Long-term intergenerational mobility in Quebec (1851-1951): the emergence of a new social fluidity regime.
Michel de Seve and Gerard Bouchard
Abstract: Using data extracted from parish registers of the Saguenay Region in Quebec from 1842 to 1971, this study explores the usefulness of the Erikson and Goldthorpe "core model of social fluidity" to describe the historical evolution of relative mobility in a "frontier" region. The conclusions of this study are two-fold: 1) relative mobility has only changed in minor ways between 1842 and 1971, 2) in so far as it has changed, it had converged toward the core model of social fluidity observed in contemporary industrialized societies by Erikson and Goldthorpe.
Resume: A l'aide de donnees extraites des registres paroissiaux de la region du Saguenay au Quebec entre 1842 et 1971, cette etude examine la pertience du "modele fondamental de fluidite sociale" propose par Erikson et Goldthorpe pour decrire l'evolution historique de la mobilite relative dans une region en voie de developpement. Deux conclusions principales sont proposees: 1) la fluidite sociale a peu change entre 1842 et 1971, 2) en autant que celle-ci a change, elle a evolue vers le "modele fondamental de fluidite sociale" observe dans les societes industrielles contemporaines par Erikson et Goldthorpe.
Two main theses are currently discussed by analysts of intergenerational social mobility in a historical perspective:
- according to some authors (Erikson and Goldthorpe, 1992), not only contemporary industrial societies share a similar regime of relative mobility but this "fluidity regime" has mostly remained constant since the beginning of the XXth century,
- according to others (Ganzeboom, Luijkx and Treiman, 1989), intergenerational exchanges between social positions in these contemporary societies are increasingly easy, even after eliminating the effects of the evolution of the social positions' distribution.
Recent debates published in the European Sociological Review (vol. 8, no. 3, December 1992) show that statistical models reflecting these two conceptions explain approximately the same proportion of observed relative mobility in contemporary industrial societies (Jones, 1992). Unfortunately, as far as we know, historical data required to test both hypotheses are very rare. On the one hand, contemporary data used by Erikson and Goldthorpe (1992) or Ganzeboom, Luijkx and Treiman (1989) have been collected during the second part of the XXth Century and they can at best provide a picture of social mobility since the beginning of the XXth Century. On the other hand, recent analysis of historical data using sophisticated models as those of van Leeuwen and Maas (1996), Miles (1993) or Fukomoto and Grusky (1993) do not allow for firm conclusions concerning the changing or stable character of the fluidity regime during the XIXth Century: while the first two suggest that this regime has changed, the last one supports the hypothesis of a constant relative mobility regime.
In Quebec, there exists only one study having compared the occupational mobility over a relatively long time period: Garon-Audy (1979) has examined three cohorts of young married men between 1954 and 1974 and pointed to a weak equalization of mobility odds over time. Reanalysed by Beland (1987), these data seem rather to indicate a stability in the association between fathers' and sons' positions. It seems necessary to complement these two last studies for at least three reasons: 1) the occupations of both the sons and fathers used in these studies are observed at the beginning of their careers (son's occupation at his wedding and father's occupation at the birth of his son, usually both before the age of 30), 2) the categories used for classifying these occupations do not permit comparisons with studies of other societies and 3) the period in question (1954 to 1974) is relatively short and recent.
This paper pursues two main objectives. First, we wish to present to social mobility researchers a new set of data facilitating the study of intergenerational social mobility for a relatively long time span. Indeed, the data set used in this paper contains information dating from 1840 to 1970 and allows the systematic observation of the stable part of the population in a region of Quebec since the early beginnings of its colonization to a more industrialized period after the Second World War. Second, as a first step towards the comparison of the two main hypotheses on the evolution of social mobility, we will test the capacity of the "core model of social fluidity" proposed by Erikson and Goldthorpe (1992) to shed some light on these historical data and on the validity of the hypothesis of a constant regime of social fluidity.
As with other historical studies on social mobility, the main limitation of this analysis is that the data used are limited to a specific region and it is uncertain whether its conclusions could be generalized to other regions of Quebec or of Canada (Garon-Audy, 1979: 195-218). Until more extensive historical data become available, these conclusions must be seen only as a first step towards a better understanding of the historical evolution of social mobility in Canada.
The Saguenay Region
The Saguenay Region is located in northeastern Quebec. Its settlement began in 1838-40 through migration from other parts of Lower Canada. The population of the region now numbers 300,000 as continuous migration losses have been more than offset by a very high birth rate. At its beginning, the economic activities of the region were a mix of agriculture and lumbering for Montreal and Quebec-based companies. At the end of the XIXth century and the beginning of the XXth Century, large pulp and paper factories as well as aluminium plants were being built. This gave rise to a dozen small towns throughout the region, most of them typical of the "milltown" pattern (St-Hilaire, 1996; Bouchard, 1996). Modern industrial activities and market involvement did not seem to have much influence on the local culture and society until the 1930s when the old "traditional" order was broken up by a number of factors like a shortage of agricultural lands and the Great Depression (Bouchard, 1996). The data used in this study should thus provide an opportunity to observe some of the effects of industrialization on social mobility.
The BALSAC Databank
The data used in this paper were extracted from the BALSAC databank developed by researchers of the "Institut interuniversitaire de recherches sur les populations" (I.R.E.P.) (Bouchard, Roy, Casgrain and Hubert, 1995, 1989). BALSAC contains information collected in the Saguenay and Charlevoix parish registers (birth, marriage and death certificates), census manuscripts and various other social, cultural and economic sources between 1842 and 1971. Today, the number of records in BALSAC for the Saguenay region is approximately 665,000. As these records have been systematically linked for each individual, it is possible to know many of the different occupations (between 15 and 20 of them) held by virtually everyone who lived in the Saguenay Region between 1842 and 1971 (other information is also available, such as the dates and places of birth, marriage, death, etc.).
This first attempt at evaluating BALSAC's potential usefulness to study social mobility over an extended amount of time could not consider all the individuals included in this databank. We have therefore limited this study to individuals who married during 6 distinct periods: 1842-1856, 1866-1876, 1886-1896, 1906-1916, 1926-36 and 1946-56 (hereafter designated by their midpoints: 1851, 1871, 1891, 1911, 1931 and 1951).
For each year in a given period, we pulled out all marriages at which the bridegroom was 30 years old or younger, in order to create cohorts of individuals at the beginning of their working life during these periods. (Bridegrooms whose age at marriage was unknown were also included in the data; they represented 8.9% of the cases.) As unigeniture and celibacy were not frequent in the region (Bouchard, 1996), this rule does not seem to bias the selection one way or another.
Historical studies based on marriage records are frequently criticized for observing the social standing of the bridegroom only at his marriage and, thus, before it is possible to observe the influence of social origin on his social position in the middle or near the end of his occupational life. So, we have tried to compare the occupations of the fathers to those of the sons when both were near 50 years of age. This was done by finding in the BALSAC databank all the occupations of both the bridegrooms and the fathers registered between 40 and 60 years of age and by chosing among these occupations the one closest to 50 years of age. This method provided known occupations for 7,018 fathers and sons: from 193 for the first period (1851) to 741 for the last (1951) (see the last line of Table 1).
Evidently, one may wonder about the many possible selection biases affecting these data, especially considering the small number of observations for the 1946-1956 marriage period (741 cases). The main cause of information loss is none of the usual ones (i.e. death, lack of sufficient information about the occupation or geographical migration), but the closure of the BALSAC databank after 1971. The closure of the BALSAC databank prevents us from knowing the later occupations of approximately 50% of the sons married during the 1946-1956 marriage period since many of them had not reach the age of 40
Table 1. Sons' EGP classes at 50 years of age by marriage periods (in
Classes Marriage period 1851 1871 1891 1911 1931 1951 I-II 0.5 1.0 1.4 2.2 3.9 7.3 III 0.5 0.5 1.1 1.4 3.6 4.9 IVab 2.1 3.0 8.4 8.3 10.3 7.7 IVc 73.1 70.3 57.9 49.7 35.8 20.4 V-VI 3.6 8.1 10.0 12.0 16.4 22.9 VIIa 1.6 7.4 12.5 18.0 20.4 25.4 VIIb 18.7 9.7 8.7 8.5 9.6 11.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 N 193 596 1229 1955 2304 741
in 1971. Yet, as the age of the fathers and sons of this period is similar to the age of the fathers and sons of the other periods, the main effect of this selection bias is that mobility observed among the bridegrooms of the last period occured nearer in time to mobility observed among the bridegrooms of the immediately antecedent period (1926-1931) than it is the case for the other periods (de Seve, Bouchard and Hamel, 1995a).
The Classification of Occupations
The BALSAC databank contains some 3,975 distinct occupational designations used by the Saguenay inhabitants in the parish registers. A first step was necessarily to regroup these titles into a smaller set of categories allowing for a comparative analysis of social mobility. Given the importance of the debates surrounding the conclusions proposed by Erikson and Goldthorpe, we have chosen to use the seven classes scheme developped by these authors for the study of social mobility in various industrial societies (Erikson and Goldthorpe, 1992: 35-47).
The categories of occupations are the followings (de Seve, 1993):
I-II the service class (mainly professionnals, managers and owners of large enterprises);
IIIab the routine non-manual employees;
IVab the owners of small enterprises outside the primary sector;
IVc the owners in the primary sector (mostly farmers);
V-VI the classes of lower-grade technicians, foremen and skilled manual workers outside the primary sector;
VIIa the semi- and unskilled manual workers not in the primary sector;
VIIb the workers in the primary sector.
Our adaptation of the EGP class scheme to this historical dataset raised two problems. Firstly, the data did not contain all the information that would be judged necessary by a researcher using contemporary data to classify all the occupations among the 7 categories of the scheme. Two kinds of occupations in particular called for compromises: the owners not in the primary sector and the rural labourers. Since we had no information regarding the size of the businesses of the owners, the vast majority of them were classified in sub-class IVab while some of them would have been more properly classified in class I-II. In the case of the rural labourers (VIIb), class distributions of for both fathers and the sons showed a very weak number of agricultural and forest labourers; this is mostly due to the fact that these "jobs" were occupied in Saguenay mainly before marriage. So, the decision was made to classify VIIb all individuals who had indicated as their occupation only the word "manoeuvre" ("labourer" without any added information) and who resided in a "rural" area of Saguenay when this occupation was observed.
Secondly, one might question the legitimacy of applying a contemporary class scheme to historical data. However, one aspect of this issue has been thoroughly investigated by G. Bouchard for the Saguenay region (Bouchard, 1996: Chapter V) and his findings indicate that two of the main operational criteria used to classify the occupations in the EGP class scheme (the distinction between manual and non manual occupations and their relative skills level) did not varied much during the period between 1840 and 1970. Yet, one could ask if the 7 categories of the class scheme regroup occupations both socially homogeneous and distinct in the Saguenay Region at the turning of the century. At this moment, we can only offer a tentative answer. On the one hand, one of our objectives is to empirically test the relevance as well as the usefulness of applying such a definition of social positions to an historical dataset. As the following analyses will show, the EGP class scheme revealed an evolution in both the occupational structure and in total mobility reasonably plausible according to our present knowledge. On the other hand, in a previous study (de Seve, Bouchard and Hamel 1995b), we have examined the historical homogeneity of the classes distinguished in terms of social mobility and its results suggested that the EGP classes are indeed homogeneous with two exceptions: in sub-class IVab, it would have been useful to distinguish the "traditional" craftsmen ("artisans") from the other small owners, and category V-VI hides a certain amount of mobility by regrouping the foremen and the skilled manual workers. However, in the dataset used for this study, these two specific subgroups are not sufficiently numerous to allow a separate analysis.
The Evolution of the Sons' Classes
Table 1 suggests that the Saguenay Region industrialized rapidly between 1851 and 1951. In 1851, almost all (91.8%) of the sons were in the two "rural" subclasses of the primary sector (sub-classes IVc and VIIb) and the other categories were almost empty: 1% in classes I to III, 2.1% in the sub-class of the small business owners IVab, and 5.2% in manual industrial occupations (V-VI and VIIa) (first column of Table 1). One hundred years later, the primary sector accounts for a minority of the sons (31.8%), while the manual workers represent close to the majority (48.3%) (and the non-manual classes I to III only 12.2%, last column of Table 1). (This general picture, taking into account only the sons for which we also know the father's occupation is not modified when we consider the sons independently of the father's occupation [data not shown]).
Two classes seem to evolve in a less regular manner than the others: after 1891, the small business owners' sub-class (IVab) does not seem to expand, and the rural labourers (VIIb) remains quite unchanged between 1871 and 1951.
To sum up, the evolution of the sons' classes from 1871 to 1951 reveals a region shifting from an agricultural world to an industrial one. Perhaps more importantly is the absence of a trend indicating the clear emergence of a tertiary sector.
Total Intergenerational Mobility
Social intergenerational mobility can be examined from two points of view: absolute mobility and relative or "net" mobility. Total absolute mobility describes the total percentages of mobiles and immobiles among all the cases observed for each period of time. However, because these percentages are affected by the differences of social positions distributions between the fathers and the sons, they reflect the simultaneous influence of exogenous factors affecting the evolution of the class structure and the chances of mobility (and of immobility) associated with the specific advantages and disadvantages of each class. This distinction is important since proponents of the constant social fluidity mobility regime hypothesis admit that total observed mobility has changed during industrialization (or can be different between societies). As this section considers only total mobility, one expects to find differences, if not a clear trend, from one time period to another.
Two characteristics of observed mobility are presented in this section: the percentage of sons who are in the same class as their father and the percentage of those who are not. Also, among the mobiles, three kinds of mobility are considered: upward, downward and what is called "horizontal" mobility.
Since the classes defined by Erikson and Goldthrope are not ordered according to some continuous measure of social position (such as socio-economic status), the distinction between "vertical" and "horizontal" mobility needs to be explained. Even if Erikson and Godlthorpe explicitly maintain that their classes are not totally hierarchised (Erikson and Goldthorpe, 1992: 44), they believe that some of them can be ordered and they propose a threefold hierarchical division among their seven classes (see Wong, 1992, about simultaneously taking into account both vertical and non vertical dimensions of stratification):
a) upper stratum: classes I and II,
b) intermediate stratum: classes III, IVab, V and VI,
c) lower stratum: classes VIIa and VIIb.
Therefore, both upward and downward mobilities are defined by inter-generational mobility between classes in two different strata, while "horizontal" mobility describes exchanges between two classes within a same stratum.
Lastly, in order to account for the commercialization of farming that occurred between the two generations observed in their contemporary data, Erikson and Goldthorpe suggest to include the farmers (sub-class IVc) in the lower stratum when observing the fathers, and in the intermediate one when considering the sons (Erikson and Goldthorpe, 1992: 46). As our data come mainly from historical periods prior to those used by these two authors and as other studies have shown that the commercialisation of agriculture was mitigated in the Saguenay Region (Bouchard, 1996: Chapter IV), the validity of this rule has to be reexamined. Three different kinds of analysis regarding the relative position of the farmers were conducted. (These analyses are not presented here, but are available from the authors). The results suggest regrouping both the two generations of farmers in the same lower stratum in each period. Table 2 presents thus total observed mobility according to this definition of the three strata.
Table 2. Total observed mobility by marriage periods between fathers' classes at 50 years of age and sons' classes at the same age (in
Marriage period 1851 1871 1891 1911 1931 1951 1. immob. 70.5 68.1 56.8 51.1 40.7 31.9 (a) 5.2 3.5 3.7 6.2 8.8 12.7 (b) 65.3 65.6 53.1 44.9 31.9 19.2 2. upward 5.2 10.9 16.9 18.9 26.6 32.4 (a) 0.0 0.8 2.0 3.6 7.4 12.8 (b) 5.2 10.1 14.9 15.3 19.2 19.6 3. down. 7.8 3.4 3.6 4.2 5.0 5.7 (a) 3.1 1.3 1.5 2.4 3.9 5.0 (b) 4.7 2.0 2.1 1.8 1.1 0.7 3. hor. 16.6 16.6 22.6 25.8 27.6 30.1 (a) 0.0 1.7 3.0 3.3 5.2 6.1 (b) 16.6 14.9 19.6 22.6 22.5 24.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 N 193 596 1229 1955 2304 741 - upward: upward mobility, - down.: downward mobility, - hor.: horizontal mobility, - lines marked (a) contain cases which exclude the farmers (IVc), - lines marked (b) contain cases in which the father and/or the son was a
This table displays four trends in total observed mobility from 1851 to 1951:
- in the non-farmers' classes, immobility within classes and horizontal mobility within the three strata increased: from 5.2% to 12.7% for immobility and from 0.0% to 6.1% for horizontal mobility;
- downward mobility was not a significant phenomenon in Saguenay since it didn't exceed 8% in any of the time period; nevertheless, after 1851, it seems to have increased slightly among the non-farmers, from 1.3% in 1871 to 5.0% in 1951;
- upward mobility increased continuously among the non-farmers as well as among the farmers, from 5.2% in 1851 to about 32.4% in 1951;
- farmers' sons who became farmers themselves were less and less numerous: 65.3% in 1851 compared to only 19.2% in 1951 (76.6% of the mobile farmers' sons became manual workers, data not shown).
To sum up, the increase in immobility and in short (horizontal) mobility rates outside class IVc suggest the appearance of barriers between the strata and classes despite a clear rise of upward mobility. As these two trends can be explained by the influence of industrialization on the relative importance of the different categories in the class structure of the Saguenay Region, it is necessary to examine whether changes can be also noted in social fluidity, without the potential impact of industrialization.
Erikson and Goldthorpe's Core Model of Social Fluidity
The "core" model of social fluidity developed by Erikson and Goldthorpe (1992: 121-139) supposes that social fluidity can be largely explained by 8 factors corresponding to 8 distinct parameters of a "topological" log-linear model (Hout, 1983):
- two factors reflect the social distance covered by a change of strata: a change between contiguous strata (HI1), and one between non contiguous strata (HI2); at first glance, the parameters corresponding to these factors should be negative because mobility is more difficult between more distant classes;
- three factors consider "inheritance" or "stability" within the different classes: a general term (IN1) affecting all classes, a particular term (IN2) for the three categories containing owners (I-II, IVab and IVc) and, lastly, a specific one (IN3) for the category of "farmers" (IVc); usually, these parameters are positive because they measure the well known tendency of sons to occupy the same position as their father;
- one factor (SE) measures the difficulties related to transitions between the primary sector and the other sectors; as HI1 and HI2, this parameter is usually negative,
- finally, two factors express particular "affinities", whether they be negative (AF1) or positive (AF2), between some classes; AF1 describes the difficulty of mobility between classes I-II and VIIb, while AF2 reflects the easier mobility some other classes (between I-II and IIIab or IVab, between IVab and IVc, from IVab or V-VI to VIIa, from VIIa to V-VI and from VIIb to VIIa).
(Values of these parameters as observed by Erikson and Goldthorpe in nine contemporary societies are presented on the first line of Table 4.)
Testing the core model of social fluidity
To statistically test the core model, we have compared different log-linear models shown in Table 3 according to three criteria: 1) the statistical significance
Table 3. Fitting the modified core model of social fluidity to mobility tables (1851-1951, N = 7,032, sampling 0 = 0.1) Model G2 df rG2 BIC a) time constant models 1. conditional independence (ref. model) 1120.9 180 - -473.6 2. descriptive constant model 138.6 144 .88 -1137.0 3. constant core model 196.0 172 .83 -1327.6 4. trimmed constant core model 198.2 174 .82 -1343.1 b) time changing models 5. non constant core model 146.5 140 .87 -1093.7 6. non constant trimmed core model 170.2 166 .85 -1300.3 c) differences between some models G2 df rG2 7. model 4 - model 3 2.2 2 .01 8. model 3 - model 5 49.5" 32 .04 9. model 4 - model 6 28.0* 8 .05 10. model 6 - model 5 23.7 26 .02 - the descriptive constant model is formally equivalent to the common fluidity model used by Erikson and Goldthorpe; substantively, it differs from it by supposing the same relative mobility in each marriage period while the common model supposes the same relative mobility in each society, - 1851 and 1871 periods are regrouped because of the small number of available observations during these periods, - * = statistically significant at .05, - trimmed core model = core model without IN3 (supplementary stability among the farmers) and AF1 (negative affinities between classes I-II and
- non constant models: models with interactions between the periods and the factors of the core model: model 5 with period interaction with the 8 factors of the core model and, model 6, with only IN1 and SE. Table 4. Parameter values for some tested core models of social fluidity presented in Table 3 Model Parameters HI1 HI2 IN1 IN2 IN3 SE AF1 AF2 a) constant models: 1. EGP -.22 -.42 +.43 +.81 +.96 -1.03 -.77 +.46 2. cst -.15 -.47 +.50 +.66 -.26(1) -.47 -.08* +.34 3. trim. -.18 -.50 +.52 +.51 .00 -.36 .00 +.31 b) non constant trimmed core model: 1851-71 -.21 -.50 +.81 +.51 .00 +.15 .00 +.31 1891 -.21 -.50 +.47 +.51 .00 -.19 .00 +.31 1911 -.21 -.50 +.46 +.51 .00 -.33 .00 +.31 1931 -.21 -.50 +.39 +.51 .00 -.39 .00 +.31 1951 -.21 -.50 +.45 +.51 .00 -.82 .00 +.31 - EG: parameter values for the core model as observed by Erikson and Goldthorpe in contemporary societies (Erikson and Goldthorpe, 1992: table 4.4., 135), - cst: non trimmed constant core model applied to the Saguenay data (model 3 in Table 3), - trimm.: trimmed constant core model applied to the Saguenay data (model 4 in Table 3), - non constant trimmed core model: core model with IN3 and AF1 equal to zero but with interactions between periods and IN1 and SE (model 6 in Table
(1): coefficient less than twice its standard error.
of the model's chi-square value (G2), 2) the proportion of mobility and immobility "variance" explained by the model (rG2) (Hagenaars, 1990) and 3) the value of a bayesian index of fit (BIC, Bayesian Information Criterion, Raftery, 1996). BIC's usefulness appears mostly when different models are equally good according to the chi-square test or to the proportion of explained mobility indicated by the rG2. BIC takes into account not only the goodness of fit of the model but also its parsimony and its formula is such that a high negative value is indicative of a better model than a less negative value or a positive one.
In a first step, four different models were compared:
- a reference model that supposes that mobility is only affected by the evolution of the fathers' and sons' class distributions (model 1 of Table 3); obviously, as is shown by its high G2 value, this model does not fit the data; yet, this G2 value does provide a standard to measure the "variations" in mobility which cannot be explained by the evolution of the class distributions;
- a "constant" descriptive model with "ad hoc" specific parameters for each cell in the mobility tables compared; since the specific parameters of these cells are the same from one period to the next, its main advantage is to test the hypothesis that immobility and mobility remained constant between 1851 and 1951; according to the results in line 2 of Table 3, this model fits the data very well: 1) had we used a probability sample, its G2 could not be considered statistically significant, 2) its BIC value is a large negative and, 3) its rG2 indicates that a constant model can describe 88% of the variations in mobility unexplained by the reference model 1;
- the core model propounded by Erikson and Goldthorpe when it is supposed that its eight factors have the same constant effect from one marriage period to another (model 3 of Table 3); as this model is a simplified analytical representation of mobility, it is not surprising to observe that its fit is worst than the constant descriptive model of line 2 according to the its G2 value (196.0 compared to 138.6) and to the rG2; however, its own fit remains satisfactory:
- its G2 could not be considered statistically significant with a sample (less than the significant value for G2 at .05, 203.6),
- its rG2 value (83%) shows that it explains nearly the same amount of unexplained mobility variations as the purely descriptive constant model 2 (88%),
- its BIC value suggests a better equilibrium between fit and parsimony than the a-theoretical constant model 2 (-1327.6 compared to -1137.0),
- as the low ratio G2/df of model 3 suggests that it fits the observations perhaps too well, we attempted to simplify this model by eliminating two weak factors: IN3 and AF1 (probably non statistically significant ones, see their values in line 2 of Table 4); the results in line 4 of Table 3 show that this "trimmed" model fits the observations nearly as well as the "complete" constant core model 3 (and line 7 confirms that the difference between these two models would not be generally be considered statistically significant); this "trimmed" constant model reveals two things about the Saguenay Region:
- there were no special obstacles to mobility between classes I-II and VIIb other than those related to hierarchical and sectoral effects;
- more surprisingly, farmers' sons did not "inherit" their fathers' class more often than other owners; in this region where the number of farmers was rapidly decreasing, it seems a reasonable hypothesis that farmers' sons were not as intergenerationally stable as they were in the contemporary societies observed by Erikson and Goldthorpe.
The fit indexes of the three models tested until now suggest that it is not necessary to suppose that fluidity between the classes has changed between 1851 and 1951 to obtain a satisfying picture of net mobility. At most, they indicate that changes were minimal (less than 12% according to the unexplained variations of descriptive model 2). Nevertheless, in order to determine whether some changes, even small ones, could be detected by the core model, we tested other models which supposed that all or some of the eight parameters of the core model varied historically. Line 5 of Table 3 shows the fit indexes for a model in which each of the eight factors could vary by periods and line 6, the same indexes for the best non constant model we could find (according to the BIC index):
- comparison of the G2's values for the constant core models and the non constant ones indicates that supposing temporal variations in the parameters increases significantly the goodness of fit of the models (lines 8 and 9 of Table 3); however, the "complete" non constant model (model 5 in Table 3) is less satisfying than the two constant core models according to the BIC index (-1093.7 versus -1327.6 and -1343.1);
- the elimination of the statistically less significant parameters measuring changes (smaller than twice their sampling error) provides a model (model 6 in Table 3) which is almost as good as the descriptive model according to the rG2 (85% compared to 88%) and as the "trimmed" constant core model according to the BIC (-1300.3 compared to -1343.1).
Since model 6, which suggests a certain amount of changes, seems nevertheless to be globally satisfying, it is useful to examine the values of the parameters of the different variants of the core model in order to gain a better understanding of the fluidity regime in the Saguenay Region.
The Saguenay's Fluidity Regime
Table 4 contains the values of the parameters of the core model as estimated by three models presented in Table 3:
- line 2 shows the values obtained with the Saguenay data and line 1, the values observed by Erikson and Goldthorpe in 9 contemporary societies; despite the obvious similarities between the two sets of estimates, some differences exist:
- as said earlier, the inheritance parameter for the farmers (IN3) and the negative affinities between classes I-II and VIIb are not statistically significant (less than twice their sampling error);
- three parameters have lower values than those observed by Erikson and Goldthrope: stability among the three classes containing owners (IN2), sectoral effects (SE), and, perhaps, positive affinities (AF2);
- as there are no published studies that use the core model to describe social fluidity in Canada or in Quebec, it is difficult at this time to decide whether these differences are specific to Canada (or to Quebec) or to the Saguenay Region;
- line 3 contains estimates for the effects of the core constant model which excludes IN3 and AF1 (model 4 of Table 3); it shows that the elimination of these two factors does not have an important effect on the other estimates,
- the last five lines present parameters values for the "trimmed" non constant model (model 6 of Table 3):
- even during the early beginnings of this region, hierarchy (HI1 and HI2), ownership (IN2) and positive affinities between classes (AF2) affected social fluidity;
- stability in all the classes (IN1) was more present in the first marriage period (IN1 = .81); from then on, it oscillates between .39 and .47 at approximately the same level as Erikson and Goldthorpe (.43);
- sectoral effects (SE) which diminish mobility between the primary sector and the other sectors increase continuously (from +.15 to -.82): it became more difficult for the farmers' sons (as well as the rural workers' sons) to have access to the non rural classes; at the same time, fewer and fewer "urban" or "industrial" workers joined these two categories.
Since the core model is defined by 8 different parameters, it is difficult to discern their total impact on the fluidity between the different classes. In order to show their combined effect, we have summed the values of these parameters in each cell of the mobility tables and compared these combined effects to those presented by Erikson and Goldthorpe (1992: Table 4.4, page 135).
When the total effect is calculated with the estimates obtained by Erikson and Goldthorpe, it is possible to distinguish two main areas of fluidity in a mobility table (see the first part of Table 5):
- a first area in which expected frequencies are higher (this area contains only positive total effects and appears in bold numbers) than when origins and destinations are statistically independent; it comprises the cells on the diagonal of the mobility table and those in two smaller zones of preferential exchanges (between classes V-VI and VIIa on the one hand and between classes I-II, IIIab and IVab on the other);
- a second area in which exchanges are less "numerous" (as indicated by negative terms) or equal to those predicted by statistical independence between origins and destinations (indicated by null terms).
The second part of Table 5 presents the combined effects of the same parameters observed in the Saguenay Region while supposing a constant model (model 4 of Table 3). It shows that the two areas of social fluidity are equally present in the Saguenay Region. To get a better understanding of the evolution of social fluidity in this region, the combined effets were obtained for each marriage period using the parameter values of model 6 in Table 3 and two indexes aggregating these combined effects were calculated:
Table 5. Combined effects of the core model of social fluidity parameters in each cell of the mobility tables (rows = fathers, columns = sons) a) combined effects observed by Erikson and Goldthorpe (Erikson and Goldthorpe, 1992: table 4.4., 135): I-II III IVab IVc V-VI VIIa VIIb I-II 1.24 0.24 0.24 -1.25 -0.22 -0.64 -2.24 III 0.24 0.43 0.00 -1.03 0.00 -0.22 -1.25 IVab 0.24 0.00 1.24 -0.57 0.00 -0.22 -1.25 IVc -1.67 -1.25 -0.79 1.98 -1.25 -0.57 0.00 V-VI -0.22 0.00 0.00 -1.03 0.43 0.24 -1.25 VIIa -0.64 -0.22 -0.22 -1.25 0.24 0.43 -1.03 VIIb -2.44 -1.25 -1.25 -0.22 -1.25 -0.57 0.43 b) combined effects observed in the Saguenay data (constant trimmed model 4 of Table 3): I-II III IVab IVc V-VI VIIa VIIb I-II 1.03 0.13 0.13 -1.04 -0.18 -0.68 -1.04 III 0.13 0.52 0.00 -0.54 0.00 -0.18 -0.54 IVab 0.13 0.00 1.03 -0.23 0.00 -0.18 -0.54 IVc -1.04 -0.54 -0.23 1.03 -0.54 -0.05 0.00 V-VI -0.18 0.00 0.00 -0.54 0.52 0.13 -0.54 VIIa -0.68 -0.18 -0.18 -0.36 0.13 0.52 -0.36 VIIb -1.04 -0.54 -0.54 0.00 -0.54 -0.05 0.52 c) differences in absolute value between the combined effects observed by Erikson and Goldthorpe and those observed in the Saguenay data (EG/Sagu.), and total combined effects in the Saguenay by marriage period (Total
1851-71 1891 1911 1931 1951 All EG/Sagu. 24.9 18.7 16.5 15.7 9.5 16.4 Total Effects 4.1 -5.1 -8.0 -9.7 -17.9 -7.3
- the first one is defined by the sum of the absolute differences between the combined effects observed by Erikson and Goldthorpe and those observed in the Saguenay for each marriage period: as seen on the first line of the third part of Table 5 ("EGP/Saguenay"), the values of this simple index decrease over time from 24.9 in 1851-71 to 9.5 in 1951; in other words, the differences between the combined effects observed in the Saguenay converge toward those observed in other modern societies;
- the second one is even more simple as it adds the values of the combined effects observed in each marriage period; the second line of the third part of Table 5 ("Total effects") shows that, with time, negative combined effects become clearly more important than positive ones; as the only changing negative parameter in model 6 pertains to mobility between the primary sector and the other sectors (SE), this reduction indicates increasing cleavages between the "more urban" classes and the "rural" ones.
The existing differences between the combined effects observed by Erikson and Goldthorpe and those observed in the Saguenay Region in 1951 (9.5) suggest either that the fluidity regime of this region had not yet completed its evolution towards a more contemporary regime or that it is distinct from other previously observed regimes.
Absolute Mobility and Social Fluidity Compared
The small differences between the proportion of mobility explained by models with and without temporal changes, as presented in section 5.2., suggest that changes in social fluidity were less important than its stable components. Similarly, one can question the relative importance of the changes in "absolute" mobility compared to stable "absolute" mobility. There is no known satisfying technique that can provide an answer to this question (Sobel, Hout and Duncan, 1991; Erikson and Goldthorpe, 1992: 205; Marks, Pittelkows and Jones, 1996). As a temporary solution, however, one may turn to a technique proposed by Breen (1985) in order to decompose the communalities and differences between different mobility tables as indicated by their G2 values.
The percentages presented in Table 6 suggest that 95.6% of the variations observed in the mobility tables can be explained by stable components: 90.5% is due to the differences between the fathers' and sons' class distributions, and the remaining 5.1% to the stable characteristics of social fluidity. Thus, only 4.4% of the variations could be associated with changing mobility: 4.3% to the evolution of the fathers' and sons' distributions, and 0.1% to changes in social fluidity.
To sum up, notwithstanding their intrinsic interest, changes in social fluidity were minor in the Saguenay Region and the total mobility observed is mainly explained by the the numerical increase or decrease in the different classes.
Three main conclusions can be drawn from this first analysis of social mobility in the Saguenay Region:
- the core model of social fluidity has proven to be a dependable tool for the study of mobility in an historical perspective: it has provided a reasonably good outlook on the observations and has permitted the identification of certain plausible changes in the fluidity regime of the Saguenay; yet, it remains
Table 6. Decomposition of total mobility according to Breen's technique G2 Unexplained variance of total mobility and total variance to be explained: Unexplained variance (random error) (1) 170.2 Total variance to be explained (2) 22,107.5 % Decomposition of total variance to be explained in percentages: Stable mobility: stable absolute mobility 90.5 stable fluidity 5.1 sub-total 95.6 Changing mobility: changing absolute mobility 4.3 changing fluidity 0.1 sub-total 4.4 Total 100.0 (22,107.5) (1) G2 of the best model (model 6 of Table 3), (2) difference between the G2 of the "best" model (model 6 of Table 3) and a model with only a period effect.
to be seen whether no other theoretical models of social fluidity could be applied to this or to similar datasets with the same success;
- even if a model allowing for changes in social fluidity has certain merits, it seems that the stable aspects of social fluidity predominates over the changing ones; therefore, the explanation for the changes in total observed mobility shown at the beginning of this study rests mainly on changes in the fathers' and the sons' class distributions;
- moreover, if the hypothesis of changes in social fluidity is retained, the observed data suggest that the fluidity regime in the Saguenay Region converged toward the fluidity regime observed in contemporary societies by Erikson and Goldthorpe; unfortunately, this convergence implies a "closing" rather than an "opening" of the "relative mobility regime" as mobility difficulties have increased for the agricultural classes, while the existing mobility channels between the other classes remained stable between 1851 and 1951.
The last result is therefore at odds with the hypotheses according to which the evolution of industrial societies is characterized by a general reduction in mobility obstacles. Indeed, our results are perhaps slightly more congruent with the hypothesis that proposes a sharp change in the fluidity regime at the beginning of the industrial period followed by stabilization afterwards.
The dataset analyzed here presents three characteristics limiting its usefulness for studying this last hypothesis: firstly, 1840 may be a too recent period to effectively observe the effects of the beginnings of industrialization in contemporary societies; at the same time, a "society" in a settlement process is perhaps not the best context to compare preindustrial and industrial mobility; lastly, these data came from a single region of Quebec and it is impossible to know if relative mobility was as stable in other regions or if geographical mobility was not in anyway a prerequisite to more open social mobility.
1987 "A comparison of the mobility structures of francophones and anglophones in Quebec: 1954, 1964, 1974", Canadian Review of Sociology and Anthropology, 24(2): 232-251.
1996 Quelques arpents d'Amerique, Population, economie, famille au Saguenay 1838-1971, Montreal, Boreal.
Bouchard, Gerard, Raymond Roy, Bernard Casgrain and Michel Hubert
1995 "Computer in human sciences: From family reconstitution to population reconstruction". In Ephraim Nissan and Klauss M. Schmidt (ed.), From Information to Knowledge, Conceptual and Content Analysis by Computer, Oxford (UK), Intellect: 201-226.
Bouchard, Gerard, Raymond Roy, Bernard Casgrain and Michel Hubert
1989 "Fichier de population et structure de gestion de base des donnees: le fichier-reseau BALSAC et le systeme INGRES/INGRID", Histoire et Mesure, IV(1/2): 39-57.
1985 "A framework for comparative analyses of social mobility", Sociology, 19(1): 93-107.
de Seve, Michel, Gerard Bouchard and Martin Hamel
1995a "Selection biases in estimating social mobility with historical data: The case of Saguenay, 1851-1951", Birmingham: paper presented at the Historical Social Mobility Conference, Birmingham, UK, April 1995.
de Seve, Michel, Gerard Bouchard and Martin Hamel
1995b "Mobilite sociale intergenerationnelle et classification des professions, le cas du Saguenay, 1871-1931", Chicoutimi: Paper presented at the annual meeting of the Association canadienne des sociologues et anthropologues de langue francaise, 63th ACFAS Conference, Chicoutimi, Canada, May 1995.
de Seve, Michel
1993 Les professions relevees dans le fichier BALSAC (Saguenay) et les classes d'Erikson, Goldthorpe et Portocarero, essai de classification, Chicoutimi (Canada), IREP, Document no. II-E-217.
Erikson, Robert and John H. Goldthorpe
1992 The Constant Flux, A Study of Class Mobility in Industrial Societies, Oxford (UK), Clarendon Press.
Fukomoto, Ivan K. and David H. Grusky
1993 "Social mobility and class structure in early-industrial France". In Miles, Andrew and David Vincent (ed.), Building European Society, Occupational Change and Social Mobility in Europe, 1840-1940, Manchester (UK), Manchester University Press: 40-67.
Ganzeboom, Harry B.G., Ruud Luijkx and Donald J. Treiman
1989 "Intergenerational class mobility in comparative perspective", Research in Social Stratification and Mobility, 8: 3-84.
1979 Mobilites professionnelles et geographiques au Quebec, 1954-64-74, Montreal (Canada), Centre de recherche en developpement economique, Universite de Montreal.
Hagenaars, Jacques A.
1990 Categorical Longitudinal Data, Log-Linear Panel, Trend, and Cohort Analysis, Thousand Oaks (USA), Sage.
1983 Mobility Tables, Thousand Oaks (USA), Sage, Quantitative Applications in the Social Sciences, #31.
1992 "Common social fluidity: A comment on recent criticisms", European Sociological Review, 8(3): 233-237.
Marks, G.N., Y.E. Pittelkow and F.L. Jones
1996 "Structural changes and association in mobility tables: An integration and extension of existing approaches", Quality and Quantity, 30: 61-86.
Miles, Andrew and David Vincent D (ed.)
1993 Building European Society, Occupational Change and Social Mobility in Europe, 1840-1940, Manchester (UK), Manchester University Press.
1993 "How open was nineteenth-century British society? Social mobility and equality of opportunity, 1839-1914". In Miles Andrew and David Vincent D (ed.), Building European Society, Occupational Change and Social Mobility in Europe, 1840-1940, Manchester (UK), Manchester University Press: 18-39.
Raftery, Adrian E.
1996 "Bayesian model selection in social research". In P.V. Marsden (ed.), Sociological Methodology, vol. 25: 111-163.
1996 Peuplement et dynamique migratoire au Saguenay, 1840-1960, Ste-Foy (Canada), Presses de l'Universite Laval.
Sobel, Michael E., Michael Hout and Otis Dudley Duncan
1985 "Exchange, structure and symmetry in occupational mobility", American Journal of Sociology, 91(2): 359-372.
van Leeuwen, Marco H. D. and Ineke Maas
1996 "Long-term social mobility: Research agenda and a case study (Berlin, 1825-1957)", Continuity and Change, 11(3): 399-433.
Wong, Raymond Sin-Kwok
1992 "Vertical and nonvertical effects in class mobility: Cross-national variations", American Sociological Review, 57(3): 396-410.
|Printer friendly Cite/link Email Feedback|
|Author:||De Seve, Michel|
|Publication:||Canadian Journal of Sociology|
|Date:||Sep 22, 1998|
|Previous Article:||New dimensions of social movement/countermovement interaction: the case of scientology and its Internet critics.|
|Next Article:||Significance of family, work, and power relations for mothers' mental health.|