The Determinants of Local Population Growth: A Study of Oxfordshire in the Nineteenth Century

Mark Casson

Key words: POPULATION GROWTH RAILWAY ENCLOSURE OXFORDSHIRE

Acknowledgements: I am grateful to Simon Townley and Janet Casson for comments on an earlier draft. The stimulus for this project came from a Conference organised by Kate Tiller at Rewley House, Oxford, in 2010. I am also grateful to the staff of the Bodleian Map Room for their assistance.

First draft: 28 February 2011

Address for correspondence:

Mark Casson

School of Economics

University of Reading

PO Box 218

ReadingRG6 6AA, UK

Tel: 0118 378 8227

E-mail:

1. Introduction

This paper examines why different towns and villages in England have grown at different rates. It focuses on one particular county - Oxfordshire – in the nineteenth century, but it pilots a research methodology which can be rolled out to other counties. Growth is measured by the proportional increase of population over a decade at the parish level. Differences in population growth are explained in terms of exogenous factors, such as proximity to major towns and access to major channels of transport and communication (river, canal, road and rail). Type of soil is also taken into account, and so too is the ownership and control of land.

The conceptual framework is set out in section 2 and the sources are described in section 3. Statistical results are presented in section 4. The wider implications of these results are examined in section 5, with special reference to enclosures, railways and the growth of Oxford city. Conclusions and implications for future research are discussed in section 6.

2. The conceptual framework

General principles of modelling population change

The distribution of population can be analysed in terms of the demand and supply for places of residence. This approach is useful because census data record residence rather than place of work. This approach also suggests an important distinction between two levels of analysis. The higher level is that of market town and its hinterland, while the lower – more disaggregated - level is that of the individual parish.

Employers and the self-employed often benefit from ready access to a local market, and this suggests that employment opportunities will tend to cluster in the hinterlands of major market towns. Since the demand for residence is often derived from the demand for work, this suggests that population too will tend to cluster around market towns.

There are considerable degrees of freedom, however, in the way that clustering works. Geographical distance from a market centre is not the only consideration for employers. Access to suitable soil, or to roads and rivers, may be important too. Employers can therefore trade off various characteristics across a range of locational options near a given centre.

Employees too have choices, because they do not have to reside in the parish where they work – they can walk, or commute, to work if necessary. Employers will tend to locate near to where employees choose to live, and employees will prefer to live in parishes that provide cheap accommodation and good local facilities. In the case of unskilled workers residential decisions will also be influenced by systems of poor relief and the way that they are applied in various parishes. The precise locations that develop near a market centre will therefore reflect residential opportunities as well as characteristics that appeal directly to employers.

Demand for residence does not stem entirely from demand for work, however. Leisure pursuits are important too, such as landscape gardening, country sports, housing art collections, and establishing model farms. Wealthy residents with an interest in the landscape may prefer to live where the density of population is very low, and may use their control of land to restrict local population growth – e.g. an aristocrat who limits the availability of housing in their estate village.

Competition in the land market will tend to put scenic land into the lands of wealthy residents, and will force poorer people into less attractive locations. People who own less attractive land will be more willing to sell it for housing development. Some establishments, such as union workhouses and asylums, may well be located on very poor land simply because it is so cheap.

If the population is reasonably mobile, workers will gravitate to cheap residential locations in the hinterland of major market centres, and in the long run this will produce an equilibrium distribution of population across parishes. The growth of a parish occurs when there is a change in residential pattern in response to exogenous factors. This leads people to either enter or leave the parish.

A change in birth and death rates will not necessarily alter the distribution of population across parishes, because it will just scale up numbers by equal proportions across all parishes. A sudden increase in the birth rate, or reduction in the death rate, will, however, alter the age distribution, and may stimulate migration – e.g. prime age workers with many dependents leave their families in rural areas and move to towns in search of work.

Some specific factors precipitating change

A potential cause of differential population growth is the relative growth of different market towns. Many market centres in England emerged between the late twelfth and early fourteenth centuries, and were subsequently subjected to periodic ‘shake outs’ during which markets became concentrated on fewer larger centres. Early shake outs were precipitated by population decline – e.g. the Black Death (Casson and Lee, 2011). Later shake outs tended to be driven by advances in transport, such as canals and turnpikes, which encourage producers to send their product to more distant larger markets than to nearby small ones. As a result, smaller markets lost critical mass and disappeared.

In the nineteenth century the building of railways was a major driver of market shake out. In theory, parishes on the railway system benefited at the expense of those off it, and so the arrival of the railway should have been a stimulus to population growth. The stimulus should have been greatest in the early years, when access to a railway was a privilege enjoyed by a few, and lowest later on when lagging parishes finally joined the system to which others already belonged. Railway access was particularly important to market towns, which explains why towns were so prominent in promoting rival schemes at the time of the Railway Mania (Casson, 2009). In the counties covered in this study there were significant differences between towns in the dates at which railways arrived, and so if these effects are important it should be easy to capture them.

The impact of transport improvements on market shake out could be mitigated by general population growth, as this could allow smaller markets to maintain a critical size. Nineteenth century population growth was concentrated in the industrial districts of the North and Midlands, however, and Oxfordshire population remained relatively static and predominantly rural.

The growth of towns has implications for agricultural demand. The town is both a market centre where local goods are sold for re-export to other towns and a place where imports (and locally produced goods) are consumed. Town growth provides opportunities for supplying urban dairies with milk, butchers with meat, brewers with hops and grain, market traders with fresh vegetables, and so on. At the same time, variations in soil type present opportunities for agricultural specialisation. Urban growth may therefore stimulate population growth in parishes that have a comparative advantage in cattle grazing, market gardening, and so on.

Some important qualifications

The determinants of population growth are not the same as the determinants of productivity growth. A switch from one type of agriculture to another may produce more valuable output, but it does not necessarily require more workers to produce it. Substituting grazing for arable may actually reduce the workforce required, as many studies of enclosure have emphasised (Mingay, 1968; Turner, 1980).

In the aftermath of the Napoleonic Wars there were substantial numbers of unemployed, many of whom became itinerant gypsies and tinkers, who sought refuge on open land or private land where absentee owners maintained lax security. In such cases population growth may reflect the marginality of a parish rather than its prosperity.

Outline of an econometric model

The interaction of these effects can be captured by a relatively simple model in which, at any given time, there is an equilibrium long-run level of population in each parish. This level reflects the size of the parish, as measured by its acreage. For a given size, population is likely to be higher if it is a market town or a borough . Population will also depend on a range of other characteristics – in particular location. Key aspects of parish location include proximity to major channels of communication – rivers, road and railways – and distances from established major centres such as London and the county town (Oxford in this case). Land ownership, enclosure and soil type are also potentially important factors.

In a long-run steady state, the impacts of these factors on population growth will remain fixed. If the total population of the county remains steady then the population of each parish will remain steady too. If county population grows, the parish populations will also grow, but they will all grow at the same rate, and relative population growth will be zero; there will be no migration between towns.

In practice, of course, the factors do not remain constant. The building of railway, for example, connects some parishes to the network, but not others. Such change impacts differentially on parishes, and this differential changes will alter long-run population levels and precipitate differential parish population growth.

The impacts of factors may also change. If consumer incomes rise nationally, for example, then relative agricultural prices will change; as a result, grazing may become more profitable and grain production less profitable. The impact on parish population of sand and clay soils, suitable for arable cultivation, will decline, and the impact of gravel and alluvium, suitable for grazing, will increase.

Some of these changes may have lagged effects. Although the construction of a railway may boost population, once construction has ceased and the railway has opened population may decline as the construction workers leave for other jobs, and it may only be some time later, when new industry has moved to the area, and the station has developed as a local railhead, that population growth resumes.

This model suggests, therefore, that population growth in any decade can be explained as the additive effect of changes in the local factors that govern long-run population. Because of lagged impacts, population change in a given parish during a given decade may be due, not only to changes occurring during that decade, but in earlier decade too. When a railway is built, for example, population may rise initially, and then fall, before rising again to a new equilibrium level. Thus population change in any given period may be due not only to railways built during that period but to railways built in earlier periods too.

The impact of such changes in local factors will be mediated by national factors such as agricultural prices, which can alter the impact of factors even when those factors (e.g. soil type) do not change. This effect will tend to show up between consecutive periods, when a given change has a greater or lesser impact than before.

To implement this approach, a regression equation has been formulated. Although the parishes in this study form a panel whose population is tracked over several decades, a conventional panel regression specification is not appropriate for this study because its assumptions about fixed effects at the parish level are far too strict. This study uses a more flexible approach in which parish characteristics can change over time and where the impacts of these characteristics can also change. Thus the proportional rate of population increase in any given parish in any given decade is explained as the sum of the impacts of changes in key characteristics of the parish in both the current and previous decades. Regressions are estimated separately for each decade to allow for the fact that impacts may vary over time in response to changing conditions at county or national level. The coefficients in these regression provide estimates of the changing impacts of various parish characteristics (and of changes in those characteristics) on population growth over time.

3. Sources and their limitations

Level of disaggregation

As already explained, the parish is the basic unit of analysis in this paper. Compared to population studies at the level of the region, county or hundred, a parish study is highly disaggregated. The parish is the smallest administrative unit on which good quality official data is consistently available (although for large parishes information on constituent townships hamlets and chapelries can sometimes be obtained as well). The parish is an ancient administrative unit, and so a great deal is known about the history of many individual parishes.

Parish data has a number of limitations, however.

  • There are different concepts of parish; this study follows the precedent of using the civil parish rather than the ecclesiastical parish; it is the basic administrative unit that levies rates to provide local services.
  • Parishes vary considerably in size (as measured by acreage), and some are very large. As a result, the type of soil may vary from one part of the parish to another. Parishes near a river are often configured as strips running up the sides of the valley – an arrangement that gives each parish access to the river and the surrounding meadows, as well as to uplands and woods. This appears to have reflected an Anglo-Saxon and early medieval view that parishes, and the estates from which they were derived, should, where possible, be self-sufficient in agricultural terms. As a result, the effects of differences between the soils at different altitudes are sometimes difficult to capture using parish data.
  • The thin and elongated nature of many parishes also means that it is difficult to accurately assess the access enjoyed by residents to rivers, roads and railways, since an access point near one end of the village may be a considerable distance from a residence at the other end of the village.
  • Parish boundaries can be irregular. Parishes sometimes have detached portions – some of which may even lie within a different county. In the nineteenth century portions of Oxfordshire parishes lay within the boundaries of Buckinghamshire. By resolving some of these anomalies, the Divided Parishes Act created new problems because transfers of land created discontinuities in the parish population figures.
  • When parishes are very small the migration of a single family can have a large proportional impact on population (as the census enumerators often note). There is therefore a case for grouping small adjacent parishes into a single larger unit, even though this reduces the number of observations available (see below).

When estimating decadal population growth it is crucial that parish boundaries remain the same throughout the decade, unless it is only waste or water that is affected by boundary change. It is necessary to either correct for the boundary change or eliminate the parish from the sample for the decade in which the boundary change occurred. Although the Victoria County History (VCH) claims to adjust population figures for boundary changes using information supplied by the registrar, these adjustments are not implemented on a consistent basis, and so VCH population tables have not been used.

The following strategies have therefore been used.

  • Eliminating parishes subject to frequent boundary changes. This can cause bias, however, because the parishes affected by boundary changes are often the fastest growing ones – especially new suburbs. Elimination is therefore used only for parishes with fairly stable populations.
  • Merging adjacent parishes whose boundaries have changed; this is applied to small parishes but not to large ones.
  • The main method involves creating two separate sets of data for all the parishes affected by a given change. One set of data relates to censuses taken before the change and the other to censuses taken afterwards. Population growth rates for decade before and after the change can then be estimated for all parishes, and only the growth rates over the period of change have to be treated as missing observations.

Selection of the county

Because this study is intended to be a pilot, the initial choice of county was made on grounds of convenience. The county chosen is certainly not representative of England as a whole, since it is rural rather than industrial, and its largest towns – Oxford and Banbury - are of only moderate size. The secondary literature on individual parishes is reasonably comprehensive, however. The Oxfordshire VCH, although not yet complete, encompasses most of the county, and its coverage of economic issues – including land ownership, enclosure and local industry – is good. A new historical atlas has just been published, that not only provides useful mappings, but also offers tentative generalisations that bear upon the theme of the paper (Tiller and Darkes, 2010). This paper can be used to assess how far these generalisations remain robust when subjected to statistical as well as cartographic analysis. Finally, the author has been able to visit every village in order to collect qualitative information ‘on the ground’.