skip to content

Keynes Fund

Summary of Project Results

The substantive aim of this project is to examine how and why failures by the market to invest adequately in either public highways or public health gave way to state successes between 1830 and 1912 in England. The intermediate objective was to collect new data on mortality, to develop a dynamic electronic map of the boundaries of urban districts, to construct new measures of investment in local public health infrastructure, and, as a pilot study, to create an electronic map of the roads in the county of Essex. Most of grant funds were used to pay research assistants who collected and digitalised the data from primary sources and for research associates within the Cambridge Group for the History of Population & Social Structure who created the GIS files underlying the maps.

The exponential growth of cities in England between 1801 and 1912 generated a host of economic and social problems, including poor sanitation, lack of clean water, and urban slums. The well-being of urban dwellers was seriously under threat from market failure and public policy responses were urgently required. Within the evolving framework laid down centrally, local authorities found the resources and the political will to deal with these problems, but the responses were varied. To shed new light on the effect of these local sanitary investments on local mortality across urban districts in England (1870 to 1911), the project collected new data on mortality patterns that enabled us to overcome measurement problems affecting previous studies. Preliminary findings suggest that the contribution of local investment in water and sewage systems by local government had little impact on mortality from the 1870s onwards. This points to the possibility that the permissive legal framework operating in the 1850s and 1860s (rather than the state compulsion of the later decades) played a more important role than commonly suggested. Many private waterworks were municipalised between 1850 and 1870. However, some towns, including London, retained private waterworks throughout the nineteenth century, and in these towns it would appear that private water works and thus the market did respond to the emerging public health crisis.

In the 1830s, many trunk roads in Britain were funded by privately-run turnpike trusts. Competition from the railways led to almost complete collapse of the turnpike trusts and a reversion to local authority control. The scope for coordination failure amongst the thousands of local government units put in charge was huge. Yet, it appears that the road mileage was extended and road quality improved, while accommodating a massive increase of local road traffic occasioned by the railways between 1830 and 1911. This is a largely unexplored area of research and our project is the first to undertake a pilot study of one county to quantify expenditure on roads and to produce dynamic maps of the evolution of the network. This is the first step to a fully-fledged study of market failure and state success in road provision.


A new perspective on public investment in sanitation and mortality in England 1870 to 1911

The state response to the public health crises began with the Public Health Acts of 1848, 1858 and 1872-75, which enabled urban areas to set up Local Boards of Health and Improvement Commissioners, and later on to establish Urban Sanitary Authorities. These authorities could levy local property taxes to fund public investment in water, sewers and other health related infrastructure. The acts made certain activities mandatory, but most were permissive. Consequently, big spatial differences emerged in investment in public health which were matched by large and persistent spatial differences in mortality. This is suggestive of a relationship between sanitary improvements and health outcomes. The main objective of this part of the project is to investigate the effect of local improvements in public health infrastructure on improvements or otherwise in local mortality outcomes. Doing so successfully, however, requires overcoming three thorny measurement problems:

  1. Mortality at the relevant spatial resolution. The investments in local public health infrastructure which included waterworks, sewage systems and scavenging were undertaken by the Urban Sanitary Authorities. These authorities had jurisdiction over particular areas and the public health benefits of their investments would, therefore, be concentrated in these areas. The published mortality statistics compiled by the Registrar-General during the second half of the 19th century were reported at the (census) Registration District or sub-district level. These do not correspond to the areas covered by the Urban Sanitary Authorities and this makes these data unsuitable for investigating the effect of investment in local public health infrastructure on mortality. Figure 1 shows a map of Wolverhampton that indicates the potential scale of the problem (in this case, the Registration District contained twice the population of the Urban District, in 1911). To overcome this problem, we digitalized the weekly returns of the Registrar-General, that were compiled from reports of local public health officials in the major urban areas. These data were reported for the same spatial units as the data on investment (namely for the Urban Sanitary Districts). This is the main benefit of these data, but it is also an advantage that they are recorded at a much finer temporal scale (weekly rather than quarterly, yearly or decadal) than any other available mortality data for this period. The downsides are (i) the data are recorded from 1870 only; (ii) they record only a limited number of causes of death, and (iii) only 16 towns are covered for the entire period 1870-1911, although by the end of the sample period the information is available for almost 100 towns. The 16 towns are Norwich, Nottingham, Portsmouth, Salford, Sheffield, Sunderland, Newcastle on Tyne, Manchester, Wolverhampton, Bradford, Bristol, Hull, Leeds, Leicester, Liverpool, and Birmingham. These were the largest urban areas outside London.

  2. Time-dynamic GIS of the boundaries of the urban sanitary districts. The second half of the 19th century saw significant urban growth and, as a consequence, the boundaries of the Urban Sanitary Districts expanded in many places. This complicates matching information from the Population Census to the areas covered by the Urban Sanitary Districts. Starting with a complete boundary map for 1911, we have constructed a time-dynamic GIS for the 16 towns in our main sample. This allows us to link to the individual census returns (the I-CeM datasets) held by the Cambridge Group and to use a bottom-up aggregation approach to allocate all the census information to the spatial units of the Urban Sanitary Districts. We had hoped to construct the time-dynamic map for a larger sample of Urban Sanitary Districts but the existing 1911 GIS turned out to be so inaccurate that we had to reconstruct it. Thus, we now have a complete set of accurate Urban Sanitary Districts boundaries for 1911 and a time-dynamic map for the 16 major Urban Sanitary Districts in our sample. This is a firm basis for constructing time-dynamic GIS dataset for more Urban Sanitary Districts in future work and is sufficient for undertaking the work planned for this project.

  3. Investment in local public health infrastructure. The data needed to quantify the temporal and spatial variation in these investments are recorded in the Local Taxation Returns from 1867 to 1912. For investment in water and sewers, it is not current spending or investment (which is recorded in the accounts) that matter for mortality but the stock of water and sewer capital. The Local Taxation Returns do not record capital and current expenditure on water and sewers separately till 1883. This complicates constructing these stock measures and we had to devise an estimation strategy to overcome this problem. Our solution is to estimate an econometric model that links total spending on each item to capital spending on that item for the years after 1883 and, then, use this model to decompose total spending into current and capital spending for the years between 1875 and 1882. With these data we can construct a unique measure of the stock of water and sewer capital. One of the challenges with this is that six of the towns in our sample had private waterworks throughout the period and three municipalised the town waterworks within the sample period. The time-dynamic GIS enables us to draw on detailed information about occupation from the census (using the I-CeM data collected by the Cambridge Group). In particular, we can construct the number of individuals employed in waterworks as an alternative measure of water provision. This will potentially allow us to extent the analysis back in time to the 1850s and 1860 where fiscal data are not systematically recorded in the Local Taxation Returns.

Figure 1. The Urban Sanitary District of Wolverhampton and the Census Registration District in which it was located, 1911.

The preliminary analysis of these data explores within-town over-time variation in the stock of water (or sewer) capital to explain within-town over-time variation in all-cause mortality and diarrhoeal deaths per births in the 16 town between 1875 and 1912 at the annual frequency. The baseline specification is


baseline specification code


where M is mortality, S is the measure of the stock of water (or sewer) capital in town i in year y. We find that there is no statistical significant association between annual reductions in diarrhoeal or all-cause death rates and the stocks of water (or sewer) capital in the sample of these large urban districts. This finding is robust to alternative measures of the investments, in particular to the use of total loans outstanding, which has been used previously in the literature as a proxy for accumulated investment in sanitation, to different periods and to different estimation techniques. Figure 2 reports some representative estimates of β. The dot is the point estimate while the lines are the 95% confidence intervals. Starting from the left, the first results show that there is a negative raw correlation, but the three next results show that this is not statistically significant when the two-way fixed effect model is fitted either on the sample from 1875 to 1900 or from 1875 to 1911 and with different control variables.

Figure 2. Selected Regression Results

This result is surprising when viewed in the context of existing work on the mortality decline in England and in the USA (see, e.g., the work by Cutler and Miller (2005)). Recent work by Jonathan Chapman (published in the Economic History Review in 2019), which statistically investigates the relationship between outstanding loans (as a proxy for investment in sanitation) and mortality measured at the level of the census registration districts (not at the level of Urban Sanitary Authorities) and as decennial averages (1861 to 1900) for a much larger sample than ours, reports that sanitary investments can account for at least half the decline in mortality. We are able to reproduce this result for our sample of the 16 largest towns if we (i) aggregate the data to decadal level, (ii) use crude mortality rates for the registration district units (rather than for the Urban Districts) and (iii) proxy past investment in sanitation with outstanding loans which capture all types of capital spending, not only those related to sanitation. The difference in results is, therefore, not due to the smaller size or composition of our sample, but to the improvement in spatial match between past investment in sanitation and mortality, to the finer-grained time variation (years rather than decades) and to the refined measure of accumulated investment in water and sewers.

The absence of a sanitation-investment related reduction in mortality, which is also reported on a more impressionistic basis in work by Hinde and Harris (2019), should not automatically be taken as evidence that the huge investments undertaken by the Urban Sanitary Authorities between 1868 and WWI did not contribute to improving public health in urban England, although this is a possibility and more statistical work is needed before firm conclusions can be drawn. We speculate that a number of factors can explain why we cannot find a quantifiable effect. First, the permissive legislation that from 1848 allowed the towns to establish local boards of health and to take out subsidized loans to fund investments in water, sewers and other public health amenities may have been sufficient to induce local policy interventions that reduced waterborne diseases. In other words, these early and more modest investments, which took place before our sample period, were the ones that really mattered, not those that took place later on from the 1870s. If this were the case, it would suggest that municipalisation and legal compulsion were less important than it is commonly thought. Second, the Weekly Returns which are the source of our mortality data did not record typhoid separately. Arguably this was the most relevant waterborne disease after cholera was eliminated by the 1860s, rather than diarrhoeal mortality which was our main outcome measure. Third, the Local Taxation Returns enable us to quantify the stocks of sanitary capital in each town but it does not capture the “quality” of the supply of water or the treatment of sewage or the effectiveness of scavenging. Evidence from the U.S. suggests that filtration and chlorination of water had a hugely beneficial effect on public health (e.g., Cutler and Miller (2005)). While chlorination is not relevant in the English context, other aspects of the quality of supply of water -- filtration and constant supply in particular regular -- might have. We plan to investigate all these possibilities in future work.


Roads: From market to state provision

The project contained a pilot project on the provision of roads in the nineteenth century. Our starting point was the market failure and (presumed/hypothesised) success of state provision of roads in the nineteenth century. Remarkably this has never, to our knowledge, been the subject of any previous investigation. Across the eighteenth century the main trunk roads had been privatised, via (not-for-profit) Turnpike Trusts. This was done because local government (parishes) had proved persistently incapable of maintaining adequate roads as long-distance traffic increased over the seventeenth and eighteenth century. By 1830 around 22,000 miles of roads had been “turnpiked”, covering virtually all long-distance routes. The coming of the railways, from 1830, undermined the toll incomes of the Trusts and over several decades Trusts went bankrupt and roads reverted to parish or other local authorities (a clear example of market failure in changed conditions) with virtually all roads under public control by 1890. We hypothesised that this was a story of state success because there were no major complaints about the state of roads in the nineteenth century (despite a known three-fold increase in road traffic) in sharp contrast to the endless complaints of the seventeenth and early eighteenth centuries. We undertook a pilot study for the historic county of Essex as a preliminary step to a larger study on the state’s success in running and improving highways as private sector provision collapsed.

The pilot study had two elements. The first was to digitise three maps indicating road quality in 1709, 1890 and 1911 to which the greater part of the budget was allocated. The second was to digitise rural highway expenditures in the late nineteenth century, to combine these with urban data already created by Aidt and Grey and to link these to specially created GIS boundary datasets for the relevant authorities, so that we could compare the improvements of roads against highway expenditures.

Figures 3, 4 and 5 show the results of the first part of the pilot. Figure 3 maps roads deemed suitable for artillery during the invasion scare in 1798 in and around the Dengie peninsular. Figure 6 shows the first class roads mapped by the Ordnance Survey in 1890 while Figure 7 shows the first, second and third class surfaced (i.e. macadamised – rammed gravel) roads and the unsurfaced roads in 1911. This exercise took rather longer than anticipated and as a result we were unable to digitise the second and third class roads in 1890. Figures 4 and 5 show the turnpike roads in 1798 and 1830 arising from previous work. One key aim of the pilot was to identify the total improvement to roads between 1670 and 1911 in Essex. In 1911 there were 635 miles of ‘first class surfaced roads’ (65% more than the maximum length ever turnpiked in the county), 2,325 miles of ‘second class’ surfaced roads and 1,402 miles of unsurfaced road and 1,185 miles of unsurfaced roads – mostly leading to isolated farms. The turnpike road network, at its maximum extent in 1819 consisted, by contrast, of only 385 miles of surfaced road. The surfaced roads were almost certainly macadamised in rural areas and stone setts in urban areas. These were the optimal surfaces for horse drawn traffic and the switch to tarmac(adam), the best surface for cars, lorries and bicycles had yet to begin on any scale.

What is immediately clear is that, whilst the privatised turnpike roads dominate the economic history literature on highways, the vast majority (89%) of the improvements to road surfaces was not carried out by the Trusts, but by public authorities. A 1911 parliamentary paper indicates this pattern held nationally. This is not to diminish the undoubted importance of Turnpike Trusts in vastly improving some 22,000 miles (the major trunk routes) by 1830. But, once the railways spread, long distance routes became of limited importance (because virtually all long distance overland freight and passenger travel went by rail). However, we know that the volume of road traffic tripled, but this vast increase in traffic must have been overwhelmingly local funnelling people and goods in and out of railways stations and hence largely on the 80,000 miles of public highways which had always been controlled by public authorities and were never “turnpiked”. Thus, as a result of the pilot, it is now clear that there is a very large story to be uncovered about state success (and private failure) in relation to the road network in the nineteenth century. We can find no literature on this topic beyond some cursory discussion in the works of Sidney Webb and Beatrice Webb.

Two further finding of the pilot may be mentioned here. The 1798 military map shows road qualities on the Dengie peninsular in relation to the capacity to move artillery in winter – a rather demanding quality threshold. It is clear from a comparison of the 1798 military map and the 1798 turnpikes that there were significant mileages of good roads, even in 1798 that were not turnpiked. If this was true in the marshy Dengie peninsular, it is likely to have been true across much of the country. Second, broadly speaking all former turnpike roads were first class roads in 1890 and first class surfaced roads in 1911.

For the second part of the pilot, we digitised rural expenditure data by public authorities from 1875 to 1911 to match the urban highway expenditure data collected by Aidt and Grey. However, it was not possible to match this to the areas covered by the Highway Authorities on the ground. As reported elsewhere, it proved possible to do much less in relation to creating boundary data to match expenditure because of the poor quality of the existing GIS. It was judged best to focus efforts here on the core of the project on sanitary expenditures. In consequence, we could not digitise the boundaries of the rural Highway authorities in Essex. We have not yet had time to analyse the expenditure data.

We undertook a third element, not planned at the time of the grant application, because we realised that we could use occupational data to help pin down change on the roads. Table 1 reports male employment on road work across the period 1851-1911. Macadamised roads required considerable maintenance, but is likely that roads which were not macadamised had very little labour employed on them. Two pieces of national context are striking here. First, the national mileage of turnpike roads in 1830 was about 20% of the national total macadamised mileage of public highways reported in 1911. Second, following the formation of county councils in 1889, there were massive increases in expenditures on roads. In this light, Table 1 may be tentatively interpreted as indicating that (i) the public highway authorities broadly maintained the existing road network 1851-1871 including the former turnpikes, but there is no evidence of improvement; (ii) significant increase in employment began shortly before the formation of county councils in the 1870 and this must have been in the public sector given that most of the turnpike road network disappeared during the 1870s; (iii) the huge jump between 1891 and 1901 is evidence of considerable road improvements under the aegis of the county councils; (iv) the fall-back by 1911 probably represents a reduced level of new road building but there must have been a much greater mileage of surfaced road to maintained by this date; (v) given there six times as many men employed on the roads in 1911 as in 1851, we can hypothesise that the vast bulk of improvements to the 80 per cent of the road network that was not “turnpiked” in 1830 took place after 1871. All these occupational data, and that for earlier years, can be mapped at a parish level, which can then be related to the incidence of turnpike roads in the early period and the road expenditures and mileage of macadamised roads in later periods, once have a GIS into which we map rural expenditures. We will need to find some other funding to create the rural district GIS dataset and then to use this to try to match road expenditures with roads. In the light of our experience with creating the 1911 road GIS for 1911 we plan to explore using new digital technologies for the automated capture of map features in collaboration with colleagues in Istanbul and elsewhere. We are hopeful that this will provide a way to digitise the entire historic road network in both 1890 and 1911 (and perhaps earlier dates) without having to raise large sums to pay for expensive RA time. This will make a large scale follow on project considerably more feasible.

Project Impact

The grant enabled foundational data collection and measurement work. The work on the outputs and dissemination is ongoing.


Two presentations of preliminary work:

  1. Historical Demography Workshop, London School of Economics (February, 2019)

  2. Seminar at Faculty of History, Cambridge (May, 2019)

Planned academic outputs

  1. Aidt, Davenport, Grey. A new perspective on the role of public investment in sanitation and mortality decline in urban England 1870-1911.

    This paper will present the findings of our investigation of the public-health-investment-mortality relationship and be the main output of the project. It will be aimed at a field journal in economic history (the Economic History Review).

  2. Aidt, Davenport, Grey. The transmission of faecal-oral diseases in Victorian cities, a seasonal analysis.

    Diarrhoeal mortality displayed very marked seasonal (summer) peaks in all towns regardless of water supply or sanitary arrangements. The causes of these epidemics (in particular, the sources and modes of transmission of infection) remain unknown. This paper will model the timing of epidemics in relation to meteorological conditions and the dynamics of fly populations to test whether fly-borne or person-to-person transmission provide a better fit to the annual cycles of diarrhoeal mortality.

  3. Aidt, Davenport, Grey. Water, sanitation and health in the diarrhoeal epidemics of 1911.

    This paper will exploit the very hot summer conditions of 1911 to compare summer diarrhoeal and typhoid epidemics in the 97 towns for which we have weekly mortality data in 1911, together with data on water sources and quality c.1910 (Parliamentary Papers, 1914).

  4. We plan to write a VOXeu piece on the mortality debate which will showcase our work on the mortality decline. This will draw out the lessons for less developed countries today and, we believe, stimulate debate.


The project has produced four primary datasets:

  1. The Registrar-General’s Weekly Returns, 1870-1911. This dataset contains mortality data, temperature and precipitation data for 16 to 97 towns in England and Wales at a weekly frequency from 1870 to 1911.

  2. GIS boundary data for the Urban Districts of England and Wales in 1911 (shapefile). Also a dynamic GIS dataset (shapefile) that includes all changes to the boundaries of the 16 major urban districts between 1870 and 1911.

  3. GIS datasets of: the road network of Essex in 1911 (four classes of roads)), first class roads in Essex in 1890) and 1798 roads in the Dengie peninsular classes as to suitability for moving artillery).

  4. Annual expenditures on roads by Rural Districts in Essex, 1870-1911)

We plan to deposit datasets 1-3 for others to use, in the first instance via The Cambridge Group for the History of Population & Social Structure and subsequently in the UK Data Achieve. Datasets 3 and 4 will not be deposited until the pilot work is completed

Research Output

New Perspectives on the Contribution of Sanitary Investments to Mortality Decline in English Cities, 1845–1909

New Perspectives on the Contribution of Sanitary Investments to Mortality Decline in English Cities, 1845–1909, Toke S. Aidt, Romola J. Davenport and Felix Gray, Economic History Review, Vol. 76(2) pp. 624-660 (2023)


Health improved in English cities in the last third of the nineteenth century, in tandem with substantial increases in public spending on water supplies and sanitation. However previous efforts to measure the contribution of public expenditures to mortality improvements have been hampered by difficulties in quantifying public health investments, and the lack of mortality data for specifically urban populations. We improve upon the existing evidence base by (1) creating measures of the stock of Urban District sanitary capital, by type, based on capital expenditure flows, rather than loan stocks; (2) using mortality and capital stock data that relate to the same administrative units (Urban Districts); and (3) studying the period 1880-1909 as well as the earlier period from 1845. The stock of sewerage capital was robustly related to improvements in all-cause mortality after 1880. The size of this effect varied with the extent of public investment in water supplies, suggesting complementary between the two assets. For the period 1845-1884 investments in water were associated with declines in infant and child mortality but the effect was much smaller and less precisely estimated in later decades. Our results suggest that improvements in water and sewerage targeted different transmission pathways for faecal-oral diseases.

Future Plans

The project opens up many avenues for future research in the area:

  1. The dynamic GIS dataset of urban districts will be an important input into ongoing work on social learning and the spread of urban amenities.
  2. The weekly returns dataset offers a major new source of mortality data for the period 1970-1911. It is especially useful for the study of the dynamics of infectious diseases in this period, because it provides much higher temporal resolution than other sources. It also pertains specifically to the larger urban centres, which are aggregated with rural populations in other data sources.
  3. A limitation of this project and other work on the mortality decline in England and Wales is the absence of systematic econometric evidence on the period from 1848 to 1870. Our findings highlight the potential importance of this period, but we cannot use the weekly returns of mortality nor the Local Taxation Returns to pursue this any further. However, the new data on loans collected by Hinde and Harris (2019) for this period can be combined with either the Registrar General’s annual or decadal returns to open up this black box.
  4. The key outcomes of the roads element of the project are that we can now definitely identify the cumulatively massive improvements made to Essex roads by public authorities before 1911 and, have strong grounds to suspect that the improvements began in the 1870s but mainly came in the wake of the founding of county councils in 1889. We are now in a position to cost scaling up the project to national level. With some further pilot work, we will be in a strong position to apply for funding for a larger project on a major state success underpinning economic development in the nineteenth century.


The work carried out during this Keynes Fund project will be followed up in the ESRC grant "Water, Sanitation and Health in the first industrial society: Britain 1780-1930", Dr. Toke Aidt and Dr. Romola Davenport.


Chapman, Jonathan, The contribution of infrastructure investment to Britain's urban mortality decline, 1861–1900, Economic History Review, 72(1), 233-259.

Cutler, David and Miller, Grant, The role of public health improvements in health advances: The twentieth-century United States, Demography, 2005, 42 (1), 1—22.

Harris, Bernard and Andrew Hinde, Sanitary investment and the decline of urban mortality in England and Wales, 1817-1914, History of the Family, 2019, IN PRESS

Figure 3. Roads passable for artillery in the Dengie peninsular Essex, 1798
Figure 4. Turnpike roads in Essex, 1798
Figure 5. Turnpike roads in Essex, 1819
Figure 6. First class roads in Essex, c. 1890
Figure 7. Roads of different types in Essex, 1911

Project Information

Project Code: JHOM
Project Investigators
  • Dr Leigh Shaw-Taylor
  • Dr Romola Davenport
  • Dr Toke Aidt
Research Round
Ninth Round (September 2016)

Project Investigators

Dr Toke Aidt is University Reader and Director of the Keynes Fund at the Faculty of Economics, University of Cambridge. His research areas are Public Choice and Political Economics. His work focuses on four main themes: The Causes and Consequences of Democratization; Development and Corruption; Political Business Cycles; and the Political Economy of Environmental Policy.

Dr Leigh Shaw-Taylor is Senior Lecturer in eighteenth and nineteenth century British economic and social history at the Faculty of History, University of Cambridge. His primary research interests are in (i) long-run economic developments in England from the late medieval period down to the late nineteenth centuries with a particular focus on occupational structure; (ii) comparative work in the same field (iii) the development of agrarian capitalism and (iv) the contribution of transport improvements to the Industrial Revolution.

Dr Romola Davenport is Senior Research Associate, Cambridge Group for the History of Population and Social Structure at the Department of Geography, University of Cambridge. Her research interests are in Mortality, Urbanisation and Migration, Particularly the Long-run Epidemiological Consequences of Urbanisation and Rural-Urban Migration.