The Judge Who Bounds Only Equals: Early Seating of a Judge of the Change, Imperial Favour and Bystander Loss in 203 Registered Kanly, 9719–10190 AG

1. Introduction

We find that a Judge of the Change seated within a year of a declaration of kanly reduced the odds that the feud would kill anyone outside the two Houses to about a quarter, provided neither House stood in Imperial favour. In feuds where either House did, an early Judge made no measurable difference to bystander loss or to the use of forbidden means.

The question comes from the registry study of Roquefeuil et al. (10242 AG), which read every closed declaration in the Landsraad Registry of Kanly and House Losses for 9719–10190 AG and found that feuds in which a Judge was seated recorded any bystander death in 30.2% of cases, against 52.5% where none was. From the registry alone they could not tell which way the association ran. A Judge might be seated because a feud was already heading toward settlement, so that the Judge followed restraint instead of producing it. And Judges might be seated more readily in feuds between Houses of comparable standing, which would spill outward less for reasons of their own. The registry gives neither the date a Judge was seated nor the rank of most Houses. Both are recoverable elsewhere. The commission files of the Judges of the Change at Kaitain date each seating. The Landsraad precedence rolls, issued every year of the window, place every House. Linking them to the registry allowed us to fix the timing of each Judge and the standing of each pair, and, through the Imperial grant registers, to ask a third question the registry study did not raise, namely whether either House stood in the Emperor's favour when the feud began. Our analysis of the linked records was completed in 10247 AG.

2. Methods

The starting set is the 284 feuds declared in 9719–10190 AG that had closed by 10190 AG (Landsraad Registry of Kanly and House Losses, 9719–10190 AG; Roquefeuil et al., 10242 AG). For each we sought three linked records. Rank at declaration came from the precedence rolls (Landsraad College of Heralds, 9719–10190 AG) and was recovered for 231 feuds. The commission files (Landsraad High Council and Imperial Chancery, 9719–10190 AG) give the date on which a Judge was seated or confirm that none was; one or the other was recovered for 219. Both were available for 203 feuds, which form the analytic sample. Linkage used the House names and declaration dates of the registry entry and was checked by hand against the herald's name in each.

Early seating, the exposure, means a Judge commissioned within twelve months of the declaration. Feuds with a Judge seated later, or never, form the comparison group. Twelve months was fixed before linkage, as the interval within which a declaration's first renewal fell due (Roquefeuil, 10228 AG). Early seating occurred in 108 of the 203 feuds.

Imperial favour was coded from the Imperial Chancery's registers of grants (9699–10190 AG). A House stood in favour if, in the twenty years before the declaration, it had received a fief or an office by direct Imperial grant, a CHOAM directorship recorded in the Chancery's register as held at the Emperor's instance, or a marriage into House Corrino (Carrowmore, 10235 AG). Either House in favour placed the feud in the favoured stratum; this was true of 61 feuds. Rank asymmetry was coded as a pair in different precedence tiers or separated by two or more places within a tier, following evidence that forms of address in council tracked recognised standing (Aldevash and Reyes-Okafor, 10240 AG), though precedence lagged deference in the one case they document; 89 pairs were asymmetric.

The outcomes are those of the registry study, namely any death recorded under the third head of a closing return (persons belonging to neither House), and any finding of forbidden means. For bystander death we fitted a logistic model with early seating, Imperial favour, their interaction, rank asymmetry and three declaration periods (9719–9875, 9876–10032 and 10033–10190 AG), reading the interaction as a ratio of odds ratios (Achalde, 10238 AG). Findings of forbidden means are reported as exact two-by-two tests within strata, because 28 events do not support the full model; all effects are odds ratios with 95% confidence intervals. Every estimate was computed by the third author without written aids and again on paper by the first, and the two agreed to the second decimal place.

3. Results

Of the 203 feuds, 83 (40.9%) recorded at least one bystander death and 28 (13.8%) drew a finding of forbidden means (Table 1). Among the 142 feuds between Houses outside Imperial favour, any bystander death was recorded in 17 of 78 with an early Judge (21.8%) and 35 of 64 without one (54.7%; odds ratio 0.23, 95% CI 0.11–0.48; exact test p < .001). Among the 61 feuds in which either House stood in favour, the proportions were 15 of 30 (50.0%) and 16 of 31 (51.6%; crude odds ratio 0.94, 0.34–2.56; p = 1.00). The model gives the same division (Table 2). Outside favour, an early Judge carried an adjusted odds ratio for bystander death of 0.23 (95% CI 0.11–0.48, p < .001). Within favour, the corresponding odds ratio was 0.94 (0.34–2.58, p = .91). The ratio of the two odds ratios, which measures how far favour cancels the Judge's effect, was 4.06 (1.17–14.14, p = .028). Rank asymmetry, the explanation the registry study could not exclude, carried an odds ratio of 1.54 (0.83–2.83, p = .17), and adding it changed the early-seating estimate only in the third decimal place. Declaration period made no material difference (odds ratios 0.99, 95% CI 0.48–2.04, and 0.96, 0.47–1.99, against the first period).

Findings of forbidden means divided along the same line. Outside favour they fell from 11 of 64 feuds (17.2%) without an early Judge to 2 of 78 (2.6%) with one (odds ratio 0.13, 95% CI 0.03–0.60; p = .003). Within favour, 7 of 30 (23.3%) and 8 of 31 (25.8%) drew such a finding (odds ratio 0.88, 0.27–2.81; p = 1.00). What favour changed was the feuds with an early Judge, which it left as violent as those without.

Two checks bear on the timing problem. Feuds that closed within twelve months could only have had an early Judge or none, so we excluded the 21 such feuds outside favour; the early-seating odds ratio became 0.19 (0.09–0.42; 16 of 69 against 32 of 52). The long feuds of the registry also sort by favour, since 16 of its 19 feuds of fifty years or more fall in the linked sample, and 11 of the 16 are in the favoured stratum. Favour also had a direction, in that the declaring House was the favoured one in 38 of the 61 favoured feuds.

4. Discussion

The Judge of the Change bounds a kanly. Seated early, in a feud between Houses outside Imperial favour, a Judge cut the odds of killing outside the Houses to about a quarter and the odds of forbidden means to about an eighth. The registry study's preferred reading was right for these feuds, and its alternative fails, because parity of rank, as the precedence rolls record it, does not account for the effect, which survives adjustment for rank unchanged. The effect also survives removal of the briefest feuds.

Imperial favour removes the effect. A Judge seated early in a feud where one House stood in the Emperor's favour left bystander loss and forbidden means close to their levels with no early Judge. Within favour the odds ratio was 0.94 (0.34–2.58), so the data exclude an effect as large as the one outside favour, and the interaction (p = .028) shows that the strata differ. Favour was also over-represented among the registry's long feuds, the nineteen that carried almost all of its bystander deaths, with 11 of the 16 in the linked sample, against 61 of 203 feuds overall. The division found here reaches into the long tail the registry study described.

The commentary on kanly has located its restraint in the forms, the herald's notice and the interval before violence (Vesk, 10232 AG). The forms were observed in favoured and unfavoured feuds alike. What differed was whether the arbiter the forms provide could act on the parties in front of him.

5. Limitations

The linked sample is 203 of 284 feuds. The 81 unlinked feuds are older on average, since precedence rolls and commission files survive less well for the first two centuries of the window, and the result applies to the linked period with most force. Imperial favour is coded from formal grants and marriages; favour conferred informally, by audience or by unrecorded patronage, is missed, which would place some favoured feuds in the unfavoured stratum and weaken, not create, the division we report. Bystander loss is reported by the Houses themselves, as in the registry study. The outcome counts deaths over the whole feud, so some preceded the Judge's seating; removing the briefest feuds narrows, without closing, the possibility that Judges were seated in feuds already settling. Findings of forbidden means are entered by Judges, so a seated Judge is in a position to find more than an absent one, which biases against the effect reported.

The finding has a boundary, in that it shows that Imperial favour cancels the effect of an early Judge; it does not show why. A Judge might be unwilling to rule against a House the Emperor favours, or unable to make a ruling hold, or a favoured House might expect any adverse ruling to be overturned. The data cannot distinguish these. The question this paper leaves open is the doctrinal one: what, in the law that appoints a Judge and reviews a Judge's rulings, gives the Emperor's favour this power over the arbiter of a feud.