour maps, which are widely used in depicting spatial patterns, can be readily adapted t o r e p r e s e n t any surface t h a t is defined over two dimensions.
Thus, a s illustrated in Vaupel, Gambill and Yashin (1985), contour maps can summarize a variety of age-specific demographic data t h a t vary over time, such as population size and r a t e s of fertility, marriage, divorce, immigration, morbi- dity, and mortality.
Contour maps have been used only occasionally by demographers, perhaps because of t h e computational effort required and because of the lack of detailed d a t a over long s t r e t c h e s of age a n d time. In a pioneering study. Delaporte (1941) published a s e t of contour maps t h a t summarized mortality patterns in several European countries; Federici (1955) directed attention t o Delaporte's work in h e r important survey of demographic methods. Recent advances in com- puters, including the development of powerful micro-computers, and the c ollec- tion and publication of extensive arrays of demographic r a t e s for single years of age and single years of time (e.g., Natale and Bernassola (1973), Vallin (19?3), Heuser (1976) and Veys (1983)), may lead to greater use of demographic contour maps in t h e future.
The contour maps presented in this paper a r e all based on annual mortality rates, , q , t h a t a r e available for single years of age from 0 to 79 and for single years of time from 1870 t o 1979. The data a r e discrete, but a surface is continu- ous: we defined the surface q (z .y ) by linearly interpolating between adjacent d a t a points. The values of q (z,y) give the height of t h e mortality surface over age z a n d time y.
The lines on a contour map connect adjacent points t h a t a r e of equal height; t h e s e lines a r e sometimes called level lines or isograms. In Figure 2, one of t h e level lines r e p r e s e n t s a mortality r a t e of about 11 percent: t h e line s t a r t s in 1870 a t age 28 a n d ends in 1979 a t age 64, indicating t h a t 64-year-old Italian women in r e c e n t years faced the same chance of mortality t h a t 28-year- olds faced about a century ago.
1910 1930 Year
Figure 1. I t a l i a n Male Mortality Rates, n o t smoothed, f o r a g e s 0 t o 79 and years 1870 t o 1979, w i t h contours from .000667 t o -195 a t m u l t i p l e s of 1 . 5 :
An important consideration in designing a contour map is how many dif- f e r e n t levels to use. The computer program t h a t we employed to draw t h e maps, which was developed by Bradley Gambill u n d e r one of o u r (Vaupel's) supervision at Duke University and a t t h e International Institute for Applied Systems Analysis, allows lines t o be drawn a t 15 levels, separating t h e surface into 16 tiers. Use of fewer lines sacrifices detail, whereas use of more lines makes t h e m a p less intelligible: 1 5 levels seems t o be a reasonable compromise, although u s e of 10 o r 20 levels might be considered. Delaporte draws lines a t 19, 20, or 2 1 levels on his various maps of European mortality.
What specific elevations t h e contour lines should connect is a second important design decision. On mortality surfaces, where mortality r a t e s might approach a minimum of t h e order of magnitude of 0.0001 and a maximum of close t o 1, use of equally spaced lines-say a t 0.01, 0.02, a n d so on up t o 0.15-- will result in a map where t h e contours a r e clumped together a t t h e youngest a n d oldest ages, with a largely empty expanse in-between. It is more informa- tive to s p r e a d t h e lines out more evenly. This can be done by placing t h e lines a t c o n s t a n t multiples--e.g., e a c h line representing a level 50 percent higher t h a n t h e previous line. Alternatively, a convenient scale can be used: Dela- porte places his lines at levels of mortality of 1, 2, 3,
...,
9, 10, 12, 15, 20, 30, 50, 100, 150, 200, 250, 300, 350, and 400 per thousand.Demographers often work with transformations such as t h e log or logit, so i t might seem reasonable t o transform t h e surface q (z,y) i n t o t h e surface of, say, log q ( z , y ) a n d t h e n t o draw level lines at equal intervals on t h e transformed surface. If t h e transformation is monotonic, like t h e log or logit transformation, an identical contour map can be drawn by spacing t h e level lines at appropriately unequal intervals on t h e original surface. In t h e case of logarithms, t h e level lines should be at multiples of each o t h e r r a t h e r t h a n being equally spaced.
An innovation i n t h e computer program we used is t h e shading of regions according t o the height of t h e surface, t h e shading varying from light t o dark as t h e surfaces rises from low t o high levels of mortality. Such shading, which is time-consuming t o do by hand but easy with t h e help of a computer, makes t h e overall p a t t e r n of a mortality surface more immediately comprehensible, espe- cially if t h e m a p is viewed at a distance, but t h e details of small peaks a n d pits
Year
F i g u r e 2 . I t a l i a n Female M o r t a l i t y R a t e s , n o t smoothed, f o r a g e s 0 t o 79 and y e a r s 1870 t o 1 9 7 9 , w i t h c o n t o u r s from .000667 t o . 1 9 5 a t m u l t i p l e s of 1 . 5 :
a n d of t h e twists a n d t u r n s of the contours lines a r e still t h e r e t o be scrutinized at close range. Literature, critics note, can be profitably r e a d a t different levels of understanding; we suggest t h e reader view t h e maps in this paper a t t h e two levels of 5 m e t e r s and 25 centimeters.
In Figures 1 a n d 2 some rough black blots a r e smeared across level lines.
These r e p r e s e n t virtual plateaus where t h e mortality surface is repeatedly crossing a n d recrossing a level line. To eliminate this kind of noise a n d t o suppress t h e details of local fluctuations so t h a t global p a t t e r n s c a n be more clearly perceived, i t may be useful t o smooth a surface. Delaporte presented both raw and smoothed contour maps of mortality r a t e s in various European countries: on his "adjusted" maps, Delaporte drew smooth contour lines based on his feeling for the data. We used a mechanistic, computer algorithm t o pro- duce the smoothed maps shown in Figures 3, 4, a n d 5. In each case t h e height of t h e surface a t age z in year y was replaced by t h e average of t h e 25 heights in t h e 5 by 5 square of points from z -2 t o z +2 a n d from y -2 t o y +2. Use of t h i s smoothing procedure c u t s t h e contour map by two years on e a c h edge, so t h a t o u r smoothed maps range from age 2 t o 77 a n d from 1872 t o 1977.
A variety of alternative smoothing procedures might be used, including procedures t h a t replace points by a weighted average of adjacent points, t h e weights diminishing with distance. For our purposes, however, the conceptual simplicity a n d computational convenience of straightforward averaging over a 5 by 5 square outweighed t h e advantages of more elaborate algorithms. We also produced maps, not shown in t h i s paper, based on smoothing over a n 11 by 11 square: on these maps, global p a t t e r n s were somewhat c l e a r e r t h a n on t h e maps smoothed on a 5 by 5 s q u a r e b u t much interesting local detail was lost.
Contour maps can be used n o t only to analyze a mortality surface b u t also t o compare two surfaces by computing a new surface t h a t represents a t every point e i t h e r the difference in heights of t h e two surfaces or t h e r a t i o of t h e height of one surface to t h e o t h e r . Figure 5, for example, depicts t h e ratio of male t o female mortality in Italy.
Year
F i g u r e 3 . I t a l i a n Male M o r t a l i t y Rates, smoothed on a 5 by 5 square f o r a g e s 2 t o 77 and years 1872 t o 1977, w i t h c o n t o u r s from .000667 t o -195 a t m u l t i p l e s of 1 . 5 :
1900 1920 Year
Figure 4. Italian Female Mortdity Rates, smoothed on a 5 by 5 square for ages 2 to 77 and years 1872 to 1977 with contours from .000667 to .195 at multiples of 1.5:
1920 Year
Figure 5. Ratio of Italian Male to Female Mortality Rates, smoothed on a 5 by 5 square, for ages 2 to 77 and years 1872 to 1977, with contours centered on 1.0 at multiples of 1.1:
DATA
The maps in this paper a r e based on life table data for single year cohorts born from 1790 to 1978 (Natale and Bernassola (1973) supplemented by Caselli (forthcoming)). These life tables give the standard probabilities of death, i.e., for each w e t h e probability t h a t a person born in a particular calendar year and alive a t exact age z will die before his or h e r z+1-st birthday. As is well known, this type of probability refers t o events t h a t affect each single year cohort a t each age i n two successive calendar years; for a discussion of this, see Vallin (1973) or Wunsch a n d Termote (1978).
Standard calculations yield period life tables for two consecutive years from a diagonal reading of t h e cohort life tables. From our data, this allowed computation of period life tables from 1869-70 to 1978-79. For convenience in constructing mortality surfaces, we assumed t h a t each * q calculated by this method described the height of t h e mortality surface a t integer age z and at a time y halfway through the two-year period which we took t o be January 1, 1870 for the 1869-70 period and so on. Thus our maps run from January 1, 1870 t o January 1, 1979.
We were able t o check some aspects of the quality of t h e underlying data by scrutinizing the contour maps produced from the data. Wherever there was an island on t h e map-i.e., a local peak or depression on the mortality surface-one of us (Caselli) looked a t historical records and previous demographic analyses to t r y t o account for the abnormality. Based on this research we believe t h a t t h e r e a r e plausible explanations for t h e local patterns shown on the maps, some of which are discussed below, although we realize t h a t plausibility by no means guarantees veracity and t h a t some of the patterns may be artifacts of the pro- cedures used t o collect and process mortality and population statistics.
We also took a close look a t the small black blemishes isc!-,tcZ f r s m contour lines on t h e unsmoothed maps, because these dark spots indicate outliers-very localized peaks or pits--that might be due to erroneous data values. Consider, for instance, the black spot in Figure 1 a t about age 54 and year 1878: it turned out t h a t this blemish was indeed produced by an error made i n transcribing t h e ltalian mortality data t o a computer tape. (The e r r o r was corrected, but we left the spot as a n illustration.) On t h e other hand, the mark a t about age 20 in 1962 represents a point where t h e mortality surface barely crosses a contour
level, like t h e top of a s e a mount t h a t appears as a small island just rising above t h e level of t h e surrounding ocean.
JDCAL PATTERBE AND FLUCTUAZlONS
Figures 1 and 2 p r e s e n t contour maps of Italian male and female mortality r a t e s . The maps, which a r e n o t smoothed and based on d a t a for single y e a r s of age and time, convey a n image of persistent i f somewhat jagged trends with some islands a n d peninsulas of relatively high or low mortality. Some of t h e variability may be essentially random noise a n d some is undoubtedly due t o t h e procedures used t o estimate t h e mortality r a t e s . For t h e major fluctuations, however, i t is possible t o establish meaningful links with external events and developments.
The serious crises i n mortality due t o t h e two World Wars can easily be t r a c e d on t h e maps in t h e dark shafts of high mortality penetrating younger ages. These shafts, especially in the years 1916-1918 and somewhat less dramatically in t h e period 1915-1920, may be i n t e r p r e t e d as a high b u t narrow mountain of mortality t h a t suddenly confronts a population with levels of mor- tality t h a t in normal times would be first e n c o u n t e r e d a t much older ages. Dur- ing t h e s e years during and immediately after World War I, women as well as men experienced risks of death from t h e age of 15 t h a t t h e y would have h a d t o con- f r o n t only from t h e age of 50 or 60 i n t h e absence of war.
For t h e male cohorts most directly involved in t h e First World War, a diago- n a l spur of high mortality indicates serious negative repercussions of t h e war on mortality for a n o t h e r 10 to 15 years. This is n o t surprising in light of a study by Gini a n d Livi (1924): out of t h e five and a half million Italian males who sur- vived military service in World War I, about 85 p e r c e n t were invalids a t t h e end of t h e conflict. A t h i r d were both ill and wounded a n d more t h a n half were ill although n o t wounded.
Women also h a d t o pass through extremely high levels of mortality in t h i s period, although they were less directly involved in t h e war. As Mortara (1925) discusses, t h e epidemic of Spanish in.fluenza in Italy and t h e disastrous economic a n d nutritional conditi.ons resulting from t h e war caused more t h a n 600 thousand female deaths in 1918, compared with a mean of ar0un.d 330
thousand in preceding and subsequent years. This experience appears to have had some negative effect on the survival of the cohorts of females who were young adults a t t h e end of the war, somewhat delaying progress in reducing mortality rates.
Variations in mortality from 1935 to 1948, peaking in the years of the Second World War, are apparent on the maps. The increase in the risk of death is notably less marked than the increase during t h e First World War. World War 11 resulted in about half as many deaths among soldiers as World War I and sub- stantially less debilitation among surviving soldiers. Furthermore, although the civilian population was directly involved in the Second World War, it was decimated neither by the war itself nor by a major epidemic: Spanish influenza resulted in 274 thousand deaths in 1918 alone, whereas less than 150 thousand civilian deaths attributable to the war occurred in t h e entire period 1940-1945.
The variable and often unfavorable mortality rates suffered by young males, especially between ages 20 and 25, can be traced across Figure 1. As analyzed by Mortara (1912), Gini (1932), and Federici (1940, 1941, and 1959), the male mortality curve in some periods not only increases steeply a t these ages but also reaches a local maximum; the reasons include reckless behavior, violent conflicts, and unfavorable Living conditions and life styles resulting from socioeconomic disruptions. Islands and peninsulas denoting prominent local maximums of mortality for males in their early twenties occur around 1870-75, 1880, 1690-95, 1905-12, 1914-25, 1938-48, and 1976-79. From 1920 to 1935 the diagonal spur of high mortality, which, as i n b c a t e d above probably resulted from debilitation during and immediately after the war, implies local maximums in period mortality a t ages increasing from 20-25 in 1920-25 to 30-35 in 1930-35.
The reappearance after 1975 of a prominent local maximum in male mortality rates between ages 20 and 25 is, as discussed by Caselli and Egidi (1984), princi- pally due t o the persistence of high rates of accidental and other violent causes of death among these young males.
PERsISrlENT TENDENCLES
The contour maps in Figures 3 and 4 are smoothed versions of t h e maps shown in Figures 1 and 2; as discussed above, t h e smoothing was done on a five by five square. The smoothing, by reducing random noise and variations of short duration and small magnitude, makes general patterns in t h e evolution of Italian mortality more apparent.
For instance, 3, compared with Figure 1, provides a clearer and more striking impression of t h e effects of t h e two World Wars and t h e lingering cohort effect among t h e cohorts who were in their early twenties during the First World War. An analysis of French.mortality by Vallin (1973) indicates t h a t higher r a t e s of mortality for t h e males most directly affected by World W a r I can be detected in France; a recent study by Horiuchi (1983) indicates t h a t a simi- l a r pattern, centered on t h e birth years 1901 and 1902, appears for Germany a n d Austria but not for Sweden and Japan, two countries t h a t were not deeply involved in t h e First World War. It would be interesting t o display male mortal- ity data from various countries on contour maps to gain visual access to t h e relative strength of this cohort effect in different countries.
Clearly evident in Figures 3 and 4 is t h e persistent progress, interrupted only temporarily by t h e two World Wars, in reducing mortality r a t e s a t all ages but most dramatically before age 50. As discussed by Bellettini (1981) and Del Panta (1984), t h e introduction of public health reforms by the first govern- ments of t h e Kingdom of Italy played an significant role in t h e progress in reducing mortality r a t e s a t the end of t h e 19th century. ' Progress, however, is evident even before these reforms, suggesting the importance of improvements in living conditions and personal hygiene. Ample statistical documentation of these effects can be found in Di Comite (1974).
The slow progress made i n reducing mortality r a t e s among younger women in t h e period up until t h e First World War stands in marked contrast t o the rapid progress in reducing mortality r a t e s for these women aft;er t h e Second World War. The drastic r e c e n t reduction in t h e risk of death from complications of pregnancy and child birth, due, in part, to reductions in fertility, plays a key role in t h e progress made after 1945; earlier progress was also associated with a reduction in t h e risk of death from various infectious diseases such as tubercu- losis.
Among teenagers considerable resistance to a decline in mortality is apparent to t h e very end of the last century for females compared with males.
Biological differences and differences in social roles help account for this.
Reductions in t h e labor-market participation of young boys, which, as Felloni (1961) discusses, was a consequence of social progress following industrializa- tion, allowed an improvement in health conditions t h a t particularly benefitted young boys already active in work outside the household. Female children and teenage girls continued to labor a t t h e demanding tasks of domestic life; under existing hygienic, sanitary, and nutritional conditions, this certainly did not lighten t h e physical stress following menarche.
In addition to the striking progress in reducing female death rates before age 50, Figure 4 reveals systematic if less dramatic trends in reducing female mortality a t older ages. For males, however, mortality rates after age 50 have fallen slowly and even show signs, a t least since 1910, of more or less leveling off. This stagnation, particularly during the last 30 t o 40 years, is discussed a t length by Egidi (1984); related patterns in France a r e described by Vallin (1983) and an increase in mortality rates among Japanese men 35 to 55 years old is documented by Okuro (1981). If the evidence of persistent progress in reducing female mortality were not available, one might be inclined to tbink t h a t little or nothing could be gained in the struggle against death a t older ages. An impor- tant research task is to disentangle t h e lingering effects on male mortality of debilitating events, like the First and Second World Wars, from period effects associated with current patterns of economic activity, environmental condi- tions, and styles of behavior.
In Figure 5 a contour map is presented of the ratio of male to female mor- tality in Italy; t h e map is smoothed on a five by five square so that, as on Fig- ures 3 and 4, each point on i t represents t h e average of the surrounding 25 points. The overall picture, which reveals t h e evolving levels of this classic index of excess male mortality, is strildng but no surprise. The light areas in t h e 19th century in the decade of age between 10 and 20 and again between 30 and 40 depict t h e substantial female disadvantage, discussed above, in adoles- cence and in the later childbearing years: t h e disadvantage is less in the decade from 20 t o 30 because of the elevated male mortality a t these ages.
These light areas are transformed in less t h a n a century into t h e black moun- tain of excess male mortality from ages 10 to 35 and especially ages 15 t o 30, attributable largely t o violent causes of death.
The &agonal swath of excess male mortality running roughly along the life line of t h e cohort born around 1900 reflects t h e sufferings of the males who were children or of fighting age during t h e First World War, young adults during the difficult years between the wars, and 40 to 50 years old a t the end of the Second World War. The black slant of excess mortality around age 60 in 1960 dramatizes the continuing disadvantage faced by the males in this cohort. The decline in relative mortality after this age mag be due to t h e death of the most frail or debilitated members of t h e cohort, i.e., due to selection.
Compared with t h e cohort born in 1900, t h e male cohort born in 1910 experienced less excess mortality a t every age up until age 60 or so. The cohort born in 1920, on the other hand, suffered substantial excess mortality, although less than t h a t suffered by the 1900 cohort, including heavy mortality during the Second World War. This perhaps provides some indirect evidence of some debili- tation occurring during t h e World War 11 among t h e males most directly involved. The relatively favorable experience of t h e 1910 male cohort, and, equivalently, t h e relatively unfavorable experience of t h e 1910 female cohort, deserves further attention.
Finally, it might be noted t h a t t h e males between the ages of 45 and 70 in 1975 experiencing high excess mortality a r e t h e males who were born between 1905 and 1930, who experienced World War I or 11 or both, and who were the main agents of t h e modern e r a of reconstruction and industrial development in Italy.
The black mountain of excess male mortality in middle age in r e c e n t y e a r s may be attributable in p a r t t o t h e s e events a n d hence be a temporary phenomenon r a t h e r t h a n evidence t h a t t h e r e a r e intrinsic, biological reasons preventing t h e decrease in male mortality a f t e r 50 years of age. However, since both male a n d female children born in t h e first decade of t h e twentieth c e n t u r y suffered from t h e privations of t h e First World War during t h e i r period of growth, i t would appear t h a t t h e female population is b e t t e r able t o cope with s u c h adverse cir- cumstances.
CONCLUSION
The contour maps p r e s e n t e d in this paper clearly and efficiently display both persistent global and prominent local patterns of mortality, simultane- ously over age, by period, a n d for cohorts. Italian mortality has been carefully studied by a series of talented and perceptive analysts, so it is no surprise t h a t most of t h e p a t t e r n s a p p a r e n t on t h e contour maps have been previously recog- nized and discussed. The maps highlight some of these p a t t e r n s in such a visu- ally revealing manner, however, t h a t what was understood before c a n now be seen, Furthermore, t h e maps focus attention on some neglected aspects of the evolution of ltalian mortality. Efforts t o b e t t e r understand t h e social and bio- logical determinants of mortality could benefit from more careful a t t e n t i o n t o some of t h e prominent features of t h e contour maps, including t h e following:
-
A series of d a r k islands a n d peninsulas punctuate Figures 1 and 5 over time a t ages of roughly 20 t o 25. These c a n perhaps be explained by reckless behavior, various violent conflicts, and various social a n d economic disrup- tions t h a t afflict men more t h a n women.-
The consequences of World War I a n d t h e accompanying influenza epidemic t h a t raged a r o u n d 1918 were more dramatic in Italy t h a n t h e consequences of World War 11. Especially prominent in Figures 1, 3 and 5 is t h e high mor- tality, during t h e 1920's and early 1930's, among t h e cohort of males who were i n t h e i r early twenties during the First World War. This effect i s prob- ably due to debilitation resulting from illness a n d injury.--
The extraordinarily rapid progress against childhood mortality after t h e Second World War is dramatically displayed in Figures 3 a n d 4. This pro- gress is undoubtedly due t o improved social a n d economic conditions and b e t t e r health care, a s well as t h e beneficial effects of a decreasing birth r a t e , b u t t h e relative importance of t h e s e various factors deserves f u r t h e r study.--
The persistent p a t t e r n of progress in reducing female mortality r a t e s in t h e early adult ages, clearly shown in 4, is perhaps also associated with t h e decline in fertility r a t e s a n d in t h e risk of death in childbearing, as well as t h e virtual elimination of fatal infectious diseases, like tubercu- losis, t h a t took a heavy toll among young adults.Attention t o o t h e r patterns, including t h e slow r a t e of progress in reducing mortality a t older ages among men compared with women, and t h e resumption, after World War 1 a n d again a f t e r World War 11, of t h e fairly regular and per- sistent t r e n d s apparent i n most of t h e contour lines before World War I, might also deepen understanding of t h e social a n d biological determinants of mortal- ity.
Additional contour maps may help demographers in four promising and important directions of r e s e a r c h o n mortality:
First, much can be learned by decomposing mortality r a t e s into cause- specific rates. Contour maps could be used t o display p a t t e r n s in mortality from various causes of death, as well a s t o compare p a t t e r n s from different causes, much as male a n d female p a t t e r n s were compared in Figure 5.
Second, i t would be useful t o compare t h e evolution of Italian mortality with t h e evolution of mortality in various Italian regions a n d in various o t h e r countries, such as France, Belgium, a n d Sweden, for which extensive mortality d a t a is available. Again, contour maps could be profitably .employed in making t h e s e comparisons.
Third, p a t t e r n s of fertility and of nuptiality could be studied using contour maps a n d comparisons could be ma.de of t h e s e patterns with mortality patterns.
Not only do fertility, nuptiality, a n d mortality i n t e r a c t with e a c h other, b u t various social a n d biological factors affect each of t h e s e demographic phenomenon, in differing degrees a n d sometimes in opposite directions.
Comparisons of different kinds of demographic r a t e s could shed light on the importance on both t h e internal and external interactions.
Fourth, models need t o be developed t h a t relate measures of external vari- ables to observed mortality rates. Appropriate models would allow decomposi- tion of mortality p a t t e r n s into t h e effects of (1) aging, (2) persistent progress against mortality over time, (3) shocks a n d disasters like t h e 1918 influenza epi- demic, (4) cohort effects due t o selection resulting from t o heterogeneity in frailty, a n d (5) debilitation of cohorts due t o illness, injury, inadequate nutri- tion, etc. Such models would be valuable n o t only for explanation of past pat- t e r n s b u t also i n forecasting f u t u r e p a t t e r n s and in evaluating policy alterna- tives. Contour maps, by efficiently summarizing many numbers in a compact space a n d by revealing both strong global a n d local mortality p a t t e r n s , can stimulate thinking a b o u t t h e kinds of external variables t h a t may be important.
Furthermore, t h e fit of a model t o t h e d a t a can be displayed by using compara- tive contour maps (like Figure 5); seeing where a model fits poorly may help t h e analyst develop a more a c c u r a t e model.
Tufte, in his lucid exposition of The Irisunl Display of Quantitative Informa- tion (1983), concludes t h a t graphic designs should give "visual access t o t h e subtle a n d difficult, t h a t is, t h e revelation of t h e complex". Demographic surfaces-which may be defined over nearly a c e n t u r y of age and more than a c e n t u r y of time, comprising close t o ten thousand d a t a points t h a t may vary. as in t h e case of mortality surfaces, over four orders of magnitude-are complex;
contour maps a r e a strilung, efficient, and clear means of giving demographers visual access t o them.
Bellettini, k (1981) Aspetti della transizione demografica in ltalia n e l primo period0 post-unitario, in S t u d i in onore d i L u Q i del P a n e , CLUEB, Bologna.
Caselli, G. a n d V. Egidi (1984) La m o r t a l i t & in Italia: e v o l u z i o n e e prob- l e m i a t t u a l i , Materiali di Stu&o e di Ricerche, n. 3, Dipartimento di Scienze Demografiche, ~ n i v e r s i t ; degli Studi di Roma "La Sapienza".
Delaporte, P. (1941) ELuolution de l a n o r t a l i t e ' e n Europe d e p u i s 1 'origine d e s s t a t i s t i q u e s dE l l E t a t c i v i l (Tables d e m o r t d i t ~ de g 6 n ~ r a t i o n s ) , Imprimerie Nationale, Paris.
Del P a n t a , L. (1984) Euoluzione denzografica e p o p o l a m e n t o n e l l ' l t a l i a
dell 'Ottocento (1 794- 1914), CLUEB, Bologna.
Di Comite, L. (1974) La mortalit& in Italia, Istituto di Demografia, Univer- sita di Roma, n. 26, Roma.
Egidi, V. (1984) Trent'anni di evoluzione della mortalit; degli adulti in Italia, Genus, a.
XL,
n. 1-2.Federici, N. (1940) %l rialzo della mortalit; maschile nel quinquennio d i et& 2 0 2 5 anni, Istituto di Statistica della R. Universita, s. I, vol.
N ,
Roma.
Federici,
N.
(1941) Variazioni temporali di un' anomalia della curva di mortalit;, Assicurazioni, a. VIII, luglio-ottobre.Federici, N. (1955) Lezioni d i Demografia, I Edizione, De Santis, Roma.
Federici, N. (1959) L'andamento della mortalit\a giovanile nelle recenti tavole italiane, Proceedings of XV.II1 fiunione della ~ o c i e t ; Italiana d i Statistica, Roma.
Felloni, G. (1961) Poplazione e sviluppo economico della Liguria nel Secolo X7X Collana dell'kchivio Economico dell'unificazione Itali- ana, s. 11, vol. IV, ILTE, Torino.
Gini, C. and L. Livi (1924) Alcuni aspetti delle perdite dell'Esercito ltali- ano, Metron, vol.
TV,
n. 2.Gini, C. (1932) Sul rialzo della mortalit; maschile dai 20 ai 25 anni e sulla minore resistenza vitale dei nati da padri di tale etg, Genesis, a. XI, n. 3-4.
Heuser, R.L. (1976) f i r t i l i t y Tables for Birth Cohorts by Color, United B a t e s , 191 7-1 973, U.S. Department of Health, Education, and Wel- fare, National Center for Health Statistics.
Horiuchi, S. (1983) The Long-tern Impact o f War o n Mortality Old-age Mortality o f First World War Survivors in the Federal Republic of Ger- m a n y , Population Bulletin of t h e United Nations, n. 5.
Mortara, G. (1912) 11 monopolio delle assicurazioni e la mortalit; italiana, Giornale degli Economisti, Gennaio.
Mortara, G. (1925) La salute publica in ltalia durante e dopo l a guema, Laterza, Bari.
Natale, M. and
A
Bernassola (1973) La m o r t a l i t ; p e r c a v s a n e l l e r e g w n i i t a l i a n e . k v o l e p e r c o n t e m p o r a n e i 1965 66 e p e r g e n e r a z i o n i 1790-1969, Istituto di Demografia, Universit; di Roma, n. 25, Roma.
Okoro, S. (1981) h c r e a s e in Mortality of Middle-Aged M d e s in Japan, Nupri Research Paper, Series n. 3, June.
Tufte,
E.R
(1983) m e Visual DLsplay of Quantitative h f o r m a t w n , Gra- phica Press: Cheshire, Connecticut.Veys, D. (1983) Cohort 3 u m i v a l in B e l g i u m in the P a s t 150 Y e a r s , Catholic University of Leuven, Sociological Research Institute: Leuven, Bel- gium.
Vallin, J. (1973)
La
m o ~ t a l i t e / p a r gGn6ration e n f i a n c e , d e p u i s 1899, Tra- vaux e t Documents, Cahier n. 63, Press Universitaires de France, Paris.Vallin, J. (1983) Tendance r6centes de la mortalit; francaise, P o p u l a t i o n , n. 1,
anv vier-fi&ier.
Vaupel, J.W., B.T. Garnbill, A.I. Yashin, and A.J. Bernstein (1985) "Contour Maps of Demographic Surfaces", International Institute for Applied Systems Analysis, Laxenburg, Austria, Forthcoming Working Paper.
Wunsch, G.J. and M.G. Termote (1978) h t r o d u c t i o n to Demographic A n a l y s i s . P r i n c i p l e s a n d Methods, Plenum Press, New York and Lon- don.
