A parity-specific analysis of completed fertility in Canada“

N A T A L I E K Y R I A Z I S Concordia University

Ce texte remet en question I’hypotMse statique des prises de decision dans le domaine de la fertilitC, qu’on retrouve dans les modeles microiconomiques. Cette hypothese est remplacee ici par un cadre siquentiel qui permet d’obtenir des effets diffCrents pour des sequences natales diffCrentes. I1 est donc possible de faire une analyse s6parCe de chaque progression paritaire en utilisant une mCthode de rCgression logique. Vu l’interac- tion entre certaines des variables indkpendantes et la religion, les analyses ont CtC ex6cutCes skparCment pour les protestants et les catholiques. Les donnCes utilisCes dans cette Ctude ont CtP puisCes dans 1”Cchantillon d‘usage public d ioo ’ du recensement du Canada de 1971. L’echantillon comprend des femmes mariCes une seule fois, vivant tou- jours avec leur mari, et nCes entre 1902 et 1936. Les rCsultats demontrent que le revenu du mari a un effet positif sur les deux parites les plus basses chez les protestants et sur les trois parites les plus basses chez les catholiques, tandis que l’effet sur les paritCs plus ClevCes dans les deux groupes est nkgatif. Les autres variables comprises dans le modele, soit l’fge des Cpoux au mariage, le niveau d’instruction de l’Cpouse, le lieu de rCsidence et le groupe de naissance, dCmontrent aussi des inCgalitCs paritaires qui appuyent l’approche skquentielle utilisee dans ce texte.

This paper questions the static assumption of micro-economic models of fertility deci- sion malung and uses instead a sequential framework that allows for different effects at different birth orders. As a result, a separate analysis of each parity progression is conducted using the logit regression method. Due to the interaction between some of the independent variables and religion, separate analyses for Protestants and Catholics are carried out. The data utilized in this study were obtained from the dioo Public Use Sample of the 1971 Canadian census. The sample is comprised of married-once, spouse- present women born between 1902 and 1936. The findings show that husband‘s income has a positive effect at the lower two parities for Protestants and lower three panties for Catholics, and negative effects at higher panties for both groups. The other variables included in the model, age at marriage, wife‘s education, place of residence and birth cohort also show variation by parity supporting the sequential approach utilized in this paper. * I am indebted to Michael Hout for his advice regarding the method utilized in this

paper. I also wish to thank Costas Ranos and the two anonymous referees for their helpful comments. This paper was received September 1979, and accepted October 1980.

Rev. canad. SOC. 8r AnthKanad. Rev. SOC. 8r Anth. 19(1) 1982

30 NATALIE K Y R I A Z I S

Microeconomic models of marital fertility are generally cast in a static decision- making framework.= As a result, they focus on total children ever born as the relevant dependent variable (e.g., Becker and Lewis, 1974, De Tray, 1974). Several researchers have challenged this static assumption of micoeconomic models and have proposed instead a dynamic approach to fertility decision making (Nam- boodiri, 1972; Rozenzweig, 1976; Hout, 1978; Seiver, 1978). Accordingly, they have conducted separate analyses at different parities. Their results indicate that the effects of economic and non-economic variables on fertility vary considerably over successive parities, refuting the existence of constant effects implicit in the one-period optimization approach. In a recent study of Canadian period fertility using 1971 census data, I similarly showed that the relationship between explan- atory variables and fertility differs by parity (Kyriazis, 1979). The sample was comprised of younger women so that the effects of the independent variables were interpreted as influences on the timing of fertility rather than ultimate family size. In this study, the target population is women 35 to 69 years who have for the most part completed their fertility; as a result, it is possible to isolate the determinants of number changes unconfounded by timing shifts, except insofar as timing decisions voluntarily or involuntarily affect ultimate family size. Since the microeconomic framework utilized in this paper has been presented in the previous article, only a brief review will be attempted here.

Microeconomic theory treats number of children as a function of three main factors: income, costs, and tastes. According to this theory, couples attempt to maximize their utility by choosing the optimum combination of children and other goods, given the economic resources available, the relative prices they face, and their preferences for children as opposed to other goods. Consistent with this framework, the model utilized in this study incorporates these three factors as determinants of fertility.

The major postulate of microeconomic theory is a positive effect of income on fertility. Analyses of this relationship, however, particularly those based on individual-level data, have produced contradictory results. Simon (1975a) suggests that this discrepancy might be due to the variation of income effects on completed fertility by parity. This variation, according to Simon, is masked when the dependent variable is total children ever born, producing results that are highly dependent on the birth order distribution of the sample. The major interest of this paper is to determine whether there is in fact a different relationship between income and fertility by parity, or whether the positive income effect postulated by microeconomic theory is constant, thus confirming the existence of an equilibrium family size.

In addition to income, the relative cost of children constitutes another economic constraint on fertility decision making. The higher the cost of children, the lower the number of children the cou le decided to have. Costs include not only the

time devoted to child-related activities, i.e., opportunity costs. Since the wife is assumed to spend more time in household activities than the husband, Mincer (1962,1963) argues that the price of time should be measured in terms of wife‘s potential earnings in the market place. The greater the wife’s potential income, the greater the cost of children, and consequently, the lower the demand for children. In this study, education is utilized as an indicator of the wife’s market earnings

price of goods and services direct Y y incurred in child bearing, but also the value of

31 A N A L Y S I S O F C O M P L E T E D F E R T I L I T Y I N C A N A D A

potential. It is understood that education has an effect not only on the wages that a woman can command in the market place, but also on tastes for other goods relative to children (Hawthorn, 1970; Turchi, 1975; Namboodiri, 1972)’ efficiency in household production (De Tray, 1974; Simon, 1974)~ and contraceptive effectiveness and usage (Michael, 1974; Turchi, 1975). Parity-specific effects of education, however, parallel closely the effects of wife’s potential earnings, suggesting that education is primarily an indicator of opportunity costs in fertility decision malung (Kyriazis, 1979). Based on this finding we assume that the main effect of education as it relates to fertility is on the wife’s earning potential, and hence the opportunity costs of children.

Rural-urban place of residence is included in the model to account for differences in tastes for children. Although Becker (1960) and others of the Chicago school do not explicitly incorporate tastes in their analyses, exclusion of this variable given a correlation between income and tastes might bias the income-fertility relationship (Thornton, 1978).

Two additional variables included in the model are age at marriage and birth cohort. Age at marriage eliminates differences in length of exposure to child bearing; the possible correlation of this variable with husband‘s income and wife’s education necessitates its inclusion in the model. The effect of birth cohort reflects such contextual influences as changes in attitudes toward family planning, technological improvements in methods of contraception, and changing trends in labour force opportunities for men and women that may have a differential impact on fertility. It should be noted, however, that cohort effects cannot be separated from aging and period effects due to the linear dependence between these variables (Mason et al., 1973). Caution should therefore be exercised when interpreting the cohort coefficients. In order to eliminate interaction effects between the exogenous variables included in the model and religion, and to facilitate comparison with previous results (Kyriazis, 1979). the model is estimated separately for Catholic and Protestant women.

THE M E T H O D

The dynamic perspective adopted in this paper dictates the use of a parity-specific model. Separate analyses are therefore conducted at each parity. The dependent variable is a dichotomous variable where o is assigned to women at party i and I to women progressing from parity i to parity i + I. Goldberger (1964: 249) indicates that the use of ordinary least-squares regression when dealing with dichotomous dependent variables is inappropriate. Because the homoscedasticity assumption is violated, the coefficients are ’statistically inefficient.‘ In order to eliminate this problem, Goldberger and others (e.g., Theil, 1970; Goodman, 1976; Hout, 1978) suggest that a logit transformation of the dependent variable be carried out. The dependent variable thus becomes the natural logarithm of the odds of progressing to parity i + 1 to remaining at parity i.

In structural equation form, the relationship between this parity progression logit (Li) and the independent variables in the model may be represented as follows: Lij = boi + bJj+ bziCj + b3iCt + b4iAj + bsiRj + u

32 N A T A L I E K Y R I A Z I S

where,

I = husband’s income C= wife’s birth cohort A= wife’s age at marriage R= place of residence

u = error term and j = an index representing the cells of the joint distribution of all variables

Due to the curvilinear relationship between cohort and fertility, a cohort squared term is included in the equation.

The method of estimation utilized in the present analysis is weighted least squares (Grizzle et al., 1969; Theil, 1970). Given that the main problem with dichotomous dependent variables is the ’systematic relationship that appears between the frequency of positive responses and the values of the determining factors’ (Theil, 1970: 105), the weighted least-squares method eliminates this problem by assigning more weight to those cells with the larger number of observations. Significance tests for the effect of each independent variable are conducted using the residual sum of squares. In the weighted least-squares pro- cedure, the residual sum of squares is distributed as chi-square with degrees of freedom equal to the number of cells minus the number of variables in the equa- tion. To determine the significance of a single variable, we run regressions in- cluding and excluding this variable and then subtract the two residual sums of squares. For example, to test the significance of husband’s income, we subtract the residual sum of squares obtained from a regression that includes husband’s income from the residual sum of squares from a regression that does not. The difference is distributed as chi-square with one degree of freedom.

DATA A N D O P E R A T I O N A L I Z A T I O N

Data The data for this study were obtained from the d i o o Public Use Sample of the 1971 Canadian census (Statistics Canada, 1975). Only married-once, spouse- present women, 35-69 years on I June 1971 were included in the analysis. Although some women over 34 bear children, the proportion is extremely small (Collishaw, 1976) so that the findings should not be significantly affected. As a result of the selection criteria utilized, 10,470 Protestant women and 10,094 Catholic women were included in the analysis.

Operationalization The variables utilized in this study were operationalized in the following ways. Parity progression logits. The computation of parity progression logits was based on reports of children ever born. For example, the log odds of progressing from parity o to parity I were established on the basis of the number of women who had at least one child ever born to the number of women who had o children ever born. In other words, women with one child ever born were included in the numerator of Lo, in the denominator of L,, and excluded from the computation of L,, L,, and L4.

33 ANALYSIS O F COMPLETED FERTILITY I N C A N A D A

Due to the small number of women with more than five children, the highest order progression analyzed was parity progression 4. Income. Most of the studies conducted within a microeconomic framework use ’permanent’ income as an indicator of economic resources (e.g., Willis, 1974; Re Tray, 1974). This refers to the couple’s perception of the average flow of resources over the life cycle. It has been argued that the use of current income underestimates the income effect, due to fluctuations that may characterize one’s income stream (Willis, 1974; Gardner, 1974; Simon, 1974). As a result, projected income at 40 is viewed as a more appropriate measure because it holds age effects constant and also because it reflects husband’s peak earnings potential, a more relevant income constraint with regard to fertility decision making (Easterlin, 1969; Bean and Wood, 1974; Willis, 1974; Hout, 1978). For compartative purposes, both current income and projected income at 40 are utilized as alernative measures in the analysis. The current income measure is husband’s total reported earnings, 1970, coded in hundred-dollar intervals. Husband’s income at 40 is estimated on the basis of education, place of residence, language spoken at home, industry, region, and age. A more detailed description of the procedure used to compute this variable is presented in Kyriazis (1979). Age at marriage. This variable is reported directly in the census. In order to facilitate the analysis, five categories were constructed and the mean for each category was used as the score. This scoring procedure was also used for wife’s education and birth cohort. The categories for age at marriage are I / 14 to 19 years, 2 / 20 to 21 years, 3 / 22 to 23 years, 4 / 24 to 25 years, and 5 / 26 to 34 years. Wife‘s education. This variable was coded into three categories representing the number of years of schooling completed. The categories are I / o to 11 years, 2 / 1 2

years, 3 / 13 or more years. Birth cohort. The women’s age at the time of the census was used to compute birth cohort. Seven five-year intervals were coded into the following categories:

6 / 1927-1931; and 7 / 1932-1936. Place of residence. The remaining independent variable, place of residence, was coded as a dummy variable where I represents rural residence.

1 / 1902-1906; 2 / 1907-1911; 3 / 1912-1916; 4 / 1917-1921; 5 / 192245126;

T R E N D S I N C O H O R T FERTILITY

The parity progression logits by cohort and religion are shown in Table I. Protestants show an increase in the log odds on first and second births by more than 80 per cent. The log odds on higher order births follow a curvilinear pattern with a net increase in the log odds on third births of 47 per cent and a net decrease in the log odds on fourth and fifth births of 76 per cent and 67 per cent, respectively. Relative to the 1927-1931 birth cohort, Protestants born between 1932 and 1936 show a decline in one child families as well as families with three children or more. The trend is toward a narrowing of the distribution and convergence toward a modal two-child family size among Canadian Protestants.

The log odds on first births show an increase of 77 per cent for Catholics from the 1902-6 birth cohort to 1932-6 birth cohort. A net decline is manifested, however,

TABL

E I

PABI

TY P

RO

GR

ESS

ION

LOG

ITS

FOR

ON

CE

-MA

RR

IED

, SP

OU

SE-

PRE

SEN

T W

OM

EN

BY

BIRTH C

OH

OR

T A

ND

R

EU

GIO

N (

CA

NA

DA

197

1)

Birt

h Cohort-Protestant/Catholic

Pant

y pro

gres

sion

Log

ets

1902

-6

1907

-11

191 2

-16

I91 7

-21

1922

-6

1927

-31

1932

-6

Firs

t bir

ths

1.82

7l1.

820

1.84

6/1.

990

1.95

9/1.

9O3

1.9W

2.22

4 2.

O44

/2.3

06

2.34

0/2.

619

2.30

6/2.

356

Seco

nd b

irth

s 1.

311/

2.18

1 1.

46V

2.12

4 1.

5W1.

740

1.61

5/2.

O03

1.

811/

2.18

5 1.

9W2.

138

2.16

912.

024

Third

bir

ths

.524

/1.5

36

.497

/1.4

86

.447

/1.2

82

5.48

A.3

29

.631

/1.2

94

.813

/1.2

11

Four

th b

irth

s .4

50/1

.818

.3

75/1

.206

.2

28/1

.167

.2

36/1

.130

.2

91/

39

9

.314

/ .780

.108

/ .5

66

Fifth

bir

ths

.458

/1.1

47

.425

/1.2

96

.223

/ .7

27

.112

/ ,9

03

.034

/ .7

54

.104

/ .6

70

.160

/ .3

25

Num

ber

of W

omen

66

3/48

7 93

7179

2 13

94/1

111

1656

/134

3 20

07/1

898

1928

1218

1 18

85/2

282

SOUR

CF.:

Sta

tistic

s C

anad

a, 19

71 Census of

Can

ada,

Publ

ic Use

Sam

ple

.n2

/ .a

96

35 A N A L Y S I S O F C O M P L E T E D FERTILITY IN C A N A D A

in the log odds on all higher order births. The considerable decline in fertility for the 1932-6 birth cohort is primarily a result of a reduction in families with more than two children although the log odds on first and second births also show a decline. In comparison to Protestants, the probabilities for Catholics of having first and second births are quite similar; it is the greater tendency of Catholics to have higher order births which accounts for the differential fertility between the two groups. It is evident, however, that both Catholic and Protestant younger women display a sharp drop in the probability of progressing from parity 2 to higher order parities, suggesting that both groups are moving toward a normative two-child family size.

What factors account for these trends in fertility? In the next section we examine the effects of economic and non-economic variables on each parity progression logit in an attempt to determine constraints operating on births of different orders. Given that the future of population depends largely on the transition from a two- to three-child family size, it is especially important to establish what effects changes in relevant economic and non-economic variables will have on the decision to have a third child.

THE D E T E R M I N A N T S O F PARITY P R O G R E S S I O N LOGITS

The Effect of h o m e The model was estimated separately utilizing husband’s current income and husband’s income at 40 as two alternative income measures. Table 11 presents the results from the equation including husband’s current income for both Catholics and Protestants. In the Protestant sample, the effect of current income is positive and significant only on parity progression 0-1. The coefficients remain positive up until parity progression 2-3 but revert to negative for parity progressions 3-4 and 4-5+. For Catholics, the effects are consistently positive although the magnitude of the coefficients is greater for the lower three parity progressions.

For reasons already discussed in this paper, the use of actual income is not considered appropriate when testing the microeconomic theory of fertility behaviour. It is simply used here as a bench mark with which to compare husband’s income at 40. Table 111 presents the results from the analysis of the model including husband’s projected income at 40, a measure considered more relevant to fertility decision-making. For Protestants, the effect of husbands income at 40 is generally stronger relative to that of current income. This indicates that a ’permanent’ income concept is indeed more relevant to fertility behaviour for Protestants. In contrast to the ’consumer choice’ model, however, which implicitly postulates a consistently positive income effect at every parity, the effect of husband’s income at 40 is positive on first and second births and negative on all higher order births. The substitution of husband’s projected income for current income in the model actually strengthens the negative income effects at higher parities in the Protestant sample. An examination of the results for Catholics reveals that the effect of husband’s projected income is stronger relative to the effect of current income only on parity progression 0-1. The effects on all higher order parity progressions are weaker. The most interesting finding with regard to income, however, is that, as in the case of Protestants, the use of projected

~ ~

~~

~ ~

~~

~ ~~

~~

~~

~~

~~

TABL

E II

MET

RIC

CO

EFFI

CIE

NTS

FR

OM

TH

E R

EG

RE

SSIO

N O

F P

AR

ITY

PRO

GR

ESS

ION

LO

GIT

S O

N T

HE

EX

OG

EN

OU

S V

AR

IABL

ES,

INC

LU

DIN

G H

USB

AN

DS'

C

UR

RE

NT

IN

CO

ME

, 19

70,

FOR

WE

D-O

NC

E,

SPO

USE

-PR

ESE

NT

WO

ME

N,

BY R

EL

IGIO

N [

CA

NA

DA

197

1)

Znde

pend

ent

Var

ibles

Pari

ty P

rogr

essi

ons -

Pro

test

ant/C

atho

lic

0-1

1-2

2-3

3-4

4-5+

Age

of M

arri

age

-.105

"/-,0

99"

-.063

'/-

,089

' - .

058"

/ - .07

4"

-.04Y /-.068'

-.070

' /-.064'

Wife

's Ed

ucat

ion

-.087

"/-.1

32"

-.W

/-,1

38"

- .06

1"/-

. 08

7"

- ,048" /

- ,0

39

-.053

/-

,056

' C

ohor

t - .

075b

/-. 0

28

- .041'/-

.102

" - .

055'

/-.

043'

-.031

/-

.071

b -.0

56'

/-.0

7gb

Coh

or?

.002

b/

.001

.0

01b/

.0

02b

.001"/ .

ooO

ol

.oO

03 /-

.001

.oO01 /

,0

01'

Hus

band

s Cur

rent

Inco

me

1970

.0

07'/

.005

,0

03 /

.014

b ,0

03 /

,006

-.

0005

/ ,001

.004

/-

,002

Plac

e of R

esid

ence

(rur

al =

1)

.208

/ ,1

12

.280

'/ ,5

33"

.548

"/

.776

' ,4

85"

1 ,5

56'

.808"

/ ,3

67'

Con

stan

t 5.

738

/ 5.

839

3.52

2 /

5.63

3 2.

689

/ 4.

200

2.27

2 /4

.059

3.

622

/ 3.

188

p s

.001

bp

s .01

p S .0

5 SO

UR

CE

: See

Tab

le I

~ _

__

_

__

_

~~~

__

_

~_

__

_

__

_

~ _

__

~

~

TABL

E XI

1 M

ETRI

C C

OEF

FIC

IEN

TS

FRO

M T

HE

RE

GR

ESS

ION

OF

PARI

TY

PRO

GR

ESS

ION

LO

GIT

S O

N

TH

E E

XO

GE

NO

US

VA

RIA

BLES

, IN

CL

UD

ING

H

USB

AN

D'S

PR

OJE

CTE

D

INC

OM

E

FOR

MA

RR

IED

-ON

CE

, SP

OU

SE-P

RE

SEN

T W

OM

EN

, B

Y R

EL

IGIO

N:

IN C

AN

AD

A 1

971.

Inde

pend

ent

Var

iabl

es

Par

ty Progressions-Protestantlcatholic

0-1

1-2

2-3

3-4

4-5+

Age

at M

arri

age

- .10

7"/-

.106

" - .

062'/

- .0

85"

- .0

54'/-

,0

69'

- .05

1" I-

,068

' - .0

64'/-

,0

67"

Plac

e of

Res

iden

ce

.579

I ,9

23'

.324

I

,554

,3

93'1

,643

" .5

92b/

.51

8"

.122

I ,8

11'

Wife

's Ed

ucat

ion

- .0

98"/

-. 19

8"

-.032

I-

.095

-.0

36 I

-,050

- .0

6ObI

- ,0

30

-.043

/-.

w

Coh

ort

- .0

61'/-

.0

36

-.031

I-.0

59'

-.039

b/-

,019

- .0

35'/-

.0

67"

-.081

"/-.0

43'

Coh

ort'

.001

b/ ,001

.001

b/ ,001'

.001

b/-.

m

.m/ ,

001

.001

b/-.0

001

Hus

band

s Pr

ojec

ted

Inco

me

at 40

.014

'/ .0

26'

,003

I ,011

-.00

1/

,001

-.001

/-.001

-.008

I-,0

03

Con

stan

t 4.

962

14.7

34

3.26

9 14

.749

2.

599

13.8

71

2.21

6 14

.058

3.

663

13.2

99

p 6

.00

1

bp s .0

1 p 6 .0

5 SO

UR

CE

: Se

e Ta

ble

1

38 NATALIE KYRIAZIS

income also yields negative effects at higher parities for Catholics. These income effects suggest that among Protestants there is decreasing marginal utility from an additional child after a two-child family size has been attained so that a given increment in income produces substitution effects on fertility at higher birth orders. The substitution effects in the Catholic sample appear only at parity progressions 3-4 and 4-5 + . Assuming that an increase in income will be invested in an additional child only if that child does not exceed the normative family size range, these findings suggest that the upper family size limit is greater for Catholics than it is for Protestants. This assumption is supported by the findings in Table I which show a greater probability for all Catholic cohorts to progress beyond parity 2 .

The income effects obtained in this study are similar to those found in an analysis of parity progressions of U.S. women conducted by Simon (i975b). He also found positive income effects only at lower parities. These results support the sequential approach adopted in this paper which necessitates a separate analysis of each parity progression. Use of completed fertility as the dependent variable leads to an averaging out of effects, the outcome of which depends on the strength of the coefficients at each parity. The findings suggest that an increase in income induces child bearing only within the normative family size range; beyond this range the utility of additional children decreases so that an increase in income is devoted to alternative goods. The difference in income effects by birth order therefore refutes the assumption of an equilibrium family size implicit in microeconomic models. An unravelling of income effects conducted within a sequential framework is essential for an adequate understanding of the income-fertility relationship.

Other Independent Variables The substitution of husband’s projected income for husbands’ current income in the equation has not substantially affected the other parameters. A comparison of Tables II and 111 indicates that the pattern of effects is very similar in the two models. As a result of this, and given the fact that projected income is considered a more appropriate measure, discussion of the effects of the other independent variables included in the model will be restricted to those presented in Table 111, i. e., results from the equation including projected income.

Table 111 indicates that wife’s education has a negative impact on every parity progression for both Protestants and Catholics. The strongest effect for both groups is on first births, suggesting that more highly educated women show a greater degree of hesitation in starting a family than that associated with subsequent births. As was indicated in the discussion of the model, we assume that the wife’s education as it relates to krtility most strongly influences the wife‘s earnings capacity. As a result, education is utilized as an indicator of the opportunity costs of child bearing. The consistently negative effects of education provide support for microeconomic theory’s postulate of an inverse relationship between the value of the mother’s time and child bearing. A comparison of the two groups indicates that the education effect is generally greater for Catholics, a finding consistent with the results from the previous study of Canadian period fertility (Kyriazis, 1979). In that study it was suggested that the substitutability between time at home and other activities might be lower for Catholic women

39 ANALYSIS OF COMPLETED FERTILITY I N C A N A D A

given the more traditional orientation of this group with regard to familial responsibilities.

The relationship between age at marriage and fertility is, as expected, negative at all parity progressions for both groups. The effects on the log odds of first births for Protestants are substantially greater, however, relative to the effects on higher order births. The explanation usually offered for the negative relationship between age at marriage and completed fertility is that a woman who marries early is more likely to bear higher order children than a woman with a later age at marriage. The influence on children ever born, however, is primarily due to the effect of age at marriage on the likelihood of a first birth. The difference in the probability of a first birth for a woman who marries in her early twenties relative to a woman who marries in her thirties is considerably greater than the difference in the probability of a fifth birth given that a fourth birth has already occurred. Although the probability that a woman who marries in her early twenties will end up having five children is considerably greater than that of a woman who marries in her thirties, the findings indicate that this overall effect might be due to the large difference in the probability of a first birth. These findings are consistent with those obtained by Hout (1977) in his analysis of completed fertility in the U.S.

The other exogenous variables included in the model show a fluctuating pattern of effects. The impact of rural place of residence is generally stronger for Catholics, with the largest effects ocurring at the first and fifth parties. The biggest difference between Protestant women living in rural and non-rural areas is in the probability of a fourth birth. In both groups place of residence does not significantly affect likelihood of a second birth. The pattern of cohort effects confirms the trends that have been documented by others (e.g., Long, 1970; Henripin and LCgarC, 1971). The positive cohort terms and the negative cohort squared terms for Protestants are consistent with the U-shaped pattern in cohort fertility characteristic of this group. Total fertility rates for Protestants were generally high but declining in the 1920s. They reached a low point during the Depression, and then started to climb again during the war, increasing sharply in post-war ‘baby boom.’ Since the late I ~ ~ o ’ s , fertility among Protestants has been steadily declining. By contrast, Catholic fertility in Canada deviates sharply from the U-shaped pattern that has been observed for Protestants. Catholics experienced a decline in family size until the 1907-11 birth cohort, but in contrast to Protestants who manifest a continuous upturn in fertility for those born between 1912-31, Catholics did not register any increase in family size. The ’baby-boom‘ period for Catholics represents simply a temporary halt in the process of fertility decline that resumed again in the late 1950s. During this period there was an increase in two-child families but the considerable reduction in Catholic fertility at higher birth orders cancelled the increase in family size at this parity. The cohort effects for Catholics shown in Table 111 confirm this historical trend in Catholic fertility. The cohort coefficients are consistently negative, but in contrast to Protestants the cohort squared coefficients are irregular in sign and significant only at parity progression 1-2.

This is consistent with the increase in fertility experienced by Catholics at this parity during the ‘baby boom’ period. In other words, a curvilinear relationship between cohort and fertility exists for Catholics at the second parity, while at higher parities there has been a continuous decline in fertility.

40 NATALIE K Y R I A Z I S

D I S C U S S I O N

The microeconomic theory of family formation is a rigorous analytical scheme that has stimulated a great deal of research, but at the same time, provoked extensive theoretical criticism. Most of this criticism focuses on the assumptions that economists make with regard to fertility behaviour. One of these assumptions, specifically that related to the time reference of the child-bearing decision was challenged in the present research. In contrast to the static approach of microeconomic theory, fertility decision making was viewed as a sequential process - a function of the couple’s changing social and economic situation. Consistent with his sequential approach, a separate analysis of each parity progression was conducted.

The findings indicate that the relationship between the independent variables and family size differs by birth order, refuting the economists’ assumption of constant effects, and supporting instead a sequential approach to fertility decision making. The effect of current income is positive only on first and second births while negative at higher birth orders for Protestants. In the Catholic sample, the influence of current income is positive at all parity progressions, although stronger on the first three births. Substituting husband’s projected income at 40 in the equation, a more relevant income measure produces negative effects beyond the second child for Protestants and beyond the third child for Catholics. Hence, changes in income constrain family size to three children for Catholics and two children for Protestants. In contrast to the consistently positive income effects postulated by economists, the findings suggest that there is increasing substitution of other goods for children at higher parities. This substitution at higher parities was also evident in the previous analysis of Canadian period fertility (Kyriazis, 1979). Such differences in the relationship between income and family size are lost in the static framework employed by economists. In utilizing such a static approach, we can not determine the threshold fertility level beyond which there is declining marginal utility associated with an additional birth.

The other variables included in the model display similar variation by parity progression. The influence of wife’s education is stronger at lower parities for both groups. The magnitude of the effect is generally greater for Catholics, however, suggesting that there may be greater incompatibility between child bearing and other activities for this group due to its more traditional orientation. Age at marriage similarly has stronger effects at lower parities for both Catholics and Protestants. This finding suggests that the strong negative association between age at marriage and births of higher order commonly found in other studies may be due to the accumulated effect of age at marriage on lower order births. The cohort effects are compatible with the curvilinear trend that has been documented for Protestants and the consistently declining trend for Catholics. The only significant curvilinear relationship for Catholics is at parity progression 1-2 consistent with the increase in two-child families experienced by Catholics during the ‘baby-boom‘ period.

In short, the analyses of parity progressions for Protestants as well as Catholics underscore the importance of using a dynamic approach to family formation. For both groups, a wide variation of effects was produced that clearly reveals the

41 ANALYSIS O F C O M P L E T E D FERTILITY I N C A N A D A

inappropriateness of a static framework in fertility analysis. The most interesting findings of this study are the negative effects of income at higher birth orders. These negative effects were interpreted as a result of decreasing marginal utility derived from children relative to other goods. Further research in this area should take into account expenditures on goods that may compete with children as an additional explanatory variable. The substitution between other goods and children inferred in this study should be made explicit by including in the model the amount of resources devoted to such goods.

Finally, it may be argued that the present study consitutes an inadequate test of the microeconomic theory of fertility due to its exclusion of child quality. Becker (1960) argues that income has a positive impact of both the quantity and quality of children but that the quantity elasticity is small relative to the quality elasticity. As defined by Becker and Tomes (1976), child quality is comprised of parental expenditures on children as well as an ’endowed’ component of such factors as inherited ability and ‘luck. ‘ The subjective nature of these elements, however, precludes the measurement of this ’endowed’ component of child quality. On the other hand, it becomes possible to explain a wide range of findings on the basis of the unknown values of this component. In other words, child quality facilitates ex-post description but does not enhance the predictive ability of the model. Unless child quality is defined in such a way so as to become operational, it cannot be utilized as a justification for the inconsistent results produced by studies using the microeconomic theory of family formation.

N O T E

I The pioneering work in this area is by Becker (1960). Other related references are De Tray (1974). Michael (i974), Ben Porath (1974), Gardner (1974), Becker and Tomes (1976), Willis (1974). These comprise the ’new home economics’ or ‘consumer choice model’ approach to fertility decision making. When referring to economists in this paper, we are referring to this shoo1 of thought.

R E F E R E N C E S

Bean, F.D. and C.H. Wood 1974 ’Ethnic variations in the relationship between income and fertility. ‘ Demography

Becker, G. 1960 ‘An economic analysis of fertility.’ In National Bureau of Economic Research (ed.),

11: 629-40

Demographic and Economic Change in Developed Countries. Princeton: Prince- ton University Press

Becker, G.S., and H.G. Lewis 1974 ’On the interaction between the quantity and quality of children.’ Pp. 81-90 in

Becker, G.S. and N. Tomes 1976 ‘Child endowments and the quality and quantity of children.’ Journal of Political

T. W. Schultz (ed.), Economics of the Family. Chicago: University of Chicago Press

Economy 84: S1434162.

42 N A T A L I E K Y R I A Z I S

Ben-Porath, Y. 1974 ’Economic analysis of fertility in Israel.‘ Pp. 189-220 in T.W. Schultz (ed.),

Collishaw, N. 1976 Profile Studies: Fertility in Canada. Vol v, Part I (Bulletin 5.1-6). Ottawa: Statistics

Canada De Tray, D.N. 1974 ‘Child quality and the demand for children.’ Pp. 91-117 in T.W. Schultz (ed.),

Easterlin, R.A. 1969 ‘Toward a socioeconomic theory of fertility.’ In S. J. Berman, L. Corsa, and R.

Economics of the Family. Chicago: University of Chicago Press

Economics of the Family. Chicago: University of Chicago Press

Freedman (eds.), Fertility and Family Planning: A World View. Ann Arbor: University of Michigan Press

Gardner, B. 1974 ‘Economics of the size of North Carolina families.’ Pp. 160-83 in T.W. Schultz

Goldberger, A.S. 1964 Econometric Theory. New York: Wiley Goodman, L. 1976 ‘The relationship between modified and usual approaches to the analysis of

(ed.), Economics of the Family. Chicago: University of Chicago Press

dichotomous variables. ’ Pp. 83-110 in D. R. Heise (ed.), Socialogical Methodology, 1976. San Francisco: Jossey-Bass

Grizzle, J.E., C.F. Starmer, and G.G. Koch 1969 ’Analysis of categorical data by linear models.’ Biometrics 25: 489-504 Hawthorn, G. 1970 The Sociology of Fertility. London: Macmillan Henripin, J., and J. Legare 1971 ‘Recent trends in Canadian fertility.’ Canadian Review of Anthropology and

Sociology 8: 106-18 Hout, M. 1977 ‘Changes in the impact of age at first marriage on cohort fertility in the United

1978 ’The determinants of martial fertility in the United States, 1968-1970: inferences

Kyriazis, N. 1979 ’Sequential fertility decision-making: Catholics and Protestants in Canada.’

Long, L.H. 1970 ‘Fertility patterns among religious groups in Canada.’ Demography 7: 135-49 Mason, K.O., W.M. Mason, H.H. Winsborough, and W.K. Poole 1973 ‘Some methodological issues in the cohort analysis of archival data.’ American

Michael, R.T. 1974 ‘Education and the derived demand for children.’ Pp. 120-56 in T.W. Schultz

States. ’ Unpublished manuscript.

from a dynamic model.’ Demography 15: 275-89

Canadian Review of Sociology and Anthropology 16: 275-86

Sociological Review 38: 242-57.

(ed.), Economics of the Family‘ Chicago: University of Chicago Press

43 ANALYSIS OF COMPLETED FERTILITY I N C A N A D A

Mincer, J. 1962 ‘The labor force participation of married women.’ In Aspects of Labour Economics.

Princeton: Princeton University Press 1963 ’Market prices, opportunity costs, and income effects.’ Pp. 67-82 in C.F. Christ

(ed.), Measurement in Economics Studies in Mathematical Economics and Econ- ometrics in Memory of Yehuda Grunfeld. Stanford. : Stanford University Press

Namboodiri, N. K. 1972 ‘Some observations on the economic framework for fertility analysis. ’ Population

Rosenzweig, M.R. 1976 ’Female work experience, employment status, and birth expectations: sequential

Ryder, N.B. and C.F. Westoff 1971 Reproduction in the United States. Princeton: Princeton University Press. Seiver, D.A. 1978 ’Which couples at given parities have additional births?‘ Pp. 309-19 in J. Simon (ed.),

Simon, J. 1974 The Effects of Income on Fertility. Chapel Hill, N.C. : Carolina Population

i975a ’Mixed effects of income upon successive births may explain the convergence

i975b Puzzles and further explorations in the interrelationships of successive births

Statistics Canada 1975 1971 Census of Canada: Public Use Sample Tapes. Ottawa: Information Canada Theil, H. 1970 ‘On the estimation of relationships involving qualitative variables. ‘ American

Thorton, A. 1978 ‘The relationship between fertility and income, relative income, and subjective

well-being.’ Pp. 261-90 in J. Simon (ed.), Research in Population Economics. Vol. I. Greenwich, Conn. : JM Press

Studies 26: 186-206

decision-making in the Philippines. ‘ Demography 13 : 339-56

Research in Population Economics. Vol. I. Greenwich, Conn: JAI Press

Center Monograph

phenomenon.’ Population Studies 29: 109-26

with the husbands’ income, spouse’s education and race.’ Demography 12: 259-74

Journal of Sociology 76: 103-54

Turchi, B.A. 1975 The Demand for Children. Cambridge: Ballinger Press Whelpton, P.D., A.A. Campbell, and J.E. Patterson 1966 Fertility and Family Planning in the United States. Princeton: Princeton Univer-

Willis, R. J. 1974 ’Economic theory of fertility behaviour.’ Pp. 25-75 in T.W. Schultz (ed.), Eco-

sity Press

nomics of the Family. Chicago: University of Chicago Press