Tue Nov 17 10:31:12 EST 1998
================= EH.RES POSTING =================
Comments on the Agricultural Revolution debate -
by Liam Brunt (Nuffield College, Oxford).
General Points.
It has been suggested by Michael Turner that a quantitative fog has grown
up around English agricultural history - in particular, whether there was
an Agricultural Revolution and (if so) at what date it occured. There are
three main reasons for this quantitative fog.
First, we are trying to gauge what happened across the whole of England
based on relatively small samples of data which are time and place
specific. It is hardly surprising that the data sometimes appears to be
contradictory - just because yields are moving up in one county and down in
another does not mean that our techniques are wrong. It just means that
there is a lot of variation across England and things can be happening
differently in different places. This does not mean that we should abandon
our quantitative approach and focus on 'processes': it just means that we
should collect more data.
Second, it is not always clear what we mean by Agricultural Revolution. It
is therefore very difficult to gather empirical evidence which can either
establish or reject its existence and timing. Like Michael, I do not find
the term 'Agricultural Revolution' very useful and I would be happy to let
it die. Its main function seems to be to give researchers something to
argue about and indicate that you work on English agriculture of the period
1500-1850. Since the term was coined by historians, I am only surprised
that it was not guillotined in the Quantitative Revolution, along with
other noble historical concepts. Instead of getting hung up on what
constitutes a revolution, I think that it would be more useful to establish
(quantitatively) how various aspects of productivity and technology
changed over time, and in what way England's experience was the same or
different to other European countries. Moreover, whether or not you
consider the eighteenth century to be an 'Agricultural Revolution', there
were a number of quantitatively important changes occuring in England which
had not occured previously and which were not occuring anywhere else at
that time. There were unusual (possibly unique) changes in land and labour
productivity which were associated with the invention and/or diffusion of
new techniques. Such diffusion *may have been* be accelerated by changes in
organisation, such as agglomeration of landholdings.
The third and most important cause of the fog which has surrounded
quantification of English agricultural history is the mis-interpretation
and over-interpretation of data by economic historians. For example, it is
often asserted that agricultural productivity went up without stipulating
what aspect of productivity we are talking about. The productivity of land?
or labour? or Total Factor Productivity? It should be noted that the
productivity of land can go up whilst the productivity of labour is going
down and TFP is constant. At the same time, total output and the yield of
any particular crop (such as wheat) could be doing anything. It is
dangerous (and often completely misleading) to try to draw any conclusions
about the path of one type of productivity from the path of another. But
historians commonly try to do just that, and end up talking at cross
purposes. One recent example is to be found in Greg Clark's opening
contribution to this debate, and I will begin my specific comments by
questioning his interpretation.
Greg's Section 1 - 'There was no agricultural revolution'.
Greg suggests that from 1500 to 1850 'agricultural productivity drifted
upwards at an average rate of less than 10 per cent in each 50 years'. I am
not sure what Greg means here by 'agricultural productivity'. If Greg is
referring to TFP, then I am willing to accept than no 50 year period
between 1700 and 1850 saw an increase of 10 per cent or more. I would
suggest that TFP growth in the arable sector over the period 1775-1845 was
approaching 10 per cent. (I am unwilling to take a position on the period
before 1700 because I have not yet seen any evidence which is strong enough
to support any statements about TFP growth whatsoever).
By contrast, I would suggest that labour productivity almost certainly rose
at a rate in excess of 10 per cent 1705 to 1775. Most workers were employed
in the arable sector, where I estimate that the increase in labour
productivity from 1705 to 1775 was 37 per cent. The rate of growth of
labour productivity slowed after 1775 and the total increase in the period
1775 to 1845 was only 4 per cent. I nonetheless still hesitate to designate
the period 1705-1775 as an Agricultural Revolution because I do not have
firm evidence on labour productivity for the period before 1705.
Interestingly, this increase in labour productivity was driven mostly by an
increase in the productivity of land rather than the amount of land
operated by each labourer. In the period 1705-1775 there was an increase in
land productivity of 25 per cent, and an increase in land per worker of 9
per cent; in the period 1775-1845 there was an increase in the productivity
of land of 35 per cent and a fall in the amount of land per worker of 23
per cent. And contrary to Greg's assertion, the increase in the
productivity of arable land was generated largely by the spread of fodder
crops and not by an improvement in grain yields. In the period 1705 to 1775
fodder crops accounted for 75 per cent of the increase in land productivity
and grain yields for only 25 per cent of the increase; between 1775 and
1845 fodder crops accounted for 56 per cent of the increase and grain
yields for only 44 per cent of the increase. (If Greg disagrees with these
figures then I would be interested to see his alternative estimates in
order that we can pinpoint the source of our differences).
It is also important to note in a comparative context that the period 1705
to 1775 saw a very substantial widening of the Anglo-French agricultural
productivity gap (both land and labour productivity). Moreover, this gap
did not widen any further in the period 1775 to 1845, when you might have
expected England to draw further ahead (given the increase in the rate of
overall TFP growth after 1800 documented by Crafts and the disruption of
the French Revolution and so on). So whether or not the years 1705-1775 saw
the onset of an Agricultural Revolution, they certainly saw an important
divergence between England and her main Continental rival.
The whole period 1705 to 1845 also saw an important contribution of new
techniques to agricultural productivity. Contrary to Greg's
characterisation of my research, to which he referred in his reply to
Michael Turner, I find a number of new techniques raised output and
productivity from 1700 onwards. It is correct that the traditional
techniques of marling and liming significantly raised wheat yields circa
1770. But even using 'significance' in a narrow statistical sense, I find
that crop rotation also significantly influenced the wheat yield (as
demonstrated by an F-test for joint significance). Crop rotation was a
primary factor causing fluctuations in wheat yield in the period 1700-1860
and the increasing use of turnips was a prominent component of that
fluctuation. The lack of fodder crops in France was also the primary cause
of their lower wheat yields (compared to England) and hence an important
cause of their overall poorer performance in arable productivity.
I also find that the diffusion of other innovations in the period 1705-1845
had a significant effect on wheat yields.There is clear evidence that seed
drilling and horse hoeing raised wheat yields by around 40 per cent where
they were employed. There is also evidence that primitive hollow drainage
around 1770 raised wheat yields by 14 per cent where it was employed; so it
is reasonable to suggest that the falling cost and rising adoption of
drainage in the 1840s (after clay pipes had been invented) had a
substantial effect on overall output.
Overall, it is difficult to think of any period before 1700 when so many
innovations were being made and diffused and there is such clear evidence
of their impact on output and productivity. If Greg believes that there was
*no* accleration of productivity growth, then one challenge for him is to
list the new techniques which were developed in earlier periods and
demonstrate (quantitatively) that their overall impact was as great as
those of the eighteenth century.
Greg's Section 3 - 'The failure of English agriculture to maintain output
per head'.
It is rather unclear to me what Greg is arguing in his Section 3, and
perhaps he could clarify this for me. He states that 'The urbanization and
industrialization of Britain in 1760 to 1860 was not spurred by the release
of labor by capitalist agriculture.' Following Crafts (1985) the Industrial
Revolution has commonly come to be defined as the transfer of resources
(especially labour) from the primary sector (mainly agriculture) to
industry (and services). If we all accept this definition, then it is
definitionally true that industrialisation was spurred by the release of
labour by agriculture (capitalist or otherwise). I assume that Greg agrees
with this statement.
In fact, Greg seems to be trying to establish whether the redistribution of
labour resources (proportionately, rather than in absolute terms) was
caused by changes occurring in the industrial sector or the agricultural
sector. His paragraph then focuses on the fact that UK agriculture did not
maintain its level of output per head of population, and he argues that
this somehow 'required' labour to move into the industrial sector.
But the transfer of labour resources is not determined by agricultural
output per head of population. Suppose that labour markets were functioning
properly (ie wages changed in agriculture and industry competitively in
response to changes in the demand and supply of labour; and workers changed
sectors in order to maximise their utility, in which wage was a primary
argument). Then the transfer of labour from agriculture to industry was
merely a function of changes in the marginal revenue products of workers in
each sector. An increase in the marginal revenue product in industry (and
hence upward pressure on wages) encourages workers to switch sectors until
diminishing returns in industry force the marginal revenue product back
down to equality with the agricultural sector. The new equilibrium has a
larger proportion of workers in industry (ie a bigger Industrial
Revolution). So Greg seems to be suggesting that Britain's falling output
of agricultural products per head of population pushed down the marginal
revenue product of agricultural workers (compared to the industrial sector)
and so more workers transferred into industry.
As it stands, I do not follow the economic logic of this point. There is
not necessarily any systematic relationship (causal or not) between
agricultural output per head of *population* and the marginal revenue
product of workers in agriculture. I suspect that Greg is actually
suggesting that if the average revenue product of UK *agricultural workers*
had been higher then the marginal revenue product would also have been
higher at the observed level of output, causing more workers to be retained
in agriculture. Hence he views the industrial revolution as a failure to
raise average productivity in agriculture.
This would be true if the average and marginal product curves shifted out
parallel to their initial position - then clearly the marginal revenue
product in agriculture would have been higher and there would have been
less labour transferred. But if we are postulating that the curves could
have moved (say, due to a change in technology) then we also have to
recognise that they could have changed shape. Then it would be perfectly
possible to have an increase in average product without any change in the
marginal product. This would depend entirely on the nature of technological
change. In principal, agricultural output per head of population (or per
agricultural worker) could have doubled without any change in the
equilibrium number of workers in agriculture and industry. Take the example
of guano, which was a powerful fertiliser imported from Chile which came
into use in the mid-nineteenth century. Guano was a land-augmenting
technology shock which left the labour and capital requirements per unit of
land unaltered (nor could farmers change their inputs of land because the
stock of land in England was pretty much fixed). So average productivity of
land and labour rose substantially but the marginal product of labour was
unaltered, and hence there was no effect on labour use between sectors.
Greg's counterfactual that higher output per agricultural worker would have
undermined the industrial revolution are therefore unsubstantiated. There
are no general theoretical results showing that changes in marginal
productivity on intra-marginal units will cause changes in marginal
productivity at the margin; nor do I know of any direct evidence from the
eighteenth century which supports such a statement.
Greg's reply to Mark - 'Mediaeval markets were already integrated'.
Greg rejects the idea of institutional change leading to higher
productivity in English agriculture, suggesting that there was no
improvement in the functioning of institutions such as markets. He presents
a regression in which the price of grain in a particular village responds
one-for-one with national prices and does not respond at all to local
yields, indicating integration between local and regional markets. But this
result is neither necessary nor sufficient to show that there was no
improvement in market integration. The basic problem is that Greg does not
consider variations in the level of prices and the volume traded. This can
be demonstrated clearly with the following example.
Situation 1200AD. Two villages have different levels of productivity - say,
the land is exogenously more fertile in one village. In consequence, there
is greater supply per head of population and downward pressure on grain
prices in high-yield village A compared to low-yield village B. The high
transport cost of 100 pence means that in equilibrium there is a price
difference of 100 pence between the two villages - and on average it is
only profitable for traders to move 1 unit of grain between the two
villages. Nonetheless, the price movements between the two villages are
highly correlated and local yield has little influence on the price because
the volume traded can rise (say, to 3 units) and keep the price difference
at 100 pence. (In Greg's equation, coefficient b on local yield Yjt could
be zero and coefficient c on the national price of grain Pt* could be equal
to unity).
Situation 1400AD. The two villages have different levels of productivity,
as in 1200AD. But a fall in transport costs to 5 pence means that in
equilibrium there is a price difference of only 5 pence between the two
villages - and on average it is now profitable to move 1000 units of grain
between the two villages. The price movements between the two villages are
highly correlated and local yield has little influence on the price because
the volume traded can rise (say to 3000 units) and keep the price
difference at 5 pence. (In Greg's equation, coefficient b on local yield
Yjt could be zero and coefficient c on the national price of grain Pt*
could be equal to unity).
I think that we can all agree that markets in Situation 1400AD are
substantially more integrated than markets in Situation 1200AD. But Greg's
model would detect no difference between these two scenarios - and he
would conclude that 'nothing happened' between 1200 and 1400. One cause of
this mis-interpretation is the constant in Greg's regression which is
absorbing the difference in price *levels* between the two villages. He
should replace it with a time trend. If the time trend were negative then
it would suggest that differences in the price level were declining over
time (ie more market integration). If the time trend were zero then it
would support his case that there was no improvement in integration -
although it would still *not* be sufficient to establish his case beyond
reasonable doubt. This is because there are other reasons why Greg's
equation might be failing.
To give but one example, grain production might contribute an important
fraction of people's income (not implausible, for mediaeval peasants). So a
fall in the local yield might create an adverse income shock. This income
effect would lead people to cut back on their consumption of grain and thus
dampen or eliminate any positive effect on local price. Therefore the
coefficient on local yield could be zero, but this need not imply anything
about the level of market integration.
Bob Allen - 'The landlords (non) revolution'.
Lest anyone should think that I have a vendetta against Greg, I shall take
this opportunity to comment also on Bob Allen's discussion. (I should point
out that one reason it is relatively easy to argue against Greg is that he
is generally more specific in his comments and therefore sets himself up as
a better target. In general, I think it is more useful to make more
specific comments because it helps everyone else to think more clearly.
Thanks Greg!).
Bob argues that a Yeoman Revolution occured between 1600 and 1750 and a
Landlord Revolution occured between 1750 and 1800. During the Yeoman
Revolution yields increased, output soared and labour productivity rose
dramatically. But the Landlord Revolution was characterised by being
completely unrevolutionary - viz, output was stagnant and rents did not go
up (although labour was released in the Midlands, at least). There are
several weakness to this characterisation.
First, what were the precise causes of this transformation during the
Yeoman Revolution? If land and labour productivity both went up, was it due
to an increase in capital inputs or technological change? (Apparently, it
was not due to organisational change, because that only occured with the
Landlord Revolution). If it was due to new technology, then what specific
technologies were responsible and can Bob document the extent to which they
were diffused and their impact in places where they were employed?
Second, there seems to be a jump in the logic of Bob's argument. He notes
that yields were not positively associated with farm size in the Yeoman
Revolution; he then suggests that the rise of larger-scale farms in the
eighteenth century would not have had any effect on the productivity of
grain production. But this implicitly assumes that technology and relative
prices were the same in both periods - whereas if we relax these
assumptions, then gains might have occured from land concentration. For
example, suppose that there was a fall in the cost of capital (such as that
documented in Bob's JEH article). And suppose that some capital projects
benefited from economies of scale (such as marling or drainage, which he
documents in his book). Then we might expect to see a positive effect of
scale on output (as well as land and labour productivity) in the eighteenth
century *even though there was no such relationship in the seventeenth
century*. In fact, I find that there *was* a small positive relationship
between wheat yields and farm size in 1770. I hesitate to claim that this
was causal until I can pinpoint the exact nature of this apparent economy.
One possibility consistent with the qualititative literature is that good
farmers were successful and accumulated enough capital to take on a larger
farm. Larger farms might therefore have higher yields because they were
systematically cultivated by better farmers. Of course, if managerial
expertise is a scarce commodity, then the formation of larger farms would
cause a genuine increase in efficiency because it would allow better
farmers to control more production.
Third, it is not very surprising that we do not observe much impact from
the landlord's revolution if we consider only the period 1750 to 1800. Most
of the of the Parliamentary enclosure had not occured by 1750 and it seems
likely that there would be a lag in investment. If the main benefit of
enclosure was that it permitted more investment in large-scale capital
projects, then this would be an important effect. The qualitative evidence
suggests that there was an increase in marling and drainage projects in the
late eighteenth century, and there was certainly a massive increase in
drainage projects in the 1840s. These would be some of the dividends
derived from the Landlord Revolution (because if holdings had still been
scattered then the cost-reducing technology which came along would have had
no impact on drainage rates, either in the eighteenth or nineteenth
centuries).
Closing remarks.
One of the problems with debating the Agricultural Revolution (or the
evolution of agricultural output and productivity more generally) is that
the available quantitative evidence is still very thin. This remains true
in spite of the efforts of researchers over the past 15 years (who have
added a lot to our stock of knowledge). For example, we have seventeenth
century yield data from probate inventories - but only for certain
counties, and they may not be very representative (such as Norfolk!).
Similarly, there is very little direct evidence about the extent to which
different technologies were employed, and even less evidence about their
effect on output. Other glaring problems include the lack of evidence on
the quantity of arable land in production before 1700 and the shakiness of
the estimates of the labour force engaged in agriculture - so what
confidence can we have in estimates of labour productivity? I could make a
much longer list of similar shortcomings.
This is accentuated by the complicated nature of agricultural production in
an organic farming system (where are the inputs and outputs are generally
interdependent). At the moment we have very little quantitative evidence
about how these elements fit together - how can you evaluate the effect of
changes in crop rotation without an empirical model? Then how can you
evaluate the take-up of new crops? The economic relationships are equally
complicated and we need data on many aspects of production before we can
talk with any certainty about changes in different types of output and
productivity. It is useless to draw conclusions about changes in labour
productivity solely from data on land productivity, and so on.
This does not imply that we should give up. But we should beware of making
strong statements to the effect that there was never an Agricultural
Revolution or that it definitely happened before 1750. There is simply not
yet enough evidence to reach a confident conclusion.
============ FOOTER TO EH.RES POSTING ============
For information, send the message "info EH.RES" to lists at eh.net.