WHAT THE AGE COMPOSITION OF MIGRANTS CAN TELL US

Luis J. Castro* and Andrei Rogers**

Sutvtuany The present paper seeks to identify some of the factors that are responsible for the widespreadregularities in age profiles exhibited by empirical schedulesof migration. It shows how family relationships among migrants are reflected into their aggregate age profiles. By disaggregating migrants into dependentand independent categories, the paper illuminates the ways in which the age profile of migrating populations is sensitive to relative changesin dependencylevels and in rates of natural increaseand mobility. Just as population age compositions reflect particular characteristicsof fertility and mortality rgimes, so do observedmigration age compositions reflect key aspectsof family structureand migration patterns. A framework for assessing impacts of natural increase, the family dependencies, and differing migration propensitiesis proposed.

)
INtRooucnoN A population pyramid graphically displays the age and sex disribution of a population; figure I presentssuch pyramids for Mexico and Sweden. The population of Mexico, with its large fraction of children and small fraction of elderly, D&y be called a "young" population in contrast to Sweden, which clearly exemplifies an "old" population. The age composition of a population retlects the past history of fetility and mortality to which the population has been exposed. For example, high rates of natural increase give rise to age pyramids that taper more rapidly with age, and zero growth rates ultimately produce age pyramids that are nearly rectangular until ages 50 or 60 and that decline rapidly thereafter as death rates increase among the aged. Thus, one may conclude that the age composition of a population tells us something about past patterns of fertility and mortality. Since migrants are a subsetof the population, does their age composition reflect analogous characteristics of recent patterns of fertility, mortality and migration? Figure II sets out the national migratron pyramids tbr Mexico and Sweden. They exhibit a fundamental common finding of countless migration studies: the age composition of migrants reflects age selectivity, with young adults and infants generally being the most mobile group in any population. Migration propensities are high among children, varying from a peak at age I to a low point about age 16. Beyond that age, migration increasessharply to another peak about age 22, after which it declines regularly until possibly interrupted by a retirement peak at the older ages. The present paper seeks to identify some of the factors which could explain the widespreadregularities in age profiles exhibited by empirical schedulesof migration rates. We begin by briefly considering the problem of migration measurement and then go on to adopt a mathematical functional description of migration age composions. Armed with this succinct representationof the age structure of migrants, we go on to examine how differences in family status patterns structure the age profile of migrants. EsTngITs}ilNGTHEREGULARITIES: MIGRATION MEASUREMENT Migration studies have in the past exhibited a curiously ambivalent position with regard to the measurementof geographical mobility. This ambivalenceis particularly striking because of the contrast it poses with respect to the conesponding studies of mortality and fertility, studies that are richly endowed with detailed discussions of measurement problems. Haenszel (1967) attributes this paradox to the strong influence of Ravenstein's early contributions tc migration analysis: "Work on migratron and population redistribution appears to have been strongly influenced by the early successesof Ravenstein in formulating 'laws of migration'. Subsequentpapers have placed a premium on the development and testing of new hypothesesrather than on descriptions of facts and their collation. . . This is in contrast tu the history of vital statistics. While Graunt, more than two centuries before Ravenstein, had made several important generalizations from the study of 'bills of mortality' in London, his successors continued to concentrate on descriptions of the forces of mortality and natality by means of rates based on populations at risk" (Haenszel, 1967:260).

,I

*Former Research Scholar, Human Settlements and Services, International Institute for Applied Systems Analysis, Laxenburg, Austria. **Chairman, Human Settlementsand Services, International Institute for Applied Systems Analysis, Laxenburs. Austria.

63

Figure I.

National population age comlDsitions: Mexico, 1970, and sweden. 1974

Mexico,1970
Age 90 80 70 60 50 40 30 20 10 0
812

L
(percentage)

128

(percentage)

Sweder, 1974
Age 90
80 70 60 50 40 30 20 10 0

16l2g4O

(percentage)
sources: Federal Statist-icalotfice, rr70; ndersson and Hormberg, 19g0.

Figure rI.

Nationar migration age compositions: Mexico, 1970, and sweden. 1974

A fnterctate migraon, Mexi@,lgl}

Age
80 70 60 50 40 30 20

J
r2B (percentage)

f0 0 8t2 (percentage)

B- Intenegionaf mignaon, swede,1974

Age
80 70 60

50 40 30 20 t0 0

16rzBi (Rercenage

0 4 8 n -r

sources: lo'ne per cent sample of the t97 Mexican population census; Andersson and Eolmberg,

r r;o- ^_ (Fercentage)
lgso.

65

It is nanral to look to the state of mortality and fertility measurementfor guidance in developing measuresof migration. Like mortality, migration may be described as a process of interstate transfer; however, death can occur but orce, whereas migration is a potentially repetitive event. This suggeststhe adoption of a fertility analogue, that is, instead of births per mother, moves per migrant; but migration's definitional dependenceon spatial boundaries and on different forms of data collection introduces measurementdifficulties that do not occur in the analysis of fertility. One of the central problems in migration measurement arises as a consequenceof the different sources of migration data. Most informatiori regarding migration is obtained from population censusesorpopulation registers that report migration data, for a given time interval, in terms of counts of migrants or of moves, respectively. Yet another source of migration data is the sample survey, which may be designed to provide information about both migrants and moves. Mgration data produced by censusesare usually in the form of transitions. Population registers treat migration as an event and generatedata on moves. A mover is an individual who has made a move at least once during a given interval . A migrant, on the other hand, is an individual who at the end of a given interval no loger inhabits the samecommunity of residence as at the start of the interval. Thus, paradoxically, a multiple mover can be a nonmigrant, if after moving several times he returns to his inial place of residence before the end of the unit time interval. Because migraon occurs over time as well as across space, studies of its patterns must trace its occurrence with respect to a time interval, as well as over a system of geographical areas. In general, the longer the time interval, the larger will be the number of return movers and non-sunriving migrants and, hence, the more the count of migrants will understate the number of interregional movers (and, of course, also of moves). Most migration data collected by population censuses come fromresponsesto fourtypical questions:place of birttr, duration of re.sidence, place of last residence and place of residenceat a fixed prior date (United Nations ,l97O). From these quesons it is possible to establish the count of surviving migrants living in a region at the time of the census, disaggregated by different retrospective time intervals. The longer the time interval, the less accuratebecomesthe migraon measure. Because population registers focus on moves and ngt transitions, differences will arise between data obtained from rcgisters and from population censuses. In the annex to the United Nations manual on Metlnds of Measuring Interno.l Migration (United Nations, 1970) it is stated: "Since at least some migrants, by census definition, will have been involved, by registration definition, in mor than one migratory event, counts from registers should normally exceed those from oensuses. . . . Only with Japanesedata has it so far been possible to test the correspondence between migrations, as registered during a one-year perid and migrants enumerated in the oensus in erms of xed-period change of residence" (United Nations, 190:50).

oN By TnsI-e I . CorrlpnrusoxoF MrcRATroN sEx ANDTypEBASED THEPoPUBETWEEN FORTT{EONE-YEAR PERTOD LATIONREGISTER AND THE CENSUS Ocrosen 1959 ll.Io Ocrosen 1960. JprN
Sex and type of migration Register dota Census data Ratio x I0O

Both s*es lnta-prefcctural Intqprcfectural

29621 2625 t35

1998 l7l 2 59075t 1001745 I l6 898 99642ft l 123853

14.47 r01.33 148.63 9t.90 148.30 104.4e

Malcs I ,f88 935 lntra-prefcctural .. Intcrprcfectural . . : : : : : : I 450 El? Fenuls Intra-prcfectural Interp'refectural | 477 86 | 1743r8

Sourcc:UnitodNaons(ln0, table42:f)).

Table 1, takenfrom the United Nations analysis,illusmigraondatais in trateshow the ratio of register-to-census generalbigger than unity, increasingwith decreasing distance, as, for example, in the case of intra- venusinterprefectural migration in Japan. In general, the ratio of register-to-census mrgrationclatasnould tend to unity as longer distancesare involved, and also as time intervals become shorter (figue III). Clearly, the rao should be greaterthan unity when short distances considered and are the close to unity when the time interval is short, because probability of moving acrosslong distancesseveraltimes shouldbe expected be lessthanthe probability of moving to And, the the samenumberof timesbetweenshortdistances. probabillty of moving several timesduring a long interval of time shouldbe greaterthan the probability of experiencing the samenumberof movesduring a shorterperid of time. A tundamentat aspect ngrationis its change overtme. of As Ryder (1964) has pointed out for the caseof fertility, periodandcohortreproduction rateswill differ whenever the age distribuon of child-bearingvaries from one cohort to another.Theusefulness a cohortapproach migration,as in of in fertility analysis, in theimportance historicalexpees of rience as an explanation of current behaviour. Morrison (1970)indicates migration inducedbytransitions that is from one stageof the life cycle to another, and "chronic" migrantsmay artificially inflate the migrationratesof origin areasthat are heavily populatedwith migration-prone individuals. Both influenceson period migration are readily assessed a cohortanalysis. by It is the migraonof a period, however,illd not that of a cohort, that determinesthe suddenredistribuon of .a national population in responseto economicfluctuations, and it is information on period migration that is neededto calculatespatialpopulationprojections. Currentperiod migrationindicesdo not distinguishtrend from fluctuaon and therefore may be distorted; current cohort migration indices are incomplete. Thus it may be useful to draw on Ryder's (1964 translation technique to changefrom one to the other. As Keyfitz (1977:25O) observes,the cohort and period momentsin Ryder's formulae can 'be interpreted,not as child-bearing,but as mortality, marriage,school attendance, income, or some otherathibuteof individuals". Migrationis clearlysuchan attribute.

Theimpctance of historicalexperience interpretingand factors, such as occupation, educational athinment and in understndingcurrentmigrationbehaviourled PeterMorriincome level. Finally, experientialstagingrefers to moveson (190:9) to definethe notion of stagingas being "any mentexperience termsof numberof previousmovesand in linkage betweena prior sequence subseguent and migraon durationsincethe lastmove;it is the '?arity" dimension of bchaq". Morrison recognizes forr kinds of staging:geo- migrationanalysis. graphic, life cycle, socio-cconomic Calculationsof migration rates of increasingspecificity ad experiential.Geographical stagingrpfcrs to rhrn migration and to what is seekto unconfound "tne" migrationratesfrom weights the conventionallyunderstood mean "stage migration", that that reflect the arithmeticalinfluenceof the past. This proto is, th idea at migrantstend to move to placesnot very cessof measuring migration dissimilar from those they left behind. Life-cycle staging ". . . at differentlevelsof specificityof occunence and views migrationasarisingout of breaksin anindividual's or cxposuryielrls productswhich draw ever finer dtstinca household's cycle, suchasentry into the labourforce, life tions betrreencurrent behavior and the residue of past marriage and rotirement. Socio-oconomic staging sees behavior reflected in the exposuredistribution at any migrationscqucnoes bcingconditioned socio-stnrctural as by tim" (Rydcr195:10). Suchproductsmaybe weighted aggregated produce and to the "cnde" rates higherlevels aggreganon. of of ForexamIturc m. Rro of rcabcr !o ccrn n[retbn drtr ple, the agc-scx-specificmigration rate is a weighted ttlh rlpcct to dllrncc Dd dn hicrvd aggregation with respect the migration "parity-duration" to distribution just as the crude migration rate is a weighted regisb/cen$s ratio aggregation with respectto the age-sex stribution. Theageprofile of a schedule migrationratesreflectsthe of influencesof two agedistributions:the agecompositionof migrantsandthatof thepopulationof which they wereapart (Rogers,lnq. This canbe easilydemonstrated decomby posing the numeratorand denominatorof the fraction that definesan age-specific migrationrate,M (x), sy. lf O (x) denotesthe number of out-migrantsof agex, leaving a region with a populaonof K (x) at that age, then M (x) : where O : total numberof out-migrants N (x) : proportion of migrants agedr years at the time of migration K : total population x C (x) : proportionof total populationaged yearsat mid-year o : cnrdeout-migrationrate We definethecollectionofN (.r) valuesto bethemigration proportion scheduleand the set of M(x) values tq be the migration rate schedule. Suuuanaxc rHE REGI LARrrrEs: MODEL MIGRATION SCHEDULES Observedage-specificmigration rate schedules universally exhibit a conunonshape(Rogersand Caso, l98l). The sameshapealso cbaracterizes age compositionof the migrants, that is, migration proportion schedules. Starting with relatively high levels druing the early childhoodtgs, migration ratesor proportionsdecrease monotonicallyto a low point atagex,, sy, increase until theyreacha high peak at age,r, aodthen decrease oneeagainto the agesof retirementbeforeleavingoff arond someconstantlevel, c, ey. Occasionallya "post-labour force" componentappears, o (x) : K (x) N (x) o'N (x) : o. K 'C (x) C (x) (l)

dis;hnce

reglsb/censsratio

meinbrual
6l

showi4geithera be[-shaped curyewith a retirementpeakat agexor anupwardslopethatincrpases mondonically to the last age includedin the schedule,age lr, say. Thus, the migrationageprofile maybe divided into child (dependenD, adultancetcerlyaomponents; however,we shallconfineour atlention in the presentpaperto only the first two. But our argumentis equally valid for profiles strowinga retirement peakor an upwardretirementslooe. The observod distributionof migrants (x), may be age ,N described a rdction of the form: by N(x):Nr6)+N2@)+c where
Nt @) 41 e-at x

Tru
Clwaraistic

2. hnrcpr, nucns DEFTNTNc AcB-spoFtc oBsERvBD MIORATION OTARACTEruSTTCS

hoportion of childe, (dcpcndants), Qt Proportion of adultsa, Q (labour

allal

force)

fra2la7

r @2t\)

Labourasymmtry, az I-abourdominancc,62 Parental-shift, Br2 Ctild-adult dependency migration ratio,Do

\21a2

(2)

a7la2 I

fa the child (dependent) component,

NZ 6) : a2 e-%k tlz ) - e-x26 ttz)

Thc reciprocal indexis alsoof interest inasmuch it reflects Otal as e number migrants adult, = of per s, ;#. stope; an p2 positions the adult component on the age axis. Taken as a group, these parameters suggest a number of useful and robust measures describing an observedmigrafor tion sclrcdule (table 2). For example, the ratio Do : lh, the child-adult dependencymigration ratio, is one of several important ratios that may be used to interpret particular patternsof dependencyamong migrants. It assumes central a role as an indicator of family dependencystructure by definittg the number of dependantsper adult migrant. The child-adult dependencymigratron ratio vanes as a function of the parameters that define the age profile of migrants. If the constantterm c is closeenoughto zero to be ignored, as normally is the case, then {" : 0 and

rbr the alt (independent) component, c is the constant and term that improvesthe fit when migration distributionsat olderages relativelyhigh. FigureIV illustratesthe female are model migraonproportion schedules the observed of data presented figure II, which by definition show an areaof in unity under eachcurve. An anernaveway of expressingequatron(2) is as a weightedlinear combinationof the densityfunctionsrepr senng the above three components(Castro and Rogers, l98l): N (x): h fi6) + hfz@ + c(Ilw) (3) wherew is thelastageincludedin the schedule,tand @2 are tbc relative shares the child and adult components, is of " tbc shareof e constanttsrm,f,(x) andfr(i) u", respectivcly, tbc single and doubleexnnential dcnsity funcons h @) : er e-at x fz @) # e-azk (4)

Do:t:ffi
Since

- Fz)- e-\26 t2) (5)

l(ql\r):W
we obtain the result
Do:

andI (o.2llQ reprcsents gamma the functionvalueof a2l\2. Note that , + z * " : I by definition. Equations(2) through(5) imply that
4I:,q,

() (7)

9,, 6r,I l + I lo)

(e)

and

c- h_
w

(8)

The six parameerse1, a2, e, ! and pz do not seem e, to have demographic inerpretations.Both c anda2 reflect the heights of their respective parts of the profile; a1 and a2 refer to the descending slopes; 12 reflects the ascending

where r : qla Fn = a,b (r2: Itrlq are te laborr dominanc, plrritll-shift and labour asymmetry indexes defined in Rogersand Castro (1981). Thesethree ratios and pzmay be usedto fully characterizeobservedmigration age profiles. Another useful indicator of the average size of family among migrants is the value so : I lS2, which reflects the total number of migrants per adult. In a single-sexformulation, for instance, if adults are considered as headsof each migrant family (interpreting single individuals as one-person families) then the sum of the two sex-specific values of so closely approximates the average size of family among migrants.

6E

IISUrc IV.

Componcntsof ihc nodel mlgrrtbn proportbn schcdute

A Inbrstate migraon, Mexi@,1970

-x
o ct) 0.03 (E
+, (tr

o E 0.02
o-

-,.2 k-p2) a2 e

^t ('-Pz) "

e e

0.01

0 ) 20.0 30.0 40.0 50.0 60.0

Age
0.05

B. Intenegional migraon, Swede, 1974

Females

s
>i o

0.04

TE o

CD (u 0.03

o-

e o.o2 e
0.01 0 30 40

Age
69

Figure V.

Model migration proportion schedules for selected cities around the world and a typical nationl immigration flow

0.05

> = x o

0.04

P o.o3
(E c

o E g o.o2
o-

0.01

0.06

0.05 -x { 0.O{ o o o
(E

E 0.03 E

I p

o.o2

0.01

l1
l

o CD (o
+. |E

c o

o-

e e

Figure V

(continuedl

Grabr Kharhm

s
e

ofH

)t

$ ooo t
.9

E oo2

3046-

Stockholm
>o(E
S-

x' P o.o4 .rt

o
.E E F

E000

lmmlgratbnb Kuvmit

s
*
{rr

ofH

$ on.

.E

Eo.oz
o,

f, e

0.01

Alberts, l!77.; gnepe_r:gg!!-t'gpte the 197-9 of polrllationcensus Mexico;Gorge of antlEigbfoh, . - --Sources.' 1973;KawabeandFarah,tgz: nr-orsso-al- Hhnberg,lffiiG-ltioris,lgTg.

7l

Trt

3.

pRoFoRTroN PlteMsrgRs AND vARIABLES DEFTNTNc oBsERvED MoDELMIGRATToN scHEDr,JLEs sELEcTED FoR crrrEs AnouND THEwoRLD ANDA TyprcAL NATToNAL rMMrcRATroN o# n
llcxico City Famls Grcatcr Klutaut Stltholm Innigration to Kuwait

Rb dc taao Pomq vtb

a1 a1 a2 ,L2 a2 lU

0.0:t o.(Xl 0.075 0.oy2 0.051 0.037 15.394 14.905 0.()69 0.059 0.6A 0.583

0.065 0.101 0.u27 15.765 0./0 1.107

0.048 0.081 0.02r 13.878 0.060 t.23

0.u2 0.099 o.2 t7.223 0.w2 0.291

0.03E 0.065 0.0,18 rr.ffi 0.098 0.318

0.031 0.099 0.054 15.053

.oog
0.531

0.025 0.116 0.088 rc.ns 0.098 0.$t

0.023 0.095 0.096 19.648 0.110 0.334

0.018 0.056 0.086 19.65r 0.108 0.229

0.u27 0.047 0.084 t9.% 0.161 0.301

a ,?--f @y'Xrr;l fiowrc, ot rtciUu did pai3 b atathis rrtictiotr. HGtrcc, u somcprotcr vluc p 'ightly birsd" d in rucb iDtcaa6c abovccqudig 4jri.lt n, W. e ndru r frcjcr fa4cnr to O oacvar. -! Dl c prllclcr vdc w6 3e1 equl to zem rn tlre non-tea paarnetcr stimarionpfocedur.

NqrE Ou .ssumplionth'r c = 0 contirs the paramcters satisrythe rclationshipflto

Figure V setsout sveralageprofiles of internalmigration flows to different cities around the world, togetherwith a qpical internationalmigration (immigration) age distribuprofiles weregenerated tion for malesandfemales.These by modbl migration proportion schedulesfitted to observed data, the parameters which are included in table 3. The of quantitative indices presentedin Able 4 confirm the regularities illustratedin gure V. For example,the migration flows to Mexico City and to Lagos differ sharplyfrom the oresllonding flows to Stoclolm. The former show about doublethe proportionof dependants exhibitedby the latter. Tbe sametable alsoindicatesthat the average sizeof family in theflow-toMexicoCity, with about2.65+ 2.20: 4.85 membersper migrating family, is the largest among the erarqtles _p_ resented . All of the migrationcharacteristics figure V andtable4 in indicatelow or high family dependency patterns.In the next secon,we sekan explanationfor suchcharacteristics by linking themwith the family characteristics the population of as a whole.

EXpITuNc TTTE REGULARITIES : FAMILY STATUS It is widely recognized that a large fraction of total migration is accounted for by individuals whose moves are dependent on those of others. Indeed, family migration is such a well-established phenomenon that Ryder (1978) has even suggested its use as a criterion for identifying family membership: a family comprises those individuals who world migrate together. To understand th influences that family and dependency relationshipshave on migration age compositions, it is useful to examine how such profiles respond to fundamental changesin dependencypatterns. To illustrate this, consider a single-sexpopulation that is divided into two groups: dependants and heads, where dependantsare simply individuals who have not left home to becomeheads. (Included as heads are independent single individuals who may be viewed as one-personfamilies.) Thus, the agedistribution of the female population C (x) may be composed by weighting the density

TrI.E 4.

EsnrrrsD cHARAgrERIsrtcs oF oBsERvEDMoDELMIcRATIoNx)puLATtoN scrrEDULEs FoR SELEcTED cITtEs AROIJND WORLD THE ANDA TYPICAL NATIONAL IMMIGRATION FI.oW
'Rb

-auu-cj

dc Julcior

I%t-rn2
Clwrctesc
Fciule llal

I%tr-t970
Fctml Malc

Lagosc 1967-IX,E
Fcnal

ciur xtuitnd l%t0-r%4

Tdol Malc

stnl,l-l^' Iy74
Frrulc

Innigruioato Kuxyait I%5-IY/O' Aab Fdrulc

hoportion of dependants (pcrecntage) hopction of adults (percentage) Total numberof migrants per adult Iborr asymrnetry Ibour dominancc Parcnal shift Gild-adult migrationratio

33.70 70.14 1.43 10.03 2.U2 l.0B 0.4t

4.8t 59.54 1.68 9.83 0.90 1.56 0.75

&.29 37.67 2.65 t5.74 0.42 t.43 t.7l

59.50 45.55 2.20 2t.0t 0.58 r.34 l .3l

42.28 59.90 r.67 3.20 t.47 1.08 o.7l

57.80 43.88 2.28 3.24

31.60 73.77 1.36 7.73

24.07 78.06 t.28 3.04 4.18 0.87 0.31

21.31 82.17 t.22 5.r8 3.55 l.l8 0.26

3t.6 70.9 1.42 2.t2 4.88 0.52 0.45

57.09 6.n 2. t 6 1. 87 3. t 2 0.29 r.23

t.n o.67 r.32

r.72
1.4 0.43

Swces: :Alberts, 197. bnc pcr cent sampleof the 19/0 Mexicanpopulationoensus. p*t!" andEigbeioh,ln3. Tkwabe andFarah,193. lAnderssonand HoLnberg,1980. Unitpd Nations, 1919.

72

functions of dependantsand heads: C (x) : t"ft"@) + r"fr,(x) where r" and r, *" the proportions of dependants and headsin the total female population andJ" @) andfr,(x) ue their correspondingage distributions, respectivelv. The ratio of tne weignts associatedwitn tne age profiles of dependants and heads defines the child-adult dependency population ratio, D", which is similar to theQ defined earlier for the migrant population:
n -t, ucr"

As in the case of migration, we can also detine the total number of personsper adult (head) ffi r" : I l2". To invesgate analytically some of the underlying patterns of "head formation " requires some mathematical theorizing. Irt yr'.denotethe age at which an appreciable number of female3irst leave home to establish their own household. Since mariage is an imporant reason for leaving the family home, it is likely that the probability density function describing the pattern of head formation by age is similar to the one found in studies of nuptiality, that is, the double exponential function defined in equation (5). Ifgly) is such a function then

Figure VI illustrates the above argument with hypothetical data. It presentsthe survivorship curve, I (x) , which is that of the Brass standardwith a : -0.80 and B : 1.75 with an expectation of life at birth of approximately 9 years (Brass, l97l); and the head formation curve G (x) is the CoaleMcNeil double exponential (Coale and McNeil, 1972) expressed by the Rodriguez and Trussell (1980) standard with mean (22years'and variance (5 years) of age of becoming a head. Figure VII shows the resulting dependant,head and population (dependantsplus heads) distributions of stable populations growing at intrinsic ratesr : 0 and r : 0.03, respectively. To derive the corresponding age compositions of migrants we inhoduce the probabilitiesp,(x) andpr(x) that a dependant and a head, respectively, migrate at agex in an interval of time. The age distribution of migrants is defined as before:

N (x) : tfi @) + f, (x)

where

f,(x):
[,

e-rx I (x)l,I - G (x\pt @)

e-rYI (Y) tt - G (Y)] Pt 0) dY

and

G@):l s0)dy tro

defines the proportion of females who have ever left home by agex, that is, who are heads according to our definition. Since fr" (x) defines the proportion of the population of headsthat are of agex , and G f) defines the proportion of the population who are heads by age , it is evident that in a stable population growing at an intrinsic rate of growth r, _ fz" 6) _e-rx I (x) G (x)

f^

fr (x) :

Jo

e-'Y I (y) G (y) pz 0) dy

The child-dependencymigration ratio Do, equivalent to equation (9), may now be defined as:

[,
Do :

e-u I (y) t - G (v)lpt 0) dy -'Y

where / l) denotes the probability of surviving from birth to agex. For similar reasons

T
@

e-a I 0 G (y) dy

I:,

I (y) G (y) pz 0) dy

r I e-'Y I (y) I G (il) dy J^ 0 populaGiven theseequations,the child-adult dependency tion ratio D" may be definedas
I

ft" @) :

e-rxI(x)I -G(x)I

I e-rY (y) II - G (y)l dy -n'I

Both child-adult dependency ratios, D" and Do, may be anatysed6y using hlpothetical populations once again. To spetify correctly the probabilities pt @) and p, l) from different sources of migraon data, it is necessaryto identify first the number of moves a person undertakes during a unit interval. However, for our purposes we may assume that both dependants and heads follow a negative exponential propensity to migrate with respectto age, with the function's Barameter reflecting the average rate of moving per unit of time. Formally, we have then
pt @) : 6, -ox

Dr:-*

TC

0) G (y) dy

and
PZ Q,Yd : o2 e-oz$ - Yo)

73

Figure VI.

Proportion surviving to age x, I (x), and proportion of individuals who have ever left home by age x, G (x)

r.0 0 0

0.900

0.800

0.700

0.600 * o o (tr
(E

0.500

o-

e e

0.400

0.300

0.200

0.t00

0.

where yo denotes, as before, the age at which an appreciable number of females first leave home to establish their own household, ord o/ and o, denote the average rates of moving per unit of time of dependantsad heads, respectively. One might expect that the average rate of moving per unit of time for dependants,o, should not exceedor,the corresponding rate for heads. In general, dependants(children) move with their parents and independent, single individuals are most likely to be found among adults. Figure VIII presentsthe vaiation of Dowith respectto D. for the hypothetical populations of figure VII, under various mobility conditions as expressedby the ratio o , lo 2.lt may be seenthat the ratioD olD,rnoreclosely approaches unity asthe migration of heads increases. The parameters dening the mobity conditions may be used to set out a typology of migration profiles that helps to identify how a particular family migration pattern may be

reflectedin a migrationagecomposition,andhow important are amongheadsand dependants the migrationpropensities in stncturing that agecomposition.FiguresIX andX present a set of profiles classifiedaccordingto two distinctly different rates of natural increase.For each of the hypothetical populationswe show threealternativecombinations proof pesitiesto migrate among headsand dependants. First, (o2 figure IX setsout, for low headmigrationpropensities : of 0.08),profiles showing significant a degree family migraon (o, : or) and also of low family dependency = @t 0.10o2and o, : 0.20o). In a similar format, gure X presents corresponding profiles for high headmigration the (o, propensities -- 0.16).lVith theaid of these figureswe two can seethat patternssuch as thoseof Stockolmindicate a with high head relatively low family migration dependency migration propensitiesand low population growth rates, profiles suchasthoseof Mexico City presentcharwhereas 74

proportionof headsat agex'f2$)' and the ase FigurevII. Proportionof dependanls.at x,fiq@), growth r of zaroend 0'03' respectively populationagecomposition,C (i,o: niinsc rates of resulting

$ o.o,
o

x- 0.o4
o

E
I

I
6

oo2

g o.os

&

F,,
fomilmso

As
oto20Oo5(r60 Age

a $ ooo I I oo3
o6 c

20 olozo3Om50 Age

30

5(, AgP

to acteristicsthat correspond high family migrafon.depenand .rd atively'high de,pendant head migraon d;t propensities.

illuminatesthe waysrn which the ageprotile tion schedules, to of .igation is sensitive relativechanges-in.dependency Viewing and levelsandin ratesof naturalincrease mobty' within a frameworkof depeldentand the migrationprocess allows one to observethat if the rirovements indepe-ndent componentmainll comprisessingle.persons' indeirendent Coxct ustoNs migrationmaybeinsignificant dependent tfr"n'tfr" associate shareof the total migration.pn the in terms of its relativl The aim of the presentpaperhas beento show how the family ,tfr.t tt"", if migfationtendsto consistprincipally of can agecompositions be that .pptut in migiation regularities maybecome children of tie share dependent what such ;;"t,'tt ,o-r-rrirrd in iseful mannerand to suggest "n natural - n-"ty importantpart of total migration' regularities may be telling us about patterns of " OUr"*in migon disributins, when analyse! in tlrc family relationshipsand mobity levels among ;;,confirm theindicaons approach, contextof thefamily status -- migrants. model proportion gi""o by the puruto"t"ts of ttr associated dir"ggr"gation of migranr*intg depe$ent andindepenwere migrationdependencies high icheduies.Fr example, dentcatei, andtheadoptionof modelmigrationpropof75

Figure VIII.

Variation of child-adult dependency ratios among migrants (Do) an the population (Ds) with respect to different levels of natural increase (r), family migration (o7lo2\ and migration propensities of heads (o2)

2.00 1.90 1.80 1.70 l.0 1.50 1.40 1.30 1.20 l.l0 { J.(X'

r = o.o3 o2 = o.16 r =Q o2 = o.16 r = o.o3 o2 = o.og

-o
o2 = o.o8

Qo o.eo
0,80 0.70 0.0 0.50 0.40 0.30 0.20 0.10 0. 0. 10.0 20.0 30.0 40.0 50.0 0.0 70.0 90.0 100.0

or lo2 (percent)
correctly indicated for Mexico City; for Stockholm they were low; and falling somewherein between these two extremes was the case of Rio de Janeiro. The degree of propensity to migrate among independent migrants is also evident from observedageprofiles. Srongly skewed distributions in the adult ages,correspondingto high tr, and a2 parareter values, indicate relatively higher migration propensities for the independent component. Profiles wittr higb dependency levels show rnuch more weakly skewed adult migration compositions due to lower propensies for individual moves among heads. Justaspopulation agecompositionsreflect particular characteristics of fertility and mortality rgimes, so do observed migration age compositions reflect key aspects of family structure and migration patterns. Although many of the relationships set out in the present paper are still conjectural, a modest start has been made. A framework for assessingthe impacts of natural increase, family dependencies and differing migration propensities has been set out. The arguments set out in the present paper are related to a number of earlier efforts by the authors to examine regulaities in age patternsof migration. For example, our focus has been on a single-sex formulation. However, it appears that differences between the age composition of migrants may be a consequenceof differences in sex-dependency structures. To study these relationships, a matrix approach has been recently proposed in Castro and Rogers (1983). Causesof migration are related to a person's age and sex. For example, migration motivated by health reasons is a phenomenon characteristic of old people, whereas educa-

76

ngurc lX.
olo

mlgrrrbn dhrlburlon for br rd hbb populrrbo gmrrh, .l,ry'ry9q f ":depcndcncig frdry nErrbn rd br hd nlrtbo ploiniltlcr
rO lspn5oog:orrtr Ltrttrdrdglrmffiy, or.Ot 0r.O.6. O.Ot6tr5 O00t

ooe

lwpgongulfr,r 'O low hcd nfgraon fp.nty, or.Ofit o.O.08, O.OIE tnd 006
OD

0.6

OO

o.g,

q@

*n*
x

toG

OO

OG

En'
oqt

F
oof
q6

ot 'oo16 orr o'c

o@

oc oo

oo

20

0 Aet

to

50

to

b. towffiydrrdct

oo

X mCongutrru O.CB low hrrd nfg[on prop.fry. or.O& ot.Ooo, O.OIO d Om

or0 OD oc
g.g,

Xeogffongwdr,r rO.Gl Lwhd nnonplup.ry, or.OOO ot.OOO 0.016 md 0003

0.6

o@

$o* oc

g o0c
I
ol. o2. 0(D

'
OG

l*.

o0a o@ o@ oo

ot 'ool8 or'ofi

ogt

ol o.o o. or0 c. F.tfngr!cl

0S|o50O Ag.

or ' o'oo8 or. OOB

oto
&tOOl0O Aot

d. Lwffiydpndrv

77

Figure X. of migration distributions for different populationgrowth, { typgloqv. : family migrntion 1Be dependencies high headmigration piofensities and
l-ovvpoptlaliongroudh,r = O High fnad migraonpropensity, or=0.16; or=O.16, 0.O32 and 0.016 Lourpoprlation grorarlh, = O r Hghheadmgration properu*ty, or=0.16; o, =0.16, O.O32 and O.016

* 0.06 t o CD
o c 0.05

$o*
= or= Q.16

R o.*
o-

F"*
It
0.

f; o.os

x'

Age b. Lowfamilydependency

2A

30 Age

40

50

60

High poprhliongrornrlfr, = O.O3 r High headmQrationpropensity. or=O.16; ol=0.16,0.032 and 0.016

Hi$t pogdation grorndfr,r= 0O3 Hagh hd rrfrdliorl proprx*V or-0.16; o, =O.16, 0.o32 and 0o16

S o.o.
x'

= or= 0.16

E E

$ b.os
o'*

o, = O.O16 o, = 0.16

o, = O.O32 02 = 0'16

20

30 Age

40

50

60

78

tion-related migration is predominantly associated with young people. Thus, in order to understandbetter why people move, it is important to disaggregate cause-specific migration databy age and by sex. This "mortality" analogy is followed in Rogersand Castro(l98la), where it is shown that it is possible to accountfor someof the differencesin age patterns of migration by linking them to differences in their underlying cause-specificstructureS. Drawing on techniques used in the corresponding literature in fertility and mortality, Rogers and Castro (l98lb) proposed proceduresfor adopting model migraon schedules

to infer migrationpatterns the absence accurate in of migration data.Suchmodelmigrationschedules may be usedto graduateinadequate data, therebysmoothingout irregularitiesandascribing the datasummary to measures can that be usedfor comparative analysis. They alsomay be usedto interpolate to single years of age observedreliability of empirical migrationdata, and indicationsof appropriate strategies their correctionare aidedby the availability of for standard families of migration schedules. Finally, such schedules alsobe usedto helpresolve problems may caused by missingdata.

Rrsncss
Alberts, J. Migracin hacia reas metropolitanas de Amrica Latina (Migration in t atin American Meropolitan Aeas). Rio de Janeiro, Centro Latinoamericano de Demografia, 1977. Andcrsson, A. E., and I. Holmberg . Migration and Scttlcnvnt: 3 . Swcdcn. Laxenburg, AusEia, International lnstitute for Applied Systems Analysis, 19t0. RR-80-5. Brass, W. On the scaleof mortality. Biological Aspectsof Demography, W . Brass, ed. London, Taylor and Francis Ltd.,lnl. pp. 69-110. Castro, L. J., and A. Rogers. Model Migration Schedules: A Simplified Formulation and an Altertutive Parameter Estination Method. I-axenburg, Austria, International Institute for Applied Systems Analysis,

1981. wP-8r-63.
Casro, L. J., and A. Rogers. Patterns of family migraon: two methodological approaches.Environnent and Planning ^r{,vol. 15, 1983. Coale, A. J., and D. R. McNeil. The distribution by age of the frequency of first marriage in a female cohort. Journal of the Amcrican Statistical Association 67 :7 43-749, 1972. Federal Stastical Office. D( Censo General de Poblacin (Ninth General Population Census). Mexico City, 1970. George, M. V., and A. A. Eigbefoh. Population growth and migration in Lagos, l9l l-1963. Urbanization and Migration in Some Arab and Alrican Countries. Cairo, Cairo Demqgraphic Center, 1973. Haenszel, W. Concept, measurement, and data in migration analysis. Demograplry 4:253-261, 1967. Kawabe, H., and A. A. M. Farah. An ecological study of Greater Khartoum. Urbanizption and Migration in Some Arab and African Countries. Cairo, Cairo Demographic Center, 1973. Keyfitz, N. Applied Mathenatical Demography. New York, Wiley,lW .

Morrison, P. M. Implicationsof Migration Histories Model Design. for Santa Monica,California,The RandCorporation,lnO. p. 4342, Rodriguez,G., and J. Trusse.Mcimumkelihood Estination of the Paratneters Coale'sModel NuptialitySchedule of from Suley Data. World Fertility Survey,TechnicalBulletin 7, Tech. 1261. Voorburg, Netherlands, InternationalStasticalInstitute, 1980. A.Two Methodological Notes Spatial Rogers, on in Population Dynamics the SovietUnion. laxenburg, Austria, InternationalInstiote fq Apped Sysrcms Analysis,196. RM-7648. Rogers,A., and L. J. Castro.Age patrcrnsof migraon:cause-specic profiles.IIASA Rcporrs l2!l-l0, l9Ela. 4: Ryder, N. B. The processof demographic transition. Demograplry, l:7482,1964. Fertility measurement through cross-sectional surveys.Social Forces,54l.7-35,l15. Methodsin Measuringthe Family Life Cycle. Proceedings the of InternationalPopulaonConference, InternationalUnion fm the Scienfic Study of Populaon lnt. , Soboleva, MigrationandSettlernent: Soviet S. Ilon.I-axenburg, Aus8. tria, InternationalInstitute for Applied SystemsAnalysis, 19E0. RR-8G36. United Naons. Department Economicand Social Affairs. Methodsof of Measuring ltertul Migration. PopulationStues, No.47. (ST/SOA/ Series A/47) Sale.s E.70.XIU.3. No. Tres and Clnracteristicsof lternartonal Migration Since1950 . DemographicStuesNo. 64. 19/9. (ST/ESA/SER.A/64) SalesNo. 8.7t.Xm.5.

79

CONTENTS

Recenttrendsand conditionsof fertility United Nations Secretariat Fertility and family structure NormanB. Ryder Populationdistribuonpolicies Harry W. Richardson. Metropolitanmigrationandpopulationgrowthin seleoted developing countries,lfbG 1970 United Nations Secretariat What the agecompositionof migrantscantell us Luis J. Castroand Andrei Rogers The long-term impact of war on mortality: old-age mortality of the First World lVar survivorsin the FederalRepublicof Germany
Shiro Horiuchi

15

80

ST/ESA/SER.N/15

DEPART]IIENT OF INTERNATIONAL ECONOMIC AND SOCIAL AFFAIRS

POPULATION BULLETI N OF THE UNITED NATIONS

No. 15 -1983

rat

WH*T*J?J*,-'