Evidence from an Early Census
DRAFT 22 OCT 95
E. A. Hammel
H.-P. Kohler
Before the 17th century it included
Moslavina, which lies west of the Ilova, and extended along the upper Sava as
far as Zagreb. The region had been the western part of Roman Pannonia, with
its capital at Siscia (modern Sisak). Overrun by Goths and others in the fifth
century AD, it was settled by Slavs around the 6th and 7th centuries and
controlled in various degrees by Croatian and Bosnian nobles until the 12th
century. In 1122 AD the Croats came under the control of the Hungarian crown.
In the late 13th century the Ottoman Turks began their drive into Europe,
taking Belgrade in 1521 and annihilating the Hungarian forces at Mohacs on the
Danube in 1526. At this point the Croats came under Habsburg control, and
defensive zones were established along the Alpine and Dinaric foothills.
Nevertheless by 1683 the Turks controlled Slavonia to the Ilova and were at the
gates of Vienna. Routed there in that year, they were driven back to the Sava
by 1691, and the prosperous population of about a quarter million persons in
that region was reduced to about 80,000. Refugees from other Habsburg and
Hungarian regions, and both Orthodox and Catholics from Bosnia and Serbia
migrated in large numbers to Slavonia and other parts of Pannonia. In 1698 the
Austrian Crown commissioned a census of Slavonia to establish the basis for
taxation and conscription, in defense against the continuing Ottoman threat.
Ultimately the population was divided into two parts, one of civil serfs in a
resurrection of mediaeval serfdom, the other of military serfs free of the
usual feudal dues but obliged to provide perpetual military service, in a
resurrection of a system that had manned Roman forts as far back as the third
century. These military serfs were settled in the Military Border, which
before 1683 lay west of the Ilova and extended toward the Drava but which after
1683 was extended along the Sava to the Danube and ultimately into the
Carpathians. In the late 17th century Slavonia was largely wilderness, its
once settled portions devastated by war, with still enormous swamps, and stands
of giant hardwoods like oak and beech surviving from the primeval European
forest.
The census information is supplemented by data on the latitude, longitude, and altitude of the villages, most of these being identifiable on modern maps.[1] Because the religion of the inhabitants of the villages is often noted in the prose summary, it is also possible to assemble a list of first and last names that occur only in homogeneously Orthodox or homogeneously Catholic villages, in both of these kinds of villages, and in villages of mixed or unspecified religious composition.[2] From this evidence we are often able to impute religion at the household level, according to the name of the head. Similarly, the prose summary often gives an indication of whether the inhabitants of villages were military or civil serfs, or were inclined to accept one or the other status; we used this information as a village-level variable. Some villages, and some inhabitants within some villages, are identified as having come from Bosnia, but this information is not consistently provided, and we use it only tentatively.
We also use last names in another way, to impute kinship relatedness. This leap of faith is justified by family reconstitution records from c. 1720-1900, in which last names are fairly regularly inherited in the paternal line.[3] Our assumptions could be upset if the ancestral Slavic system of assigning patronymics as last names, based on the baptismal name of the father, were in full force, but it appears not to have been.[4]
Initial statistical examination of the census data showed that households held on average about 3 yokes of grainland[5], variously distributed across different kinds of grain, a row and a half of grape vines, about 2 yokes in hayfield, around 1 ox, 1 cow, 1 calf, 3 sheep and goats, 2 pigs, and a beehive, plus 1 horse for every two households (Table 1).[6] The distributions are all sharply skewed; most medians are zero. There is little or no evidence for specialization or substitution. For example, the correlation coefficients between all forms of livestock are positive (Table 2). This circumstance argues for a general differentiation of the population by wealth, with the richer having more of all kinds of stock, and the poorer less. Similarly, there is no strong evidence of substitution in crops (Table 3).[7] A more technical analysis (Kohler and Hammel 1996) also shows no evidence of specialization or the possibility of interhousehold trade. The only apparently strong exception is the negative correlation between frumentum on the one hand and milli and tritici on the other. Milli and tritici are millet and wheat, while frumentum means grain in general (like the German Korn); it may have been a synonym for either millet or wheat, although it appears not to have been for oats (avenae) or barley (hordei). Thus, this substitution appears to have been performed verbally by the census takers, not actually by the peasants.
There is some weak support for the idea that maize (kukuruz) was a substitute for other kinds of grain, from statements in the prose descriptions (see below). Maize was introduced to the Balkans by the Ottomans probably in the 16th or 17th C. The Slavic term, kukuruz, is derived from Turkish kokoroz, possibly from -oroz (rice) and koko- (stench), thus the "rice of the lower classes." Note the alternative Serbian term, mumuruz, possibly from -oroz and mumu- (with meaning parallel to koko- ); see also the alternative term from Rumanian, mamaliga (Skok 1972: 228-229). The census gives no evidence of the cultivation of the potato, but it is likely it had recently been introduced. The Slavic term, krumpir, is from German Gruntbir , Grundbirne (ground pear, cf. pomme de terre Skok 1972: 215). The potato was cultivated in Spain by the third quarter of the 16th C (Salaman 1949: 143) and possibly introduced to Austria by the Habsburgs. Most of the scanty evidence suggests it reached the Balkans after maize (J. Capo, personal communication 1995). Even the Austrian censuses of 1830-47 do not mention the potato, but neither do they mention any other tubers or vegetables such as turnips, cabbage, etc., which were surely being grown. There is thus every reason to believe that some households were growing potatoes and other garden crops in addition to grain, and to conjecture that some may have been engaged in essentially swidden horticulture without growing grain at all. Thus it appears that the census concentrated on economic assets that were felt taxable or capable of commercial exploitation; this area did engage later in extensive commercial grain production on the large estates. It is also of interest to note a weak but significant negative correlation (r2 = .10, p < .0001) between the percentile rank of households in a village list and the total amount of land devoted to grain cultivation; that is, there is some weak ordering of households by wealth, the poorer being listed last.
Travellers through the region in the 19th century generally deplored the state of agricultural practice, especially in the Military Border where there was no commercial development but only subsistence agriculture. The accounts suggest that agriculture was extensive rather than intensive. Certainly in 1698 there appear to have been too few farm animals to provide enough manure for regular fertilization. Fragmentary evidence from the chronicle of the monastery of Cernik from an even later date (Jancula 1980) suggests that cattle were usually pastured in the commons or waste; without stall-feeding it is difficult to recover enough manure for fertilization. Similarly, sheep must be carefully penned on the grainfields to utilize their manure; there is no evidence of this practice in Slavonia at all at any time. The large swine herds later characteristic of the region seem not have appeared by 1698, and even in the 19th century swine were pastured mostly in forest, not on stubble. To be sure, the relationships that emerge from our analysis cannot take into account that realm of economic activity unrelated to major field crops, but we believe that they provide a reasonable description of grain agriculture.
Simple OLS regressions of the various kinds of grain or of all grain in general on potential factors of production, show that plausible independent variables have a positive effect. Households with abundant male labor tend to be those with more animals and more land under cultivation. This result, of course, could be no more than the general wealth effect already noted (Table 4). But there are intriguing clues in the prose summaries that led us to look further. Here are some examples:
Incolae hi haidonicales pro exercenda sua rurali oeconomia et terreno incolendo per defectum pecorum insufficentes sunt, ex eo vicinos fundos non usuant. (These military serf inhabitants have insufficient cattle for the exercise of their rural economy and inhabited terrain and on account of that do not use neighboring lands.) Sentences of this meaning, with numerous variations, are extremely common and almost universally found in some districts. Sometimes the statement only says that the inhabitants are incapable of cultivating their land, without mentioning the lack of cattle, but such statements are much rarer. It is extremely rare to find the statement that the inhabitants of a village have sufficient capacity to use the land of a neighboring village. The collective noun pecus (gen. pl. pecorum) means "stock" and could refer to any of the animals listed in the census. However, sheep, goats, pigs, and bees are not employed in working the land. Similarly, although horses and cows can be used to pull a plow, their use is rare except on light soils (or where plow horses like Belgians have been bred for the purpose). The best horses in Pannonia in historical times have usually been Hungarian, and they are riding horses, not plow horses.[8] Thus, it seems most likely that the lack of cattle refers particularly to oxen, who were the main source of power for plowing in mediaeval Europe and in 19th century Croatia. Oxen would have been important in Slavonia also for pulling stumps in the extensive forest clearing that was necessary prior to cultivation of field crops (but not garden crops). We concentrate on the oxen as a limiting factor of production.
A second clue is an occasional sentence that refers to annual floods of the Sava, e.g. Locus hic inter palludes et sylvas alninas collocatus.... (This place is located between swamps and alder forests). (Note that alders prefer damp, even swampy ground.) ...Quamvis fundus hic tam palludinosus sit per inundationes Savi....(This land is very swampy because of the floods of the Sava...). In the chronicle of the monastery of Cernik in the 18th century there is evidence that these swamps were malarial (Jancula 1980), and later travellers in the region also noted such conditions in Pannonia (e.g. "Banat fever" after the neighboring region of Banat in Pannonia). We use information on the altitude of villages to approximate these circumstances, which would have made tillage more difficult, not only because of disease but because of drainage and other soil problems.
A third clue is the rare sentence that suggests that inhabitants in forested
regions and lacking oxen might plant maize in the clearings as a substitute
for other grains; e.g.... ob defectum iumentorum sunt incapaces, suntque
meri fossores kukurczarii...(...because of a lack of plow oxen they are
incapable [of field agriculture] and only [plant] maize in clearings...). This
suggests the modest substitution noted in the correlations earlier.
The general wealth effect can be seen in OLS regressions of household grain land on the major factors of household production, animal power and manpower, for which we take the number of oxen and the number of adult males as indicators, as noted in Table 4. To avoid the effect that mere household size would have on these relationships we also regress per capita grainland on the number of oxen and males in Table 5 (per capita here meaning male household members listed). In Table 5 we see that additional oxen are a positive factor, while additional males are a negative factor. The marginal productivity of oxen appears to increase with additional oxen, but but the marginal productivity of human labor appears to decrease with additional labor. The "ethnic" variables are of interest. The omitted category is the small number of households living in mixed Orthodox-Protestant villages. Families with homogeneously Orthodox names have about the same per capita level of grainland as those living in mixed Orthodox-Protestant villages (where they dominate). Those families with names occurring in mixed Orthodox-Catholic villages have somewhat less, and families with names that occur in homogeneously Catholic villages have the least. This result confirms that the Catholics were on the whole poorer than the Orthodox.
These simple initial results also lead us to the indication that there was some
optimal ratio of men to oxen for plow teams, such that adding additional males
to some fixed number of oxen diminished per capita productivity. We might
imagine that this effect was simply an artifact of having more males in the
denominator of the per capita grain measure. However, if an additional male
had marginal productivity equal to the previous average, the effect of adding
another male would be zero. If an additional male had marginal productivity
above the pre-existing average, the effect of adding another male would be
positive. This might happen if there were returns to scale in important tasks.
However, the result of adding an additional male is negative, as shown;
additional males depress per capita productivity, given some number of oxen.
Unless an additional man brings additional ox power with him, he lowers the per
capita productivity of the enterprise. The optimal ratio of men to oxen
appears to be quite low, around 1:3 (see below).
Y = A * Ob0 * Mc0
However, patterns of interhousehold cooperation are common in agrarian societies. Such cooperation is often based on kinship. While in most rural societies there is virtually no rental market in livestock because of the potential for abuse of animals rented out, trust and mutual obligations between kin may overcome this barrier. This led us to speculate that the oft-mentioned insufficiency of oxen (defectum pecorum) might have been ameliorated by the lending of oxen between kin-related households. Thus we might imagine that a brother would lend an ox to a brother who had none, or to a brother who had some, in supplementation. Brothers who had some oxen but not enough might help eachother seriatim. We might imagine that human labor might reciprocate loans of ox power, either if the recipient had no oxen at all, or in supplementation if he did. We might imagine that human labor would be also reciprocally exchanged.
Oxen are useful in plowing, harrowing, stump pulling and other intense and relatively short-term activities in preparation of the land for sowing. Apart from providing traction power for haulage, they are not important in harvesting. Their use is thus episodic, but they must be maintained year-round. Oxen cannot plow alone, of course, and at least one man is necessary to guide the plow. The constraint on plowing is the weight and maneuverability of the plow, which must be handled by the man or men. Plows useful for heavy soils can be quite cumbersome. Plow teams of 2, 4, 8, and perhaps more oxen can be managed by one or two men. In 19th century Slavonia, at which time some detailed ethnographic information is available, but by which time grain agriculture was much more developed, plow teams consisted of from 2 to 8 oxen (J. Capo, personal communication 1995). An additional man, or more commonly a boy, can lead the oxen. There is a potential reciprocal relationship with grain land, because oxen are sometimes pastured on fallow land and stubble in addition to grazing on pasture and waste land. Thus there may be some problems of reciprocal causation in analysis, but we do not attempt to explore them here.
Male labor power is used not only for plowing grainland but also sowing grain,
scything grain, transporting the harvested stalks, and threshing and storing
grain. Weeding, raking, bundling, and sometimes sowing are usually done by
women and children, but we did not try to take account of female labor, since
it would be closely correlated with male labor in a household economy in which
virtually all adults were married.
We allowed putative kinship to decay exponentially with distance, so that for
some household with last name X, another household with name X in the same
village was assumed to be in the same kinship network, but inclusion in the
network fell off rapidly as households with the same name were found further
and further away. Experiments showed that the rate of exponential decay made
little difference to the outcome; it is the shape of the function that is
important.
Table 6 gives the results from the Cobb-Douglas production function, in which Y is per capita grainland[10], A includes a constant ([[alpha]]0), coefficients for altitude ([[alpha]]1), altitude squared ([[alpha]]2), and a dummy for Orthodox ([[alpha]]3), (with all non-Orthodox and those of unknown religion grouped together). d0 and d1 are the coefficients for having more or less oxen than the optimal ratio to human labor. [[theta]]0, [[theta]]1 and [[theta]]2 are the coefficients for the effects of having oxen in the household, in the network and outside the network, respectively. [[phi]]0, [[phi]]1, and [[phi]]2 are the analogous coefficients for male labor.
The principal results are as follows:
* The amount of grainland increases with altitude but decreases with altitude squared, showing that more extensive cultivation and productivity were possible above the swampy bottomlands but below the higher, mountainous regions.
* The Orthodox households are have more grainland than other groups; this advantage amounts to about 7 percent, all else equal.
* The mean optimal ratio of labor to oxen is 1.18 men per ox or conversely 0.85
oxen per man. If the average household could optimize its possession of oxen
it would have
oxen where Li is the number of males. This implies that 1626 of the
2427 households (about two thirds) have a deficit of oxen with respect to their
male labor force, while only one third have the optimal ratio or more. It can
be shown that the optimal ratio of oxen to males increases with altitude,
meaning that households with the best soil for grain farming have relatively
more oxen per male worker. They benefit twice; they have better soil to
cultivate, and they have more oxen to do it with.
* The coefficients [[theta]]0 and [[phi]]0 add to one. This means that the
optimal demand for oxen
and the amount of cultivated land T* vary proportionally with the number of
males in the household. Thus, there are constant returns to scale, as we would
expect under conditions of extensive, undeveloped agriculture in which the
technical means of intensification (other than increases in the labor supply)
were unattainable or in which the additional land coming under cultivation was
as yet of no poorer quality than the core holding.
* Households with a relative excess of oxen (more than the optimum) seem to engage in an exchange of male labor, but having males in the network seems to confer no particular advantage. Thus it seems that the traditional patterns of cooperative labor exchange (Croatian sprega) crossed kinship lines. The average household with an excess of oxen gains about 1.49 workers in help from other households.
* On the other hand, households with a deficit of oxen compared to their labor force seem to benefit from having oxen in their kinship network but get little or no benefit from having oxen just in their neighborhood. These results for the exchange of labor and of oxen confirm the expectations grounded in the idea of trust prevailing between kin more than between neighbors.
* Oxen are much more important determinants of the extent of cultivated land than are human workers; based on the coefficients [[theta]]0 and [[phi]]0, oxen are about three times as important.
* Similarly it can be estimated that the relative prices of oxen to labor are in excess of 2.5:1
Some of these results can be shown graphically. Fig. 2 shows the output and marginal productivity of oxen, against the ratio of oxen to workers. As the number of oxen per worker increases, output (per capita grainland) increases almost linearly. The marginal productivity of oxen, per one worker, declines, but only modestly.
Fig. 3 shows similar data for the output and marginal productivity of labor, per ox. Output per laborer increases more slowly and with more obvious flattening than output per ox (cf. Fig. 2). The marginal productivity of labor declines much more rapidly than that of oxen (cf. Fig. 2).
All of these results confirm the textual evidence of the census but provide an
analytic view that is not apparent merely in statements about "a lack of oxen."
Table 1
Descriptive Statistics
N= 4453
Mean Std. Std. Count Minimum Maximum
Dev. Error
Rr .383 .486 .007 4453 0.000 1.000
Ro .329 .470 .007 4453 0.000 1.000
Rpo .009 .097 .001 4453 0.000 1.000
Rro .124 .329 .005 4453 0.000 1.000
Cm .454 .498 .007 4453 0.000 1.000
Ob .026 .159 .002 4453 0.000 1.000
Equi .425 .702 .011 4453 0.000 6.000
Boves .936 1.076 .016 4453 0.000 10.000
Vaccae 1.012 .895 .013 4453 0.000 8.000
Vituli 1.258 1.304 .020 4453 0.000 16.000
Ov&Cap 2.662 6.313 .095 4453 0.000 80.000
Porci 1.829 3.329 .050 4453 0.000 52.000
Alvearia .786 1.749 .026 4453 0.000 30.000
Vin.Foss. .689 1.271 .019 4453 0.000 12.000
Frum. 1.396 1.878 .028 4453 0.000 15.000
Hordei .210 .504 .008 4453 0.000 12.000
Avenae .293 .605 .009 4453 0.000 12.000
Currus 1.830 2.018 .030 4453 0.000 20.000
Kukuruz .187 .433 .006 4453 0.000 6.000
Falcator .051 .441 .007 4453 0.000 6.000
Milli .391 .943 .014 4453 0.000 20.000
Tritici .669 1.505 .023 4453 0.000 20.000
Terr. in. 1.031 2.636 .039 4453 0.000 28.000
Grain 3.146 2.950 .044 4453 0.000 33.000
Males 2.291 1.105 .017 4453 1.000 14.000
Big Stock 2.373 2.219 .033 4453 0.000 22.000
p/cGrain 1.523 1.563 .023 4453 0.000 23.000
Smallstock 4.490 8.133 .122 4453 0.000 95.000
AllAnimals 8.907 10.736 .161 4453 0.000 118.000
p/cSmallstock 2.075 3.903 .058 4453 0.000 72.000
p/c Bigstock 1.157 1.246 .019 4453 0.000 22.000
p/c Animals 4.218 5.347 .080 4453 0.000 118.000
Table 2
Correlation Matrix for Animals
N = 4453
Equi Boves Vaccae Vituli Ov&Cap Porci Alvearia
Equi 1.000 .493 .416 .396 .323 .388 .210
Boves .493 1.000 .626 .573 .434 .468 .212
Vaccae .416 .626 1.000 .763 .237 .421 .227
Vituli .396 .573 .763 1.000 .180 .408 .205
Ov&Cap .323 .434 .237 .180 1.000 .362 .144
Porci .388 .468 .421 .408 .362 1.000 .307
Alvearia .210 .212 .227 .205 .144 .307 1.000
Table 3
Correlation Matrix
Crops
Vin.Fos Frum. Hordei Avenae Currus Kukuruz Falcato Milli Tritici Terr.
s. r in.
Vin.Fos 1.000 .291 .106 .215 .237 .017 -.042 .081 .067 .086
s.
Frum. .291 1.000 .116 .214 .246 .221 .136 -.261 -.330 -.267
Hordei .106 .116 1.000 .201 .250 .005 .056 .159 .248 .220
Avenae .215 .214 .201 1.000 .294 .005 .027 .267 .358 .266
Currus .237 .246 .250 .294 1.000 -.037 -.105 .201 .313 .299
Kukuruz .017 .221 .005 .005 -.037 1.000 .219 -.162 -.192 -.169
Falcato -.042 .136 .056 .027 -.105 .219 1.000 -.048 -.052 -.046
r
Milli .081 -.261 .159 .267 .201 -.162 -.048 1.000 .603 .492
Tritici .067 -.330 .248 .358 .313 -.192 -.052 .603 1.000 .642
Terr. .086 -.267 .220 .266 .299 -.169 -.046 .492 .642 1.000
in.
Table 4
Regression Summary
Grain vs. 6 Independents
Count 4453
Num. Missing 0
R .730
R Squared .533
Adjusted R Squared .532
RMS Residual 2.018
Coefficient Std. Error Std. Coeff. t-Value P-Value
Intercept .801 .095 .801 8.402 <.0001
Rr -.534 .091 -.088 -5.859 <.0001
Ro .064 .096 .010 .672 .5019
Rro -.435 .116 -.049 -3.755 .0002
Cm .147 .063 .025 2.317 .0205
Boves 1.842 .029 .672 62.844 <.0001
Males .345 .029 .129 12.090 <.0001
Table
5
Regression Summary
p/cGrain vs. 6 Independents
Count 4453
Num. Missing 0
R .626
R Squared .391
Adjusted R Squared .391
RMS Residual 1.220
Coefficient Std. Error Std. Coeff. t-Value P-Value
Intercept 1.974 .058 1.974 34.260 <.0001
Rr -.284 .055 -.088 -5.152 <.0001
Ro -.011 .058 -.003 -.194 .8462
Rro -.174 .070 -.037 -2.479 .0132
Cm .018 .038 .006 .465 .6421
Boves .874 .018 .602 49.307 <.0001
Males -.499 .017 -.353 -28.910 <.0001
Table
6Production Function
Log likelihood = 391.7
N Observations = 2427
Parameter Estimate Std.Err. t
Constant [[alpha]]0 .2101 .0563 3.73
Altitude [[alpha]]1 .00313 .000048 6.49
Altitude2 [[alpha]]2 -.443E-5 .816E-6 -5.43
Orthodox [[alpha]]3 .1008 .0346 2.91
d0 see footnote 1 2.2512 N/A N/A
d1 see footnote 1 -1.0262 N/A N/A
[[phi]]0 HHoxen .7632 .0330 23.09
[[phi]]1 NWoxen .0499 .0200 3.45
[[phi]]2 NNWoxen .0083 .0002 2.9
[[theta]]0 HHmales .2736 .0398 6.88
[[theta]]1 NWmales .0187 .0762 .24
[[theta]]2 NNWmales .0028 .0008 3.72
[[sigma]]2 (error) .2664 .0069 38.85
Note 1: the coefficients d0 and d1 are used to estimate the optimal ratio of oxen to labor.
Note 2: the coefficients [[phi]]0 and [[theta]]0 can be taken as the coefficients for HHoxen and HHmales, respectively, if the values of NWoxen and NWmales, and of NNWoxen and NNWmales, respectively, are held constant.
Hammel, E. A., and K. W. Wachter, 1996a, Evaluating the Slavonian census of 1698. Part I: structure and meaning, European Journal of Population (forthcoming), .
Hammel, E. A., and K. W. Wachter, 1996b, Evaluating the Slavonian census of 1698. Part II: a microsimulation test, European Journal of Population (forthcoming), .
Jancula, J., 1980, Povijest Cernika i cernicka samostanska kronika (Graficka Radna Organizacija "A. Tajkov", Slavonska Pozega).
Kohler, H.-P., and E. A. Hammel, 1996, "A model for agricultural production: Slavonia, 1698," , In preparation).
Mazuran, I., 1988, Popis naselja i stanovnistva u Slavoniji 1698. godine, Vol. 2, Radovi Zavoda za Znanstveni Rad u Osijeku (Hrvatska Akademija Znanosti i Umjetnosti, Osijek).
Salaman, R., 1949, The history and social influence of the potato (Cambridge University Press, Cambridge).
Skok, P., 1972, Etimoloski rjecnik hrvatskoga ili srpskoga jezika, Vol. 2, ` (Jugoslavenska Akademija Znanosti i Umjetnosti, Zagreb).