HO\Uø ARTIFICIAL WERE THE TAX ASSESSMENTS OF

HO\UøARTIFICIAL WERE THE TAX ASSESSMENTS
OF
DOMESDAY ENGLAND? THE CASE OF ESSEX
By
JOHN MçDONALD and c. D. SNOOKS
î
¿
Reprint from
THE ECONOMIC HISTORY REVIETø
sEcoND SERIES,VOLUME XXXVilI
No. 3, AUGUST, 1985
Werethe Tax /,ss¿ssments
of Domesday
EnglandArfficial? The Caseof
Essex.
By JOHN McDONALD and c. D. SNOOKS
there is a difference of opinion abour the main purpose of the
\Y/!il.
w Domesdaysurvey-whether it was of a fiscal or a feudai naìure-there
is generalagreementthat the assessments
for the geld (often calleddanegeld)
were an important part of the data collected in rog6.1 This data sãt ii
potentially valuableasa detailedsourceof information concerningthe manner
in which Anglo-Norman kings raiseda significantpart of their r-evenue.The
conventionalwisdom on rhis question, which haslts origins in the work of
the Victorian scholarJ. H. Round, is that theseassessmeñts
were "artificial"
in the sensethat they were imposedfrom abovewith tittle or no referenceto
t\e_egogomiccapacityof manorsro pay this impost. yet, despitethe longevity
gf F.r interpretation, rhereare gooda priorigróundsfor dou-btingits vaúdity-.
It is.hard to imagine an arbitrary taxãtion õystemsurviving foi almost two
hundred years, as did the geld, in view of the inevitable ðostsinvolved in
suppressingwidespreadopposition. Even if the suppressionwere effective,
thesecostswould have drasticallyreducedthe benefiisto be gainedfrom thó
tax's imposition. It was this doubt which led us ro tesr, üsing regression
analysis,the relationshipbetweent¿rxassessments
and the capacityoimanors
to pay this tax (measurêdin terms of both their resourceendowmentsand
their revenues).The resolution of this problem is of signiûcancebecauseit
yill make possiblethe explanation,not only of the nature of geld assessment
in ro86, but alsoof the way in which the assessment
systemevolvedbetween
A.D.99r and n62.2
. Th. data analysedin tÞis paperinvolve the 695 observationsfor lay esrates
in the County of Essex.This county was chosenboth becauseof the greater
*- Tlút anicle is part of a targer study
of the economicsof DomesdayBook, which is being funded by
.
both the Australian Rese¡¡ch Grants Schemeand Flinders University. furtór results from-this proieci
gn be found in J. McDonald and G. D. Snooks, 'The Deterurina¡ts of Manorial Income in Domesiay
England: Evidence from Essex',Jolracl of EconomicHisøry, 45 Çg8). Valuable assisunce,t"g.t¡* øtú
heþful suggestions,have beenprovided by Ms E. Aker and Á{s'8. äckers. Heþful commenrswere also
p"dJ _U_v
$r J. Carnnbe¡I,PjoI: P_.D. 4.. Harvey, Dr J. Harcher, prof. G. R. Èawke, prof. J. C. Holr,
!r9f. H. R. Loyn and Dr G. H. Martin in partituhr, and by participants in semi¡a¡s at Corpus Chrisú
Cgllege, Cambridge, the Institute of Historic¡l Researchitt tle-Uoiu.i.ity of London, and the'University
of Durham.
1_For conuastingviewssee H. Round, Fadat
Erytand (rg95), and F. v. Maìtland, DomcsdøyBooh.
J.
qP Fq^¿ (cambridge, r8g:) on rhe one hand, and v. g. d¿ur¡o,
Ttu MaÞing oj oorusdaj Boon
(Oxford, 196r), on the other- For a recentevaluation, seeS. Harvey, 'DomesdayBook and its Predecãssors',
Englith Histori¿al Ret)i¿u,86 (r97r), pp. 753-73.
2 See McDonald and G. D. Snooks,'Modelling
J.
MedievalTax Systems:An Explanationof the Nature
and Evolution of Danegeld, 9gr-tr6z' (forrhcoming).
352
TAX ASSESSMENTS
OF DOMESDAYENGLAND
353
amount of detail availablefor the tlree countiesin Little DomesdayBook
(LDB),3 and becauseof its fortunate escapefrom the devastationwhich
followed the conquestin ro66. Vhile there is little reasonto believethat the
results from studying Essex are not relevant to other counties, it will be
important to extend our analysisto include other regions.a
I
At the time of the Domesdaysurvey, public revenuewas raised from a
variety of feudal and non-feudalsources.This revenuewas.in addition to the
feudal servicesowed to the king, such as knight service and attendance
at Court. Feudal revenue took the form of aids (arising from marriages,
knighthoods, and wardships),and reliefs (paid on the inheritanceof a fief),
while non-feudal funds came from the royal demesne,various local and
national courts, rights over boroughsand minting, and finally the geld.s
The geld played an important part in the provisionof total public revenue.
After an examinationof the rights and obligationsof Norman barons,Sidney
Painter concludedthat "'fhe most important public burden of the Norman
period was the tax known as danegeld".6He showedthat in 116z, the last
year in which the geld was levied, an amount of.f4rt3z was collected,which
compared more than favourably with the fzr4o8 raised from "the most
profitable feudal aid of Henry's reign".7 M. M. Postanalso claimed that at
the sametime the geld conuibuted betweenone-quarterand one-third of the
king's annual revenue.sThese conclusionsare supported by our study of
Essex:the rate struck in ro83/84, of six shillings to the hide (which was the
trnit of geld assessment)
implies that the averagemanor was required to part
with between 15.6 (mean) and 16.3 (median)per cent of its annual revenue
to meet this impost.s The geld, therefore,was both an important sourceof
royal revenue, and a considerableburden on landholders.
Yet what is known about the nature of the geld? Initially the geld was
levied on a national basis as an extraordinary tax in order to bribe raiding
partiesof Danesto direct their attentionselsewhere.Appropriately, it became
known asdanegeld.This impost, which hasgenerallybeenregardedasa land
3 DB consists of two MS volu.rres, LDB which includes the counties of Essex, Norfolk and Suffolk,
and Exchequer DB, which includes data of a briefer nature for the remaining counties.
a Owing to differences in the way cenain variables are recorded from circuit to circuit, considerable
problems would result fro¡n attempts to aggregatedata to the national level. It is best, therefore, to proceed
on a county by county basis. A similar point, on a different subiect, is made by K. P. Vitney, 'The
Economic Position of Husbandmen at the Time of the DomesdayBook: A Kentish Perspective',Economic
Histoty Reuiat znd ser, p<xvtI (¡984), pp. 23-34.
s See S. Painter, Srudiesin the History of tlæ Englíih Fadal Baroty (Baltimore, 1943); S. K. Mitchell,
Stuli¿s in Taxøtion underJohn and Hnry IIl (ltlew Haven, r9r4); M. M. Postan, The Medinal Economy
and Society(t972); R. L. Poole, TIæExchequãin the Twelfih Cnnry (Ofrord, rgrz); and H. R. Loyn,
Anglo-SøxonEngland and tlæ Nonnøn Corryrcst(1963), pp. 3o5-¡4.
6 Painter, Englíth Feudal Barotty, p. 74. A similar claim is made for this period by J. Campbell in
'Observations on English Government from the Tenth to the Twelfth Century', Transactimtsof the Ruyal
Hisørical Soci¿ty,z5 G97S), p. 5z; while M. K. Lawson in 'The Collection of Danegeld and Heregeld in
the Reigns of Ethel¡ed II and Cnut', Eng. Hist. Rat. gg (1984), pp. 722-3 suggeststhat the burden of the
geld was even more severeprior to the Confessor'sreign than after the conquest.
7 Painter, English Feudol Børory, p. 78.
8 Postan, Mdieoal Ecottottrjt,p. t7z
e This calculation is basedon the simpli$ing assumption that there were no exemptions from this tax,
354
JOHN MCDONALD AND G. D. SNOOKS
tax,lo wasfirst leviedin 99r by King Ethelred,and until ror2 was collected
irregularly to meet the Danish threat. In the latter year it becamean annual
tax employedto finance both a standingarmy (known as royal housecarles)
and a small fleet of warships, until it was discontinued 6y Edward the
confessor i1 1o5r. During these years this tax was known as heregeld(or
army geld). rüØilliamI may haverevivedthe geld asearlyas rc66167,6utfitile
is known about its incidenceuntil the end of the reign, by whicú time it is
generallyregardedas having becomean annual tax once more.ll From this
time onwardsmore information is available..Inro83/84,for example,the geld
was levied at the rare of six shillings per hide, in-ro!6 at four ihiilings-per
hide, during l_rgltv I's reign at rwo shillings per hidé (levied annualy), änd
{}.i"g Henry II's reign at rwo shillingsper hide for the yearsr156 anâ':r.62.
Although the geld, assuch, wasnot leviedafter r 16zbecauseof the opposition
it aroused,both Richard and John attempted to raise similar taxeì-on five
(tt94, r r98, rzoo, tztT andrzzo) baseduponhidesand evenactual
occasions
ploughlands.But, owing to renewedoppositionfrom the king's barons,even
theseirregular levies finally ceasedin izzo.l2
il
Even older than the geld was the method of assessmenr
upon which it was
based.The underlying hidage systemis thought to be linkid ,.ro the more
primitive tribute taken by Anglo-Saxonkings in early days,and alsowith the
syltgm wherebyfood-rentswere extractedfrom dependentestates".l3From
evidencecontainedin the eighth-centuryt¿u(colleciors'document known as
the Tribal Hidagerla it would appearthat there was somecontinuity in the
systemboth beforeand after the geld wasfirst
þ.p."-r.E -hi4"sçassessment
leviedby Ethelred. It is the natureof this assessmenr
sysrem,asiãhad evolved
by ro86, that is the subiect of this article.
uqail ¡e1t$y ago the generalassumptionamongstDomesdayscholars
was that"the hide, being an areal measure,was directly related to a certain
number of real acres.Although there was somedebateôver how manv acres
constitutedt hide (or "homestead")rrsthis areaof land wasgenerallyregarded
-R.
levei.re rüf.
1sbeing sufficientto maintain a peasantfamily at a subsistence
Þytot was the fust to challengethis idea suðcessfully.rn Kqt to Domesday,
first published in 1878, Eyton claimed that the manor in Éthelred's reili
(979-tot6) had been"assessed
accordingto its value-according to its capaci-ty
10
,
ffThe geld was a land-tax, the fust regular and p€n¡ranentland-tax known to the Vest in the Middle
,The Last cennrry of
{Ees", -L-oyn, Angla-saxonEryctand,p. ¡os; and more recently t. A. Green,
Danegeld',-.8'ng',
Hist. Reo.,96 (¡98¡), p. z4r. Both acceptedthe inierpretationóf f . u. Stenton,.Angtr
Søxon Englanil (Oxford, rg43), p. 645.
11Round, Faulal Englan¿Lpp. 87-8, and G-albraith,
The Making of Domesda3¡
Book, p. ç.
'îit" Lrrt Century
-1] Paintgr, Eaglbh Feudal Børony, p. 79; Mitchell , Studi¿sin Taraíion; and dreen,
of Danegeld', pp. z4r-2.
13Loyn, Anglo-SøxonEngland, p.
3o5.
14lbid. pp.
3o6-9.
^ls R. w_. E¡on, Key n Domesda)t:An Anatysisand Dþest of tIæ Dorsasuntey (tg7g), p. 14, suggested
48 acres; Round, Feudal England, pg. 4r-7, demonstratesthat it was ¡zo acro.^1ó-rlfaitland,DomesdayBook,pp.35T-62;Y. stubbs, r/re constiutionalHistoryofEnglmd,r(oxford,
187+),pp. 79 and I85' Also seeR. Lennard,'The Origin of the Fiscal Cancaíe',-Econ. Hist. i?at. xw
(¡944), p. 59 for a discussion of the original meaning of the hide.
TAX ASSESSMENTS
OF DOMESDAYENGLAND
355
to beartaxation", but that while "valuesand capacitieswereliable to constant
fluctuation, we cannot find a single instancewhere hidøtionappearsto have
beenalteredto suit the specialcircumstancesof deterioration,ôimisfortunes,
or neglect".l7on this basishe arguedthat by ro86 hides no longer correspondedcloselyto measuresof areaor value (revenue):the hide wás solelya
unit of tax assessment.
_ \Øith the exceptionof J. H. Round, this new idea was resisredby other
Domesdayscholars.At the 1886conferenceto mark the 8ooth anniveisaryof
the Domedaysurveymosthistoriansconcernedwith thehidageissueattempted
to support the older view.18only Round acceptedEyton'sclaim that the hide
was a measureof tax assessment,
which had come adrift from its economic
origins.le Even so, at this time there was little evidencein Round's work of
the radical interpretation of Domesdayassessments
which was to grow from
this seed.There were just two hints of what wasto come,and thesecan only
be identified in retrospect.First, like \üØ.Stubbs,2oRound believedthat "thê
hundred was the fiscal unit for the collection of Danegeld".2l secondly, in
the same paper he made a brief referenceto the regular occurrenceof an
assessmentof five hides for Domesdaytowns.His only explanationof this
was: "Now five hides was tlte unit of assessment
for the puipose of military
seryice".22Over the following decadethesetwo separateobseivationswere to
coalesceand develop in Round's mind to produce a view of Domesday
assessments
which has prevailedfor the last century.
Round's revolutionaryinterpretation of geld assessment
appearedin 1895
i1 his major work, Feudal England. From a srudy of Infuisitia comitøius
(ICC), which unlike DB recordedthe number of hides on a
Cantabrigicnsis
village aswell ason a manorialbasis,Round saw(or thought he saw)a pattern
in the allocation of hides which suggestedan arbitrary ãdministrativè game
rather than economic reality.23 He attempted to demonstratethis in the
following way. In the first place Round õonstructeda table showing the
number of-hides, ploughlands, and aaleu for the five vills in a "typical
hundred" (staines)recordedin ICC. He showedthat while eachvill wajrãted
at ten hides, there was some variation in ploughlandsand oalets.while he
wasableto demonstratethat the relationshipbetweenhidesand thesevariables
rryasnot exact,he did not considerthat it, nevertheless,could be a strong nonexact (stochastic)relationship. Rhetoricâ[y Round arkedt
rvhat is the meaningof it? simply that ASSESSMENT
BORENo RATIO ro
AREA OR TO VALUE in a Vill, and still lessin a Manor.
17Eyton, Key to Domesilay,p. to.
by Taylor and Poll in P. E. Dove, ed. Doncxlay
Strldlies,
r and u (rgBB).
-Dove,
t'il lq_p"¡rers
H. Round, 'Notes on DomesdayMeasuresof Land', in
Dor*td"! Srudi¿;,r, p. r94.
J.
20Stubbs, ConstitutianalHisøry, r, p. lr8.
'Danegeldand
the Financeof Domesday',in Dove, Domesdø|tstudies,r, p. rr7.
?l
H. Round,
22lt
lbid. p. rzo.
23Round seemsto have regarded most admùrisEative
arrangementsof the Anglo-Noruran regime as
arbitrary impositions divorced from economicreality. He viewed, for example,knighl serviceasan artiûcial
system of military tenure, unrelated to the value of the property of those who weie required to provide it.
Prof. Holt drew our attention to tlis matter, which we are èurrently analysing st"tirticauy. See H.
J.
Round, 'The Introduction of Knight Serviceinto England', Eng. Hítt.-Reo.å (rlql), pp.
4r;-4l,6iS-lS;
'The Inuoduction
pp.
rr-24i
C.
Holt,
7 Q89z)'
of Knight Servicein England', laia-ttoniãn Sndic;,
J.
6, pp. 89-to6; and s. Harvey, 'The Knight and the Knight's Feein England-'
, pastanã present,49 e97o)',
w.3-43.
AND G. D. SNOOKS
J O H NM C D O N A L D
356
Assessment
wasnot obiective,but subjective;it wasnot fixedrelativelyto area
or to value,but to thefive-hideunit. Theaimof theassessors
wasclearlyto arrange
the assessment
in sumsof five hides,ten hides,etc.2a
After a similar examinationof a secondhundred (Radfield),which consisted
of sevenvills each rated at ten hides, he overstatedhis caseby concluding:
"The thing speaksfor itself. Had the hidation in thesetwo Hundreds been
dependenton areaor value, the assessment
would havevaried infinitely".zs
In fact ploughlandsvaried, somewhatless than infinitely, from 13{ ro 20:
onceagainhe overlookedthe possibilityof a non-exacrrelarionship.Although
he proceededto assembledata from other hundreds in ICC and in other
counties,Round'srevolutionaryview waslargelybasedupon a visualexamination of the abovetwelve observations.
Recognizingthe narrow empirical base of his new interpretation Round
sought support elsewhere.In this "chase" for the frve-hideunit, no further
considerationwas given to the relationship between hides and resources.
In the first place he extendedhis net to include other hundreds in ICC.
Unfortunately this additionalevidenceis not only ambiguous,but alsodamaging to his cause.The aboveevidenceincluded two hundreds in which the
hidage of singlevillages was divisible by five: hence the five-unit system.
Round's additional evidencewaslessaccommodating,as it becamenecessary
to group togethervarying numbersof vills to obtain a multiple of five. These
village clustersranged from z to 5 vills (and the number of resulting hides
ranged from zo to 25), thereby implicitly allowing for variation in resource
endowments.Despitethis Round concluded:"The chasenow becomesexciting: it can no longer be doubted that we are well on the track of a vast
system of artificial hidation, of which the very existencehas been hitherto
unsuspected".26Secondly,Round gavebrief and unconvincingattentionto a
number of isolated hundreds from other counties, including Bedfordshire,
Huntingdonshire,Northamptonshire,Hertfordshire, 'Worcestershire,
Somerset, and Devon. Once againhis searchwas for the illusory five-hidepattern,
and not for the relationshipbetweenhidesand other resources.In otherwords,
he attempted to reiect the idea that assessment
was basedupon capacityto
pay by this indirect method of searchingfor the five-hide unit (or, as in the
caseof the Danelaw counties, the six-carucateunit) rather than by directly
testingfor a non-exact(stochastic)relationshipbetweenhidesand both resources and ztalets.In the courseof this researchRound claimed to discovernot
only the five-hide system, but also the possibility that the hundred, rather
than the village or manor, was the basisof the assessment.
At the end of his
ramble through a limited number of selectedhundreds he concluded with
characteristicconfidence:
Indeed,thoughI definitelyadvancethe suggestion
that the assessment
was,in the
first instance,laid uponthe Hundreditself,and the subsequent
assessment
of its
Vills andManorswasarrivedat by divisionandsubdivision,
thetruth or falsehood
of this theoryin no way affectsthe indisputablephenomenon
of the five-hideunit.
On the prominenceof that unit I take my standas absoluteproof that the hide
2aRound, Fatdal England, pp.
489, (his unorthodox emphasis),
2slbid. p.49.
2ólbid. p. 5r.
T A X A S S E S S M E NOTFSD O M E S D AEYN G L A N D
357
all the
assessment
wasfixedindependmþ
of øreaor oølu¿andthat, consequently,
attemptsthat havebeenmadeby ingeniousmen to discoverand establishthe
boreto area,whetherin Vill or Manor,haveproved
relationwhichthat assessment
but, aswasinevitable,vain.27
not only contradictoryamongthemselves
preferred not to contradict
contemporaries
Is it little wonder that Round's
lüø.
gentlemanly
F.
Maitland
had no desireto do so.
him? Certainly the
propagating
played
Round's views,
a
significant
role
in
Maitland, in fact,
purpose
the
manner
in which
of
DB,
but
also
concerning
not only on the fiscal
dimension
holders
of
land.
A
fascinating
tax assessments
were imposedupon
of this processof the diffusion of intellectual ideas is that Maitland was
responsiblefor propagatingeven thoseviews of his contemporarywhich he
did not accept.The reasonfor this was that although much of the evidence
Maitland assembled,togetherwith the inferenceshe d¡ew fromit, contradicted
the artificiality interpretation, Maitland failed to confront Round's dubious
claims.28In failing to do so he rendered Domesday scholarshipa major
disservice.As Round's viewshad the apparentsupport of the most accomplished of the late Victorian Domesdayhistorians, they were adoptedby the
leadingscholarsof following generations,suchas Sir Frank Stentonand Prof.
H. C. Darby. Theseviewshave,therefore,remainedthe conventionalwisdom
on this issuedown to the presenttime.
Maitland's view on the questionof whetherthe geld in DomesdayEngland
wasbasedupon the capacityof manorsto pay is, at first sight, ambiguous.He
presentedtwo contradictoryinterpretations,togetherwith the evidenceupon
which they were based,as if there were no differencebetweenthem. In the
first place Maitland examinedthe relationshipbetweenhides and teamlands
in the countiesof Buckinghamshire,Lincolnshire, Yorkshire, Leicestershire,
Northamptonshire, \üØiltshire,Dorset, Devon, Somerset,and Cornwall. By
noting whether the number of hides was equal to ("equal rating"), greater
than ("over-rating") or lessthan ("under-rating") the number of teamlands
he discoveredthat, while there was somevariation, the equationof the two
was "a condition that we may call normal".2eTo explain the "aberration of
A [hides] from B [teamlands]" Maitland considered the effects of both
privilege, or "beneficial hidation", and "the operationof natural and obvious
causes[by which] an old rate-bookwill becomeantiquated".3oBut after some
consideration he deemedtheir combined effect insufficient to account for the
perceivedvariation, as the fust of thesecould accountfor only "exceptional
cases", and the secondfor only slow changesrather than those "that take
place suddenly at the boundariesof counties". He turned, mistakenly, to
Round to supply the completeanswer:
A masterhandhaslatelyturnedour thoughtsto the right quarter.Therecanwe
27Ibid. p. 62.
28It is well known that Maitland and Round were working on thei¡ Domesdaybooks at the sametime,
and that Maitland modestly and generouslywithheld publication of DomesilryBooh and Beyond for two
years ro allow Round's views to circulate without competition. This is acknowledgedby both authors.
Maitland said that it was done so that he and others might lea¡n from Round. After readingFadal Englan¿
Maitland said: "In its light I have suppressed,corrected, addedmuch" (p. v). Unfornrnately he suppressed
and "corrected" too much, as his stetistical work and historical insight were far superior to thoseofRound,
2eMaitland, DomesdayBook, p. 447. For a recent discussionof beneûcialhidation after ¡o66 seeGreen,
'The Last Century of Danegeld', pp.
z4S-52.
30lbid. pp. 448 and 45o.
358
JOHN MCDONALD AND G. D. SNOOKS
think beno doubtthat, asMr Roundhasargued,thegeldwasimposedaccording
to a methodwhichwe havecalledthemethodof subpartitioned
provincialquotas.
A sumcastupona hundredhasbeendividedamongthat hundred'svills; a sum
castupona vill hasbeendividedamongthe landsthat the vill conrains.3l
This was a metlod, he claimed, which was bound to produce "inequitable
results" and to lead to a systemof patternedhidesof the type discovèredby
Round. In order to "emphasizethe artiûciality of the systèm",32Maitlanã
directed attention to the frve-hide patrern apparenrin a few hundreds in
Cambridgeshireand \üØorcestershire.
secondly, Maitland examinedthe relationshipbetweenhides and values
(or revenue).He began,as before, by expressingthe view that:
jobbery,we cânnot helpfancyingrharwe
behindall the capriceandpresumable
seea certainequityprinciple. . . . Now we haveformedno veryhigh estimate
of
the iusticeor the staresmanship
of the Englishwitan [council],and whatwe are
ggingto sayiswrungfromusbythefigureswhichhavedissipated
somepreconceived
ideas;but they hardly allow us to doubt that the numbèrof hideJcasrupon a
countyhad beenaffectednor only by the amount,but alsoby the valueìf its
teamlands.33
This is, of course,a contradictionof Round's hypothesis,to which Maitland
had earlierdeferred.Nevertheless,Maitland procèededto test the relationship
between hides and values in the following rwo ways. First he compareã
Iisually the hides and values aggregatedat the county level for twenty
Domesdaycounties. Secondly,in order to avoid the próblem of including
non-agriculturalprofits in these county totals, Maitland selectedon a basié
which is not disclosed,and thereforecannotbe assumedto be random, eighty
manors from each of thirty counties.These non-randomsampletotals werè
aggregatedfor each county, expressedas values(in fs) per hide, and then
comparedvisually one with the other. According to Maitland, both these
statisticalexercisesreveal a principle underlying the observedcounty varialtoÏ " principle of "one pound, one hide". But, characteristically,no sooner
had he admitted the existenceof a relationshipbetweenhides and valuesthan
he qualiûed his discovery:
TØe,
havespokenof a tendencyon the part of the hide to be worrh a pound.Now
wehaveno wishto representthis eqùtableelementasall powerfulor verypowerful;
the caseof Kent is sufÊcientto showthat it hasbeenoverruledby favôuritismoi
privilege. . . . Still that the kings and witan haveconsideredthe-valueaswell as
the numberof teamlandsseemsfairly plain. Probablythey haveconsidered
it in a
rough,typicalfashion.3a
Even this evaluationwasfurther modified, Maitland sawin somecountytotals
þarticularly Yorkshire, Lancashire,and Northamptonshire)"the unreiormed
relics of an- age when the distribution of ûscal units among the various
prwinces of England was the sport of wild guesswork".3sAnd once again
under the influence of Round he refers to the hide (and the teamland) as
"ar':ficial and . . . remote from real agrarianlife".36 Finally, he admits to
DomesdayBooh, p.45o. He speculatesabout the way this was done oà pp. 472-3.
1i 44*d,
32
lbid. p. 452.
33lbid. p.
463.
3albid. p. 47o.
3sIbid. pp.
469-7o.
36lbid. p. 47r.
TAX ASSESSMENTS
OF DOMESDAYENGLAND
359
being "baffled by the make-believeof ancientfinance".37Therefore,although
claiming that ". . . we are endeavouringto loosenone of Domesday'sworst
knots",rs Maitland was,in fact, responsiblefor tighteningit, both by endlessly
qualifying his own interesting and independentstatisticalwork, and by not
taking issuewith Round.
It is clear from this account that Maitland acceptedwithout reservation
Round's claim that the systemof assessment
was imposedon the hundred
from above.Yet he vacillatedover the questionof artificiality versuscapacity
to pay. rüØhile
he admittedthat there wassomeevidencefor Round'sdiscovery
of the five-hide unit, he was reluctant to deny that there was a systematic
relationship betweenhides and both teamlandsand values.lüØhilerheserwo
views are essentiallycontradictory,Maitland provided not evena hint of this
conflict. He just presentedboth views, togerherwith the evidenceupon which
they were based,using oneto qualify the other. To validatehis own statistical
work it was necessaryfor Maitland to challengeand rebut the evidenceand
argument of Round. For somereasonhe chosenot to do this. Vhy? \ühile
this remainsone of the intriguing enigmasof Domesdayhistoriography,there
are a number of possibilities. It may have been the result of an abnormal
deferenceto the ideasof others, particularly when expressedas absolutelyas
Round's.3eIt may have been causedby Maitland's lack of confidenceabout
his pioneeringstatisticalwork on Domesdaydata, asexpressedin the preface
to Domcsd.ay
Book andBeyond.ao
But most probablyit derivedfrom an inability
to establish a method to determine which set of conflicting evidencewas
correct. Whatever the reason,those that followed Maitland appearto have
regardedhis failure to take issuewith Round as unqualified support for the
artificiality hypothesis.
A few notableillustrations-such asthe viewsof P. Vinogradoff,F. Stenton,
and H. C. Darby-are sufficient to demonstratethe continued influence of
Round and Maitland. Vinogradoffs method was similar to Maitland's. He
acceptedthat the systemwas artificial becauseit was basedupon a regular
pattern of hides at the village and hundred levels.alSimilarly, he detecteda
relationship between assessment("geld hides") and arable land ("ûeld
hides"), but unlike Maitland he drew attenrion to the conflict of evidence.
Unfortunately, he attempted to resolve this conflict by a number of fanciful
argumentsabout a peculiarrelationshipbetween"geld hides" and "nrmbered
hides".a2Once again, Round triumphed over conflicting evidence.
Stenton,on the other hand, acceptedwithout questionRound's interpretation of danegeld.In Anglo-saxonEngl.andhe wrote: "The fact that assessment
to gelds proceededfrom above downwards-from the king's court, through
the county and hundred, to the village-was establishedlong ago by scholars
37Ibid. p.
473.
38lbid. p. 47r.
3eV. H. Galbraith, Domesihy Booþ: Its Pbce in Administtdtioe llirøry (Odord, 1974), p.
9, made a
similar suggestion on other issues. Discussions with Prof, I. C. Holt and Dr G. H. Manin have given
further confidence in this speculation.
aoHe said (p. vt): "for my hastily compiled Domesday Statistics I have apologizedin the proper place.
Here I r¡"i[ only add that I had but one long vacation to give to a piece of work that would have been
better performed had it been spread over many years".
al P. Vinogradoff, EnglíshSocizEin the Elcvnth Cmury (Ofrord, r9o8), pp. r97, r99.
42[bid. pp. zor-7.
360
ANDc. D. sNooKs
JoHNMCDoNALD
suchasRound and Maitland".a3And again:"From a datewhich lies far back
in English history a large round number of hideshad beenattributed by the
king's ministersto eachcounty in this region,without any but the mostgeneral
referenceto its true agriculturalcondition".aaYet most revealingof all: "The
intricate combinationsof assessment-units
into neatblocksof five hidesor six
carucatessometimeslook like the result of a gamewith figures, played by
clerkswith no interestin realities".asIn other words,Stentonacceptedwithout
qualiûcationRound's view of the nature of the Domesdayassessments.a6
More recently Round's artiûcial assessment
thesis has been advancedby
H. C. Darby. This interpretation occurredin the five counry volumeswhich
first appearedbetween r95z and 1967, together with the summary volume
Domesday
Englandfust publishedín t977. As this article focusesupon Essex,
Darby's work on that county will be examinedfirst. Of the Essexhidation
Darby said:"We cannotsayto \ryhatextentthe incidenceof the hidagereflected
the realities of the time. The figures certainly bear no relation to those of
ploughteams."aTThis conclusion, with minor variations ("ploughteams"
sometimesbecome"agricultural resources"),was repeatedby Darby and his
co-authorsfor county after county, therebyassumingthe statusof a formula.as
In the caseof Essex,Darby attemptedto supportthis statementby examining
the ratio betweenhides and ploughteamsat rhe hundred level for only three
hundreds.
In his final Domesdayvolume Darby summedup his view of assessments
for all England. In a sectionentitled "Artificial Assessments",he wrote that
"the assessment
wasessentiallyartificial in character. . . . \üØe
haveno means
of knowing how equitablethe original assessment
may havebeen,but by ro66
it was manifestly not so".4eIt is quite clear, therefore,that Darby accepted
completelyRound's artificiality hypothesis,and reiectedthat of capacity to
pay. But what is surprising is that he did so even though he was unable to
find sufficient evidencein his county volumesto acceptRound's supporting
evidenceconcerningthe five-hide pattern. In his final volume Darby said:
"Not all countiesdisplay the five hide systemas clearly as Cambridgeshire,
but whether they do or not, it is clear that villageswith identical assessments
had widely differing resourcesand populations".SoIn other words, while
unable to acceptRound's evidenceof the five-hide(and six-carucate)system,
Darby appearsunwilling to reject Round's artificiality thesiswhich was based
on this "evidence". Instead, Darby attempts to retain the long-standing
conventional wisdom by claiming that there was no relationship between
assessment
and resources.Round was right, Darby çleims, but for the wrong
a3Stenton, Anglo^SaxonEngland, p, 646.
aalbid. p. 646.
aslbid. p. 648.
a6It is interesting that Galbraith n DomesilayBooÞ complains that Stenton, being "a pupil of
J. H.
Round" þ. 16), accepteduntil his death Round's view of how the survey was conducted-in a hundred
rather than a feudal basis. In discussion, Prof. Holt has suggestedthat Round played an important role in
Stenton'searly career.
47H. C. Darby, Th¿ Dotusday Geogaphy of Eastnn Englon¿ (Cambridge, rgTo), p. zzo.
a8For example see H. C. Darby et al., Dorn¿silattGeograpþ of Northmt Englafld (Cambridge, ¡962),
pp. 23, to6,244,29r:,339, passim.
4eH. C. Darby, Domesihy Ettglnnd (Carnbridge, tgTg), p. g.
solbid. p. ro.
3
TAX ASSESSMENTS
OF DOMESDAYENGLAND
36r
reasons.Since Darby's work, medievalistsinterestedin the geld have been
concernednot with the relationshipbetweentax assessments
and the capacity
to pay, but rather with matters such as its severity before ro66, a possible
reâssessment
in ro86, and the nature of its exaction(including exemptions)
together with reasonsfor its discontinuancein the twelfth century.sl
ilI
To cut through Maitland's knot of ambiguity it is necessaryto clarify the
nature of the conflicting methodsused to analyzeDomesdayassessments.
It
was demonstratedin SectionII that two methodswere employedto support
the artificial assessment
thesisand to reiect the capacityto pay principle: first
by rearranging the data on hides from a manorial to a village (or village
cluster) basisin order to examineit for a distinctive pattern; and secondlyby
visually inspectingthe data on a manorialbasisfor any relationshipbetween
hides and both resourcesand values.The frrst method is an indirect test, and
the seconda direct test of whether Domesdayassessments
were basedupon
the commonsenseprinciple of capacityto pay. It will be argued here that
while there are fundamentalconceptualproblems involved in the use of the
indirect method, thesedo not apply, in principle at least,to the direct method.
Yet the way in which the direct method hasbeenemployedin the Domesday
literature has causedincorrect inferencesto be drawn.
The basicproblem with the method devisedby Round is that it bearsonly
indirectly upon the questionto be solved.This methodis basedupon the idea
that, if the number.of hides per vill, or cluster of vills, exhibits a regular
pattern, then it is the result of administrativeconveniencerather than the
principle of capacityto pay. There are two aspectsto this problem. First, the
processby which a regular pattern is observedis highly subiective.Not only
doesit involve aggregatingvills into clustersif the predeterminedpattern is
not evident at the individual vill level (which could involve making covert
allowancefor the different resourceendowmentsof vills), but alsoit enables
a selectiveuse of data by focusing attention upon those observationswhere
the five-hide unit emerges.Vith this type of analysisthere is, of course,no
generally acceptedset of rules that can be used to indicate whether the
evidencedoes indeed support the hypothesis.Secondly,as Round's method
is indirect it does not exhaust all alternative explanationsof an observed
pattern. For example,a regularpattern of hides could be generatedby a ûxed
relationship betweenthe size of villagesand the resourceendowmentsof the
manors they serviced,rather than by administrativegamqsmanship.
The second method employed has a number of distinct advantages:it
s1 See,res¡rectively,Lawson, 'The Collectionof Dangeldand Heregeld'; S. P. Hawey, 'Domesday
l.
Book and Anglo-Nonnan Governance', Trønsactio¡ts
of tlæ Royl Historicøl SocicE,25 Gg79, pp. 17S-g3;
'The
and Green,
Last Century of Danegeld'. Of this more recent work, Harvey's is the most relevant to
our concem in this article. She advancesthe hypothesisthat the'plougblands'variable may not be what
it appearsto be (i.e. ân amount of arable land measuredin terurs of the nu¡nber of teams neededto work
it), but rather a new and disguised assessmentundenaken in ¡o86. Slhile there are a number of reasons
for doubting this hypothesis, we have not testedit here bec¿use'ploughlands'arenot recorded in the Essex
folios. Instead, the record for this county includes the numbers of 'ploughteams' and the scope for
employment of additional ploughteams.
362
JOHN MCDONALD AND G. D. SNOOKS
and capacityto pay,
d,irectlymeasuresthe relationshipbetweenassessment
therebyexcluding alternativeexplanations;and the resultscan be interpreted
more obiectively by employing standardstatisticaltests of significance.The
problem in the Domesday literature is that when the direct method has
been employed-by Maitland, Vinogradoff, and Darby-it has been used
inappropiiately. These scholarsbasedtheir analysesupon an examinationof
either only a few observations,or larger non-random samples.Often these
observationswere chosenbecausethey representedthe exUemitiesof the
observedrange.Statisticaltheory indicatesthat theseproceduresmay provide
a quite misleadingimpressionof the averagerelationshipbetweenvariables.
Beforethe adventof the electroniccomputerand the developmentof statistical
methodologyit is understandablethat the analysiswas undertakenin this
way. Today the technology is available to enable the relationship to be
analyzedusing all the data for a particular region. Accordingly, in sectionV
the direct method has beenapplied by employing regressiontechniques.By
effrcientuse of all the data for Essex,entirely different conclusionsabout the
and resotuceendowments(and
averagerelationshipbetweentax assessments
hencecapacityto pay) have been obtained.
ry
Only a brief considerationof the data in DB is necessary,becauseit has
attracted the attention of numerous scholarsover the last century and has
beensummarizedby the authorselsewhere.s2Itis generallythought that DB,
to which has beenatuibuted both a fiscaland feudal purpose,was completed
betweenChristmas ro85 and the death of Villiam I in Septemberro87, a
period of only twenty months.s3This rapid compilationwas facilitated, so it
is thought, by the use of pre-existinghidage lists, which were used as a
framework for the collection, verification, and presentationof evidence.For
this purposethe countiesof Englandweredivided into sevenor more circuits,
which were visited by separateteamsof disinterestedcommissioners.These
commissionerscirculated a set of questionsto the king's tenants-in-chief(or
their agents),obtainedeither written or verbal answersrelating to both ro66
and ro86, and finally subiectedthis information to the scrutiny of the hundred
juries (composedof equal numbers of Englishmen and Frenchmen)in each
county court. The risk of false information being supplied was minimized,
both becausethe evidencewas scrutinized by iuries under oatl, who had local
knowledge of the estatesin question, and becauseany offending party riskedseverepenalty. Even the commissionerswere spied upon by further agentsof
the king. No censusor survey of modern times can claim the samebuilt-in
safeguards.This is not to say that mistakesdid not creep into DB, either
deliberatelyor by accident, but the method of collectionwas such that they
would have been minimized, at leastfor ro86, the year of the survey.sa
The Domesdaysurvey of Essexfor ro86 provides sufficient deçailto test
s2 McDonald and Snooks, 'The Determinants of Manorial Income', section III.
s3 Some, including the author(s) of. tj¡e Anglo^SaxmChronicle, claim it took only nine months.
sa A good recent discussion of the natu¡e and methods of the survey can be found iD S. P. J. Harvey'
'Recent DomesdayStudies',Eng. Hist. Reit.' (r98o), pp. ¡zr-33.
95
T A X A S S E S S M E N T SO F D O M E S D A Y E N G L A N D
363
the relationshipat the manorial level betweentâx assessments
(measuredin
hides) and the capacity to pay (measuredboth by a comprehensivelist of
resourcesand annual values). Data are availableon a manorial basisfor the
following variables: the number of hides; the annual value (or income,
consistingof rents and profits on the homefarm); the number of ploughteams
(consistingof a plough and eight oxen);the number of additionalploughteams
that could have been employed had they been available; the numbers of
freemen,sokemen,villeins, bordars, and serfs(slaves);the amount of woodland (in tefins of the number of swineit could theoreticallysupport); pasture
(numberof sheep),and meadow(acres);numbersof real beasts(cows),sheep,
swine,goats,and horses;and finally the numbersof mills, saltpans,fisheries,
vineyardsand beehives.To reducethe number of explanatoryvariablesto a
manageablesize, some--namely the componentsof labour and livestockhave been combined,ssand others have been eliminated-namely saltpans,
fisheries,and vineyards-because they appearedincomplete or were minor
resources.The resulting setsof explanatoryvariablesinclude: values(Ð on
the one hand, and the resourcevariablesof ploughlands(actualploughteams
plus additional ploughteamsthat could have been employed)(Pf, livestock
(Li)s6,labour(À0, serfs(S), woodla\d (W, meadow(Me), pasture(P), and
mills (ll) on the other. There is, however, an ambiguity concerning the
resourceof meadow.Thê literature suggeststhat meadowwas a very valuable
resourcebecauseit wasessentialto the maintenanceof the critically important
ploughteam(by producing hay), becauseit was highly fertile, and becauseit
was scarce.sT
Also it was thought by someto have been communally owned
and disuibuted regularly by lot.s8 Finally, meadow was essentialfor the
maintenanceof horses,a key elementof England'swar machineafter ro66.
As it was vital that the area of meadow be maintained, it is possible that
meadowwas exempt from taxation, in which caseit should not be included
as an explanatoryvariable.
v
To test the hypothesisthat there was a systematicrelationshipbetweentax
assessments
and the economiccapacityof manorsto pay this impost, the 695
lay holdings in the County of Essexhave been analyzed.Attention has been
focusedupon lay holdings in this paper becausethey constituted78 per cenr
of a total which also included the manors of Crown and Church, and because
the remaining estateswere complicated by taxation exemptions
As argued in section III, one way of measuringthe capacityof manorsto
pay the geld is in terms of their resources.There is, however,a major problem
in establishingsuch a relationship, becauseof the diffrculty of determining
ss The system of weights for combining livestock was: cows,
3o; swine, 8; sheep, 5; and goats, 4. See
Mutland, DonesdayBooÞ,p.44; A. Ballard, TIu DonesilajtInpæsr(19o6),p. zl;1. H. Round,'Domesday
Survey', in V. C. H. Esser, r, p. 367; and J. A. Raftis, Ifte Estàtes
(Toronto, tg57), p.
of Ramsey.áåáey
62.
s6 Ercluding horseswhich, asshown in McDonald ¿nd Snooks,'The Determinants of Manorial Income',
rvere a non-economic resou¡ce.
s7 Maitland, Donexlq Booh, p.
44j
58Ibid. p. 348.
364
JOHN MCDONALD AND G. D. SNOOKS
its functional form. If a relationshipexists, a þriori we might expectthe tax
to be an increasingfunction of resourcesso that greatermanorial
assessment
Theoretical
with a higher tax assessment.
resourcesare, on average,associated
of the
precise
form
functional
the
indicate
however,
reasoningdoes not,
in
or
log-linear
linear
example,
is,
for
the
relationship
relationship;whether
are
used
methods
regression
arises
when
usually
the variables.This difficulty
to measurerelationships.One way of overcomingthe problem is to assume
from the outsetthat the relationshiphasa particularform, for examplelinear,
and to estimatethe parametersof the relationshipgiven that assumption.This
strategywill work well if the linear form is a goodapproximationto the actual
functional form, but not otherwise.To guard againstthe latter possibility,
the alternativeapproachof assumingthat the actual funcúonal form can be
adequatelyapproximatedby a member of a broad classof functional forms
hasbeenadopted.The data havethen beenusedto selectthe functional form
from this broad class.
The classof functional forms entertainedis defined by the Box-Cox Extended (BCE) model, which includes the conventionallinear and log-linear
functional forms as specialcasestogetherwith thosethat are linear in a power
transformationof the dependentand explanatoryvariables.Transformingthe
variablesby taking powersor logarithmsoften "straightens-out"or linearizes
a relationshipwhich is non-linearin the original variables.To seethis, let us
supposefor simplicity that the tax assessment,I,, only dependson a single
and
reiource variable,R,. If the averagerelationshipbetweentax assessment
graphed
when
observations
scatter
of
variable
is
log-linear
the
this resource
will trace out a curve rather than a line (see Fig. t). Clearly the average
relationship between tax and tÏe resourcevariable will not be linear. If,
however, the observationson tax and the resourcevariable are transformed
by taking logarithms,and the logarithmof tax is graphedagainstthe logarithm
of the resourcevariable, the scatterof observationswill trace out a line. In
other words the logarithm of tax is linearly related to the logarithm of the
resourcevariable, that is
ln T, : Br+B,lnR,*e,,
(r)
where Êoand F¡ areparametersand e,is a randomdisturbance.This log-linear
functional form is the member of the BCE classof functions which has a
transformationparameter(/.) equalto zero. For someother data set, another
transformation may "straighten-out" the curve. For instance, it may be
necessaryto transform the variablesby taking squareroots. The functional
form that is linear in the squareroots of the variablescorrespondsto the BCE
fünctional form with )" = L. If the data are linear in the original variablesthe
BCE transformationparameter(/.) is equalto one,andhenceno transformation
is required. Box and Cox explain how L together with the parameters,B;r
of the functional form can be jointly estimatedby a maximum likelihood
method.se
se SeeG. E. P. Box and D. R. Cox, 'An Analysisof Transformations',Journal of tlæ Ragal Sntistical
SocicE,SeriesB, z6 (t964), pp. 2r\-52,
T A X A S S E S S M E N T SO F D O M E S D A Y E N G L A N D
365
Figure t. An lllustration of How Trørtsforming
theData May InduceLinearity
and Homoskedøsticiw
T,
lnT,
R,
InR,
The dataon manorialtax assessments
and resourcesarecross-sectional
data.
A common problem with regressionequationsbasedon cross-sectionaldata
is that the variance of the regressiondisturbanceis not constant over the
sample but is related to the size of the unit of measurement:in this case,
the size of the manor. \üflhenthis occurs, the disturbancesare said to be
heteroskedastic.Ifthe regressiondisturbancesareheteroskedastic,
then pararneterestimatesobtainedfrom estimationmethodsthat ignorethe heteroskedasticity are inefficient and the usual parameterstandard errors and tests of
significanceare invalid. Transforming the original variableswill often make
heteroskedasticdisturbancesmore homoskedastic.For example, referring
again to Fig. r, if in the original variablesthere is greater variation in tax
assessmenfs
q[ high valuesof the resourcevariable,when tax and the resource
variablearettànsformedby taking logarithms,the variationin the transformed
tax variableat high resourcelevelswill be moderated.It is often the casethat
there existsa transformationof the original variablesthat both "straightensout" the curve and induces homoskedasticity;nevertheless,it is perhaps
optimistic to assumethis is the casefor any given data set. Consequently,the
estimation method used here allows for the disturbancesin the regression
equation, which is linear in the transformedvariables,to be heteroskedastic,
with the varianceof the disturbancedependingupon the size of the manor.
The ploughlands variable is a measure of the size of the arable land of the
manor and, becauseon most manors arable farming was the maior activity,
the ploughlandsvariable is employed as the indicator of manorial size.6o
60Formally it is assumedt\at the disturbances, et, in ttre regressionequation involving the transformed
variables, (for example equation (¡)), are a set of independent drawings from a normal distribution with
meân zero and varianced,: o2(PÐ\, where ø2 is a positive constant,(PÐ, is the valueofthe ploughlands
variable for manor t and y is the heteroskedasticityparameter, which is nou negative. Nótice that if y : q,
then d = ø2, so the disturbances are homoskedastic and the.model reduces to the conventional BCE
model. Ify : z, the variance ofthe disturbance is proportional to the squareofthe size-indicator variable,
a plausible and commonly used speciûcationfor distu¡bance heteroskedasticity.This model is very similar
to that postulated by K. Lahi¡i and D. Egy, 'Joint Es :mation and Testing for Functional Forrr and
Heteroskedasticity',J ournal of Econonetrics,r 5 (r98 ¡), pp. 2gg-3o7.
366
JOHN MCDONALD AND G. D. SNOOKS
The regressionresults are presentedin Table r. Regressionr provides
estimatesof the relationship between tax assessmentand the resourcesof
ploughlands,livestock, labour, serfs, woodland, meadow,pastureand mills
for Essexlay holdingsin ro86. The unknown parameters¿, T, þo, F'r..Fpof
the model were estimatedby the maximum likelihood method with the values
for,l and y being determinedby searchingover a grid boundedby o ( ,t < r
and o ( / < 3. This searchindicatedthat the likelihood wasmaximizedwhen
(to three decimal places)T : o, so we set }z : o and found the maximum
likelihood estimator searchingover valuesof ,t in the interval (-4r 5). The
maximum likelihood estimatorof /. is o.z3o. Regressionr in Table r lists the
BCE estimatesfor this value of )".61A number of points should be noted.
First, the results of a seriesof diagnostic tests applied to this equation are
consistentwith the hypothesisthat the equationis well-specified.The Reset
testr62which is a generaltest for misspecificationof the regressionequation,
does not reveal a misspecification,and the results of two heteroskedasticity
tests,63Hetero (a) and Hetero (b), suggestthat the disturbancesare homoskedastic.Secondly,coefficientestimates(alongline r) arepositivefor all resource
variablesexceptmeadow.The r-ratios6aindicatethat the coefficientestimates
61An inplicit assumption underlying the BCE models is that there exists a value of the transformation
parameter,1, such that the uansformeddependentvariable,7$), is normal. However,for 1>o, 1S) must
be greater than - ¡/1 and for i<o, flI) must be lessthan UX;there is a truncation problem, so flÀ)can only
be approximately normally distributed. If I : o¡ there is no truncation problem and if the absolute value
of I is small (which is the case for our estimated model) the truncaúon is small. The results of N. R.
Draper and D. R. füx in 'On Disuibutions and their Transformation to Normality',Joumal of thz Royøl
SøtisticalSocæry,SeriesB, 3r (1969), pp. 472-6suggestthat inferencesare robust with respectto the
approximation. The diagnostic tests and a Chi-squared goodnessofût test for normality suggestthat, when
l : o.23o, the distu¡bancesare indeed well-behaved and approximately normal. The Chi-squaredtest was
basedon 6o sub groups of the residuals from Regression¡ and the null hypothesis of normality was not
reiectedat the 5o per cent level. SeeC. J. Huang and B. W. Bolch, 'On the Testing of Regression
Disturbancesfor Normality', Journal of tlæAmeritan Sutisticøl Associøtian,6g(tglÐ, pp. 330-5for details
and discussionof the test. Note that four manors without tax assessmentsand rz manors without values
were omitted from the sample and the problem of ze¡o variable values was de¿lt with by adopting the
(Reading, rg77), p. lz of
suggestionof F. Mosteller and J. V, Tukey, Datø Atulysis and Regression
adding one sixth to the observationsof the resourcevariablesbefore transforming the variablesand omitting
a mQst unusual manor with a zero annual value. The r-ratios given in Table r are conditional on i.
Uncünditional t-ratios can be calculated from the likelihood fuiction or by a non linear least squares
regression. SeeJohn J. Spitzer, 'A Monte Carlo Investigation of the Box-Cox Transforuration in Small
Samples',J.Am. Støt. Assoc.73(rSZ8),pp. 488-95.They are extremelysimila¡ to the conditional,-ratios.
62See J. B. Ramsey, 'Tests for Speciûcation Errors in Classical Linear Least Squares Regtession
Analysis',J. Ragal Stat. Soc. SeriesB, 3r (1969),pp. 35c7r, and fames B. Ramseyand Peter Scbmidt'
'SomeFurther Resultson the Use of OLS and BLUS Residualsin SpeciûcationError Tests',J. Am. St¿t,
Assoe.7r (1926), pp. 389-9o. The square and the cube of the estiuratedpredicted valueswere included as
additional regressorsin ou¡ tests.
03These ¡re tests described by A, R. Paganand A. D. Hall in 'Diagnostic Tests as Residual Analysis',
Econam¿tricRetieuss,z (¡983), pp. t59-2r8. For the Hetero (a) test, the null hypothesisis that the r-th
disturbance variance, ø1, equalsø2 for all l, and the alternative hypothesisthat o1 = o2+E(y,lxòa., where
E(ylx) is the expectedvalue of the dependent variable, given valuesof the explanatory variables, The test
is designed to have high power against heteroskedasticity that is linearly related to the expectedvalue of
the dependent variable. The test involves regressingthe Regressionr squared residualson a constant and
the estimated predicted values and testing that the estimated coefficient on the estimated predicted values
is zero using the ¿-test.For the Hetero (b) test the null hypothesis ß o7 : o2 for all ¿and the alternative
hypothesisthat o2' : o2*E(tt,lx;)sn+(E(l/xò)2a,. The test is designedto havehigh power againstheteroskedasticity that is linearly related to the expected value of the dependent variable or the square of that
expectation. The test involves regressingthe Regressionr squared residuals on a constant, the estimated
predicted values and the square of thesevalues and testing that the estimated coefficients on the last two
variables are both zero using the F test.
6aThe ¡-ratios in line z (labelled t) are the usual t-ratios, that is, the ratio of the estimated coefficient to
its estinated standard error. The ¿-ratiosin line 3 (labelled (ø)) are calculated by White's method, see
H. White, 'A Heteroskedastic-ConsistentCovarianceMatrix Estimator and a Direct Test for Heteroskedasticity', Ecowmetrica,48 (¡98o), pp. 817-38.
T A X A S S E S S M E N T SO F D O M E S D A Y E N G L A N D
367
and pasturearesignifrcantlydifferent
for ploughlands,livestock,labour, s_erfs,
from zero at the one per óent significanbelevel. Thirdly, Sq f test statistic6s
estimatesare
igiu." under the colu-mnheadedF) confirmsthat the coefficient
ùïhit iïS"in.."t. For'thly, the'coefûcient of determination adjusted for
A."gr.., oî freedom (Rr) is;.626 whicþ is a good F Jot disaggregat:-d__.11t:
data of this'nature, particularly when the influencesof privrlegeanO
sec"tional
adjustment are taken into account.66
delayed
--iffrihly,-;hiÈ
the literature led us to anticipate$at $e meadowvariable
insignificant, it cameas a surplise-that it was_significantand
*;y-ü;.'beãn
It is onl/possiblé to speculateabout the reasonsfor this. It was
;õ"d"..
ruãgÃte¿ in sectioniV th.t, becauseof its importanceto the Anglo-Norman
.ãäÃ""i,V, meadow -"y trot have been taxed. However, our surprisilg
results suggestnot that mäadowwasfree of tax, but that it was the basisof a
ir* ãã¿"."tion. A possiblereasonfor the king granting a I?x deduction on
meadow could have been to encouragelandholders(who did not have per-""é"t rights to the meadow)to undeitakethe investment(draining,.clearing,
aná flood"control)required to expandthe acreageof meadow,67which.¡ras.a
critical resourcefor maintainingboth ploughteamsand war-horses'While this
explanation is consistent with- ont rèsults it must remain in the realm of
untit direct evidenceof meadow managementin the eleventh
;Ëcudtü
Finally, as the estimated regressioncoefûcientsin
;;;G-fo""¿.
äil;y
nãgt.ttio" r relate to the tranéformedvariables, their interpretation is not
is facilitatedby calculatingresourcesiñple. Interpretation of the relationship
-the
line labelledE in Table r)' These
to ärr"rr*ent elasticities(given along
of tax assessment
sample,means
the
at
fartial elasticities,which aré evaluated
importance of
the
ihdicate
clearly
äo¿ tft. respectiíe *ãtoot.. variables,
exp.lainþq .ty
in
labour,
exient
leiser
a
to
the plòughlands variable, and
is a highly
there
that
indicates
r
Regression
To summarize,
resources
the
and
"r..ir*.it.
assessments
tax
manorial
between
significant relationship
For all
dominant.
being
ploughlands
of
connibution
;i;h.-*-;;;;;*t,h-de
tft." meadowthe relatioìship is, as expected,positive' Some
;;ñtð;;tfrär
and meadow
t ag"tiu. associationbetween tax assessments
the
for
reasons
'f;.niffå::åt"å',iTïf*s
TheBCE
is orinterest.
rerationship
estimated
:
form with )ç : o.23ois intermediatebetweenthe linear (4 -I) and log-lincar
form. For
log-linear
ii-: ol?r"ctional-forms, lying somewhatcloser to the
Regression
in
presented
èompaíativ. po.poro ttt.io'g-lñ.* form estimatesare
z in Table r. Altlr-oughthe-estiriratespassthe Reset{aSnostic test, the.FForm test68 suggests"that the tog-linear functional form is inappropriate
6sThis is the test statisúc for the test of the null hypothesisthat the resou¡cevariablesdo not signifrrantly
against the
affect úranorial *
G.ã- o. co.mci*is on the ¡esource variables are all zero)
"rr"rrãlnìr
that at leasì one resourcevariable does signiûcantly affect manorial tax assessment
hypothesis
Ji"*"ti*
Pñnciùl¿s of
t"..,'åo. of the coefÊcients on rhe resource variables ls non-zero¡' See H. Theil,
G..
"t
pp'
138-9.
(Amsterdam,
r97r),
Ecotto¡netrics
66lbid. p. ¡8r.
- il" ,"op. for this can be seen in the caseof Dorset, where in the_Sooyears after ro86 the area of
meadow incräsed from 7,ooo to 95'ooo acres. Maitland, Dotttcxlty Bgo\,-v' þ3'- .
68This is rhe test describe¿ UVL. G. Godfrey and M. R. lflickens in'Testing Linear and loc-LFeT
hypothesis
negx.ssion, for FuncúonJ F ora¡1Ì
, R*i*t o¡ øcorcnit Stutlics,48 (r98r), pp. 487'96. The null
the test is designedto havehigh power againstthe alternative
for the test is that the modei it iog-1i"."r
""á of the BCE family of functional forms'
that thrc model is *o-th.r nember
iyp.,¡.rit
Table t. Tax Assessment-Resmffces
ald Tax Assessm¿nî-Income
Relationshþsfor EssexLay Holdings, rc86
Const pt
Regression
t
Box-Cox
Extended
Transforrration
n=678
p
t
4ø)
Li
N
S
tJl
M
e
P
M
8 . ¡ r r.r7 o.o5r o.39t o.zt9 o.o35-o.zo6 o.rrr o.o4g
32.3
3r.5
F
9.6
4.2
S.r
3.o
2,9
r'5
-4.6
4.5
V
I
R
2
F
o.z3o o.626 142.6
o.3
s
F-Form
2.22
r.o8
F(8,669)
7.5 4'2 4.5
r.5 -4.4 4.4 0.3
o'4j3 0.056 o.r84 o.o69 o.oz7 -o.o96 0.063 o.oro
Reset
r.96
Hetero
\a)
u)
o\
æ
Hetero
(b)
o.o7
o.34
F(2,667)
t(66Ð
F(2,666¡
3.or
+ r.96
3.or
Fl
Þ
X
Regressíonz
Log-Linear
Form
n= 678
Regressionj
Box-Cox
Extended
Transformaúon
n=678
p
t
lw)
E
4.65 o.399o.o34 o'r49 o.o78 o.02I -o'oó7 o.o5r o.o2g
5 9 . r ç . 8 4.3 5 . 3 3 . 4 r ' 7 -3.7 4.r
-3.9
55.3 7 . 3 4-3 4 . 5 3.6
4.3 o ' 8
o.399 o.034 o'r49 o.o78 o.02I -0.067 o.ojr o,o2g
0.584
o.72r
r.96
p
8.¡s r'94
, 7 r . 9 28'9
4w) 67.E 25,5
E
r2o.o
F(8,669)
o'23o 0.552 8j4.3
o'717
-5.r
Þ
q)
2.96
-5.29
16.09
4668)
F(2,667)
t(661)
F(2,666)
Xt'97
3.or
+r.96
U)
(t)
3.or
3
9.gr
-r.o4
6.t4
t(67Ð
+ r.96
F(2,673)
U)
z-l
2.43
F(r,676)
F(2,614)
3.86
3.or
U)
3.ot
3
Regressíon4
Log-Linear
Fonn
n= 678
p
2.41
29.o
t(w)
E
28-7
.7o9 o
34.9
35'5
o.709
0.642 tzr6.7 o.669 _o.r9
F(r,676)
t(67Ð
3.86
!t.g7
o.r3
_o.9r
o.jo
F(2,679
(fiÐ
+r.96
F(2,6t3)
3.or
Nores"The first row in the table
F) eives the Box'cox Extended coefôcient estimatesof respectively the intercept or consranr, ploughrands
(Pl)' livestock (rr), labo'r 9-!elled
(Àf, shvèi (s), woooan¿
pasture (p) anJi.uG ø. The second row (labelled gives trre
Ø!;;d;;lälirià2,
conventional ¡-ratios and the thi¡d row (l"beued
I
4n¡ r-raios calcular"¿Li*¿fii¡.o
vditá'. ¡rr.tloã.'äe
fourth row gives the partial resourcet¡rx assessmentelasticities evaluated at the sample mean of tax assessment(Ð
and the samplemean of the resource (R) *a defined as ô1, . Rr.
I/ refers to manorial value or income, 2 the transforrratio¡, pÍ¡rarneter
of the Box-Cox Extended uansforsration, .R2the coefûcient of do.'j*åt
adiusted for degreesof freedom, and F the t st statisti. fãi-.rd"c
t"îih-.äãä"i"o,.
on all resource variabres ale zero.The value of the F
statistic is 14z'6, F(E,66q) in line z indicates that on the nuu
lvpotlesis th.;,fürti.
is
distributed
,. . r-¿i.ãiuution with g and 669 degreesof
freedom and ¡'s6 in line r is the 5 Per cent
u¿u. for thàt F-dirrriilrü:l
i, ,r. ¿i"ur¡*..
.ltinrrì.d standard error. F-Form, Reset,
(o) ayd il-etao (Ð íelaæ;iÉ-tu";d"ä-io.-lnr*
"¡ti.J
*a it.r.Ã-rt.ããr:ãüry ,.r,, menrioned in rhe rext. In the tust row the rest
Fo?o
sratistic
is given and underneath the distribution of the test tt tirti. una"r
t¡ãnouîypoo.sis
and the 5 per cenr critical value for that disuibution.
3.or
U'
ñ
Þ
z
F
Þ
z
H
lAX
A S S E S S M E N T SO F D O M E S D A Y E N G L A N D
369
and the Hetero (a) and Hetero (b) tests indicate that the disturbancesare
heteroskedastic.Nevertheless,the Regressionz results convey a broadly
similar story about the relationship to those of Regressionr. Except for
meadow, the coefficient estimatesare positive and the rüØhite'sf-ratios, (see
the row labelled (ar)) which allow for heteroskedasticity,suggesta highly
signifrcantrelationship.Comparisonof correspondingresource-taxassessment
partial elasticitiesreveals a broadly similar pattern in the two equarions,
although the elasticitiesin the log-linear equarionare a little smaller.
As someDomesdayscholarshaveregardedthe geld asa tax solelyon arable
land, this questionhasalsobeeninvestigatedhere. If manorialtax assessments
dependedonly on the acreageof manorialarableland, the coefûcientestimates
on the other explanatory variablesin Regressionr should be insignificant
from zero.This is not the case.Regression3 presentsthe BCE estimateswhen
tax assessment
is related to ploughlandsonly. The transformationparameter.
,t is o.z3o, the sameas that for Regressionr, but the Resettest indicatesa
misspecificationin Regression3. This is the result that would be expectedif
important explanatory variables had been omitted from the regression.A
formal test that the tax assessments
are not determined by the livestock,
labour, serfs, woodland, meado\ry,pasture, and mills variablesis rejectedat
all conventionalsignificancelevels.6eThis evidencesuggeststhat in ro86 the
geld was not simply a land tax but was a tax basedupon all the resourcesof
manors.
As the capacityof a manor to pay tax can be measuredin terms of the value
of its output as well as in terms of its resources,the relationship between
manorial tax assessments
and manorial annual valuesis presentedin Regression 4. The main characteristicsof the estimatesare as follows. First, the
maximum likelihood procedureindicated that the heteroskedasticparameter
y is zero and the transformationparameter,t is equal to o.r3o. As this value
,of ,t i$, not significantly different from zero at all conventional significance
levels,the log-linear form estimatesfor this relationshiphave beenpresented
in Table r (see Regression4). Secondly, the regressionestimatespass the
diagnostic tests and exhibit a highly significant positive relationship between
tax assessment
and manorial annualvalues.Thirdly, the value-taxassessment
elasticityestimateof o.7og indicatesthat, on average,a one per cent increase
in manorial value is associatedwith a o.7og per cent increasein the tax
assessment.Fourthly, the value of the coefficiènt of determination, Rr, is
slightly higher than in Regressionr, but these coefficientsare not easily
comparablebecausethe variableson the two equationsare subjectto different
transformations.Fifthly, the value-tax assessment
elasticity in Regression4
(o.Zog)is similar to the sum of the partial resource-taxassessment
elasticities
in Regressionr (o.746). This indicatesthat similar.resultsare obtainedin the
tax assessment-capacity
to pay regressions,irrespectiveof whether capacityto
pay is measuredin input or in output terms.
6s The test is the Likelihood Ratio test of the null hypothesisthat the coefûcientson the abovementioned
variables are all zero against the alternative that at least one coefficient is non-zero and is based on the
result that z(Lo-L¿.), where Zo is the maximized log-likelihood under the null hypothesis and l¿ the
maximized log-likelihood under the alternative hypothesis, is asymptotically distributed as a Chi-squared
disuibution with sevendegreesof freedom. The test statistic has a value of 65.25t and the 5 per cent and
o.5 per cent critical values for this Chi-squared disuibution are 14.o67 and zo.z78, respectively.
370
JOHN
MCDONALD
AND
G. D.
SNOOKS
Figure z. The Tax Assessment-Mønorial
Annual Value Relationshipfor
EssexLay-Manors,to86
TaxAssessment
(in acres)
100
Value
(in shillings)
Finally, the nature of the geld assessment
can be illustrated by reference
to Fig. z in which the averagetax assessment-manorial
value relationship has
been graphed.Both averageand marginal tax assessment
rates can be calculated from the graph. The averagetax assessmentrate for a manor with an
annual value of, say, 50 shillings is given by the slopeof a line from the origin
to the point on the tax assessmentcurve corresponding to a manorial value of
that amount. For manors with an annual value of 5o shillings the averagetax
assessment
rate is 3.6 fiscal acresper shilling; while for manors with annual
valuesof r, 5, ror 2c.,4c,6c18o and roo shillings,the averagetax assessment
rates are rr.rr 7.o, S.T, 4.7, 3.8r 3.4r 3.r and 2.9 fiscal acresper shilling
respectively. As will be noticed, the decline in the averagerax rate is quite
marked at low annual values but moderates as the manorial annual value
increases.The marginal tax rate when the manorialvalue is, say, 5o shillings,
is given by the tangent to the tax assessmentcurve at the point corresponding
T A X A S S E S S M E N T SO F D O M E S D A Y E N G L A N D
37r
to that value. The marginal tax rate also declines,but at a declining rate, as
the annual value of the manor increases.These estimatessuggest,in other
words, that the geld was a regressivetax levied on the economiccapacityof
manors.
VI
Contrary to receivedopinion, which extendsback to the Victorian scholar
J. H. Round, the statisticalresults in this article demonstratea strong and
positive relationshipbetweenthe geld assessments
and the capacityto pay of
Domesdaymanorsin Essex.This is true whethercapacityto pay is measured
by the revenues("values") or by the resourceendowmentsof manors. It
in DB, for Essexat.least,cannot be regardedas
follows that the assessments
artificial or simply based.on an arbitrary principle such as adminisuative
convenience.Also, the existenceof such a strong relationshipcastsconsiderable doubt on the plausibility of Round's hypothesisthat the tax assessments
were cast down from above rather than built up from the manorial level.
Nevertheless,the empirical evidenceis not inconsistentwith the claim that
the hundred was the administrativeunit of the geld.
Our resultsalsoclarify a number of other aspectsof the geld in ro86. First,
by ro86 the geld can no longer be thought of as a tax solely on arable land.
This in turn suggeststhat the king's agentswere attempting to take account
of the growing economic sophisticationof rural settlementswhen framing
their non-feudal taxes. Secondly,the geld was a regressivetax in the sense
that the marginal tax rate declined as the revenueof the manor increased.
This reflectedWilliam I's needto placatehis powerful barons,without whose
support he could not have cometo power in ro66. Thirdly, it is possiblethat
tax deductions were granted on meadow land; certainly our results are
consistentwith this ûnding. This may have been done in order to encourage
investmentin this critically important resource.
Finally it is desirable to place our conclusionsin a wider context by
speculatillgbriefly about the nature and developmentof the geld assessment
system.7oThe fact that the geld assessmentsystem was based upon the
principle of capacity to pay is consistent with the hypothesis that the king
attempted to ma:rimize his net tax revenue, as claimed in contemporary
literature.Tl This conclusionis basedupon the argumentthat the adoption of
the equitableprinciple of capacityto pay would haveminimized the opposition
to this tax, which in turn would have minimized the costs of enforcement
(or transactionscosts) and hence maximized net tax revenue. The effective
application of capacity to pay, however, would have depended upon the
employmentof a good practical measureof this principle. This idea may be
usedto clarify the current ambiguity aboutthe developmentof geldassessment
from 99r to 116z. It can be argued that in earlier Anglo-Saxontimes, when
population was small, land was abundant, and when production was of a
subsistencenature resulting in similar mixes of inputs and outputs, a simple
meâsure of arable land (such as the number of ploughlands required by a
70This is developed more fully in McDonald and Snooks, 'Modelling Medieval Tax Systems'.
71G. N. Garmonsway,ed, Anglo-SaxonChronicl¿(t953), p. z16.
372
IOHN MCDONALD AND c.
D. SNOOKS
family for subsistencei.e. the hide of rzo acres)would have sufficed as a
measureof capacityto pay. As population grew, settlementsexpanded,less
favourable land was brought into economic use, production becamemore
specializedwith varying input and output mixes, and trade becamemore
common, a simple measureof arableland would not have reflectedcapacity
to pay. Accordingly it would have been necessaryto adopt a more general
measureof capacity to pay, such as manorial revenueor manorial resource
endowments.The evidencesuggeststhat this had been achievedby ro86.
By ro86, therefore, England was governedby a sophisticatedand highly
able administration. In this respectit was a worthy successorto the AngloSaxonregimesthat it replaced.T2lüØilliam
I's administrationwasableto assess
and record the taxable capacityof English manorsaccurately,and it was at
pains to employ this information (embodyingthe principle of capacityto pay)
as the basis of its taxation policies. DomesdayBook, therefore, is not only
evidenceof the seriousand able intent of the Anglo-Norman régime, but is
also a vital and remarkably rich sourcefor the reconstructionof the English
economyin the late eleventhcentury.
The Flinders Unizsersity
of SouthAustraliø
72Campbell, 'Observationson English Government',p.
54.