NOTES
1.
The standard sources (Hietzinger 1817, 1820; Vanicek 1875; Rothenberg 1960, 1966) are often inspecific about which regimental regions of the Border were most directly involved.
However, in these major crises, most may have been.2.
Such households are referred to in the ethnographic and historical literature (but not in peasant discourse) by the term Zfldruga (loosely, ‘co-operative’). They are typically agnatically defined, with father and married sons, or married brothers, or even cousins. They were perhaps most developed in the Military Border of Croatia, encouraged and codified by the Austrians as a way to ensure household survival while some manpower was drawn off for military purposes; their mean size in the six western regiments was about eleven persons (Roksandic 1988: vol. I, 35). (For a review of issues concerning such households, see Tomasevich 1955: 178 ff.; Hammel 1968; Hammel 1972, 1977, 1980a, b, 1984; Todorova 1993; Capo-Zmegac 1996.) Equally important in the present context was the effect of such households or their near analogs in households of closely related agnatic kin that were located in close propinquity, on the presence of multiple females associated with a core of male agnates, either as their wives or sisters. These women provided economies of scale in female labour, reciprocal child care, surrogate nursing, and mutual succour and assistance. While the universality or typicality of such households have been criticized as a facet of the orientalization and essentialization of ‘Balkan’ populations (Todorova 1993; Capo- Zmegac 1996), they were frequently attained during the cycle of household development, even when not common in the cross section. Even in medieval times households containing more than one conjugal unit were less than half of all households, in the cross section. The attainment of multiple conjugal units is strongly influenced by demographic conditions.
Thus, for example, the number of households containing a married father and a married son is a function of the overlap in the married lives of parents and children and thus strongly dependent on age at marriage and longevity. The number of households containing married brothers is strongly dependent on the net reproduction rate. Indeed, the virtual impossibility of the universality of fraternal joint families can be shown by the fact that for the average married pair to have two sons surviving to marry would require that the population double in each generation. (On these issues, see Halpern and Anderson 1970; Hammel 1990.)3.
This contention is disputed by some Croatian historians as a self-serving claim of Austrian apologists; nevertheless, the peasants of the time apparently evinced a preference for military status because feudal dues were less than in the civil zone, and the military establishment is sometimes reported to have provisioned villages in times of famine. To be sure, such reports are from Austrian sources. Much of our information about economic developments is general and insufficiently specific with respect to region. Regions were diverse. The Austrian sources give little systematic information on the local military border except 1830—47. Hungarian sources often do not differentiate between Hungary proper, Croatia, and Slavonia. There are no wage series, only some price series.
4.
The grain prices are regional prices based on the Vienna market and correlate closely with those of the Budapest market and (for a more limited time range) prices on the Zagreb market. Regional grain prices mostly reflect supply, conditioned largely by regional climatic factors. We have no information on local prices or the supply factors driving them. Neither do we have information on shifts in demand, plausibly driven by military provisioning in time of war. Peasants rarely bought or sold grain. They could buy it in time
of need only if they engaged in wage labour to raise cash; they could sell it in time of surplus to raise cash for taxes and amortization payments.
Grain was produced commercially on the large estates in the civil zone, but not in the military zone.5.
Elasticity is the proportional change in some quantity induced by a 100% change in another. For example, if the elasticity of mortality with respect to the price of grain were 0.2 (20%), a doubling of the price of grain would increase mortality by 20%. Elastiticies may be positive or negative. Such elasticities may be estimated over several time segments. For example, one may estimate the effect on mortality levels in year t of price changes in that and the previous 4 years, that is, at yearly lags 0, 1, 2, 3, and 4. The sum of these lagged elasticities, the ‘five-year lag sum’, is often used as a summary measure. See Hammel and Galloway (2000^, b) for an explication of the method and application to Slavonian data.
6.
7.
These are the heights of ‘taller Hungarian soldiers’, that is, those with heights above the historically highest lower boundary imposed. Komlos gives data for the decade of birth, for example, ‘1790’, thus with a decade centre at 1795. Under the assumption that nutritional deprivation would have its most lasting effects if it occurred during late pregnancy and before weaning, one would choose to centre the height data on the centre of the decade of birth. If, on the other hand, one assumed that the most important effects occurred during the adolescent growth spurt, when some ‘catch-up’ is possible, one would centre perhaps on age 15. I am obliged to Richard Steckel for his comments on this matter at the Arild conference, giving his preference for centring on the decade of birth. Of course, the exact mechanisms are unknown and differ for civil and military serfs. For civil serfs, feudal dues in the form of corvee labour were relatively light. Additional feudal taxes could be assessed. Church tithes were assessed and may not have changed at all. Only peasants with allodial holdings were subject to strictly feudal dues. Civil serfs also farmed so-called industrial land, land arrogated from the commons, in which the landlord also had rights, and paid a sharecropping portion to the landlord.
Such dues would not have changed with emancipation. The levels of these dues varied according to the practices of different estates. Military serfs had obligations as border guards after the Napoleonic wars, as well as in road maintenance, fortification maintenance, and haulage. These would not have changed until 1871. Such obligations also differed in practice between different parts of the Military Border.8.
The ethnographic, historical, and medical accounts suggest that herbal abortifacients were used and also that crude mechanical abortion was employed, often resulting in tetanus.
9.
Karl Kaser, personal communication.
10.
Theories about the origins and functions of complex family organization abound. The potential for such ‘patriarchal’ organization has deep roots in Indo-European social structure and kinship systems. The value of such organization to the households themselves in frontier situations has been noted above. There were also advantages to overlords seeking to maximize the availability of labour, especially in periods of peak labour demand, such as military mobilization or mass corvee labour on large estates, since a single adult male from such households could be left to tend the family plots while the rest were drawn off.
11.
Agnatic kin are those related by links through males. Patrivirilocal residence is a pattern in which a bride resides with her husband who resides with his father or other agnatic kin.
12.
For a useful summary of the economic issues, see Tomasevich (1955) and for infor mation on extant and divided households in 1890, see Zoricic (1894).
13.
If a household had no sons, the husband of at least one daughter might move into the household of the wife's parents.
14.
In the census of 1857 the Gradiska Regiment had a population of 56,402, of which 76% were Catholic, 24% Orthodox, plus 6 Uniates, 7 Protestants, and 3 Jews.
15.
Evidence from scattered libri status animarum and the Chronicle of the Monastery of Cernik indicates that infants were usually baptized at birth by the midwife.
We use ‘baptism’ and ‘birth’ synonymously.16.
Strictly speaking, this is an estimate of the probability of dying, not a rate. In the body of the analysis we use a mortality rate, defined as the number of deaths divided by exposure time in person-years.
17.
Loudon (1992: table 1.1) shows levels between 4.7 and 5.1 per thousand for England and Wales, 1851-1900. Wrigley et al. (1997: table 6.29) estimate rates in the range of 12-17 per thousand baptisms (corrected for stillbirths and no births, thus ‘per confinement’ and corrected for background mortality) for England 1600-1750. Rates for Germany, Sweden, and rural France, adjusted to achieve comparable definitions, were estimated by these authors to be in the range 9-12 per thousand between about 1700 and 1900.
18.
However, where rates of abortion and of fatal consequences of abortion are both high, some substantial fraction of maternal deaths may occur earlier, typically 3-6 months earlier. Similarly, deaths from ectopic pregnancy, eclampsia and some from infectious disease exacerbated by immune suppression will occur before parturition, not after. See Hogberg (1985: 1-22) for discussion and data.
19.
This chapter and others by Prof Ulf Hogberg and colleagues at Umea University are some of the most detailed analyses of the subject in the literature.
20.
in partu, in partus, paru, partu, Hificilis partus, parus dificilis, puerperium, in puerperio, in puerperium, in puerperis, partu infelico, post-partum, post-partum sanguini, peritonitis puerpera, pleuritis et puerper, sepsis puerperalis, texak porod (Cr. ‘difficult birth’).
21.
Notice that burial records are not written for unbaptized persons, so that there are several hundred cases in which baptism must have been immediate.
22.
Tables 57 and 58 of the census give regional breakdowns for the more common causes of death, but there are no data on those related to MM. Thus it is impossible to narrow the comparison to Slavonia or segments thereof.
23.
More precisely and for computational convenience, we count deaths and exposure time from birth up to the 61st day, and from birth (or from marriage for nulliparous women) to the 731st day. The two-year window gives a count of deaths from all causes. Subtracting the sixty-day set from the 730-day set gives deaths and exposure within the two-year but beyond the two-month window. The corresponding mortality rates are computed from these deaths and exposures. Note that what we compute is a mortality rate, m(x), not a ratio of deaths to live births or a crude rate of deaths per women in some age range. Note also that using a longer span after sixty days to estimate BM will begin to pick up deaths from pregnancy-related causes, such as abortion, ectopic pregnancy, eclampsia, as the end of the BM span begins to capture women returning to fecundability, that is, pre-parturitional deaths at the next parity.
24.
Reconstitution gives many examples of couples with linked children but no record of the marriage itself. Such ‘sibsets’, as we call them, because they consist of sets of persons with the same parents, could be the children of couples who married outside the parish and subsequently migrated in, or children to a marriage with no surviving record. We do not include such data here, because parity, age, and remarital status of mother are unknown. Similarly, although we can identify higher order marriages from the marriage
records, we have not yet fully automated linkages between successive marriages and thus do not include higher order marriages here.
25' We follow the traditional practice of assuming that a woman who did not demonstrably die within sixty days of giving birth actually survived to the end of the sixty-day period. We were concerned that we might in this way miss some women who might have emigrated and died outside the parish within sixty days, and conducted a parallel analysis in which we demanded evidence of survival in the form of some other event occurring to the mother. If that event were a last birth, we did not admit that birth to the analysis but used it only as a censoring event. The two forms of analysis did not differ importantly in their results. Similarly, we assume that women who did not die in the 61—730 day period following a birth did survive it.
26 Obviously, there is no difference between these intervals for women whose terminal parity is 1, and the expectation suggested will be clearest for women with higher terminal parities.
27 541 of 31,683 attested causes of death were for smallpox (variola, febris variolanus). Almost all of these were reported after 1850. How much of total adult mortality was smallpox related is not yet clear.
28' Bogicevci, Cernik, Nova Gradiska, Oriovac, Petrovo Selo, Stivica, and Vrbje.
2,' Note also that any aggregate analysis of MM may be sensitive to changes in overall fertility. If fertility is controlled by stopping behaviour, depending on the stopping point different proportions of women will bear children at parities of different risk. We do not attempt to take these factors into account but have opted to control for parity level at appropriate points in the analysis.
30 The probability of dying in childbirth, estimated as maternal deaths per birth, is the usual measure employed, sometimes per 1,000, or 10,000, or 100,000 births.
31' Logistic regression is one of several accepted methods for analysing dichotomous outcomes. Here the outcome is whether the mother dies or not. The coefficient for any independent variable is the logarithm of the relative odds ratio, or;
∏Λ√(i-⅛))
where ■ is the probability that the outcome (death to the mother) will occur to a mother with characteristic x. Confidence intervals for the coefficients are estimated in the usual way, as some multiple of the standard error of the coefficient.
32 We are indebted to Mike Hout for suggesting this.
33' The collection of birthings analysed here is not a formal random sample, so that the application of the traditional rules of inference can only be approximate.
34 It is important to note that in some degree, the movement of grain prices parallels our narrative of socio-economic change. One could plausibly offer a model in which grain price changes, in the short and long run, were the driving forces, rather than the structural changes in the labour market and in the households of peasant producers. Some changes in grain prices were of course correlated with the structural shifts. Grain prices sometimes increased in time of mobilization, either because of military buying or because of drops in supply as labour was drawn off the farm, or both. As peasants were drawn off the family farm into wage labour to raise cash for taxes and amortization, or to buy food as family farms became smaller through inheritance division, production may have dropped, again raising prices. Simply substituting grain price movements for a broader story of change seems to us to sacrifice historical richness for numerical precision.