The Database Arable Land in Sweden, 1810 and 1870
Sweden's historical demographic and agricultural source material is in many respects unique in Europe and the world. A digitizing of this material would make it more accessible to scholars. The current project aims to create an agrarian history database of the arable land and meadow area in Sweden at the parish level between 1810 and 1870. The arable land and meadow area provides important information on land use and production, which can be used in a range of subject areas, for example economic development, environment protection, sustainable development and more. From a Swedish perspective, the nineteenth century is particularly interesting since it is the era of the agrarian revolution. The best remaining opportunity to calculate the arable land and meadow area is contemporary Land Survey maps. By analyzing information in these, the arable land and meadow area can be calculated at the härad (a jurisdictional unit) and parish level. The project will result in a national database freely available to scholars. Owing to earlier projects and initiatives, with this project we would get a digitized quality statistics of the arable land at the parish level area every sixtieth year from 1570 to 1990, something few other countries would be able to produce. The database thereby will be of great interest for scholars also outside of Sweden.
Final report
1. The purpose and development of the infrastructure
1.1 Purpose
The purpose of the project was to use the rich Swedish Land survey material to create a database of arable and meadow areas for the period around 1810 and 1870. The project tried to achieve a much more reliable estimate of the total agricultural output in the country than has previously been possible.
The starting point of the project was the shortcomings of the official agricultural statistics of the era, which was largely based on information from the farmers themselves. As they feared increased taxes, they tended to information that was lower than reality. Researchers have estimated that the actual arable area was 100% larger than reported in 1810 and for 15–30% larger than reported in 1870. However, local studies have cast doubt on such general upward adjustments. The degree of misrepresentation varies greatly within the country.
The Land Survey maps cover Sweden's agricultural area for a number of time periods, though only part of the material was related to the project's focus of 1810 and 1870. The map stock is available as a digital archive on the National Land Survey's website, which makes a macro study possible. The database is based on over 9,500 maps.
The methodological key that allows data from a variety of maps to be combined is the “mantal” of the farms: a measure used in the assessment of land. By combining data on arable and meadow areas per “mantal” –at a sufficiently large number of farms within a certain area, with the total number of “mantal” for the same area, the total arable and meadow area can be estimated.
Based on the arable area, agricultural output was calculated. These calculations needed to involve credible information from official statistics on cultivating systems, the proportion between different crops, and yields. Thereafter, the output was set in relation to the nutritional needs of the population, not least as a way of testing the plausibility of the results.
1.2 Development
The empirical work started in the spring of 2017 with work on the maps. The level of study was the “härad”, the local jurisdiction district, but a lack of data in northern Sweden made it necessary to study larger areas. A total of 231 areas were examined. In the winter of 2019, raw data became available for all parts of the country.
The project then went into a process of quality assurance to detect and fix errors, weaknesses, and uncertainties in the compiled data. An important element was map analysis where data from the project was linked to a GIS program, which provided an overview of regional patterns. Deviating districts were checked to find out if the difference was due to specific local conditions or to errors or pitfalls in the data collection phase.
In 2020, production and consumption (calories) calculations were carried out. They involved analyses of many agricultural historical issues such as yields, grain weights, grain as fodder, grain waste, and calorie contents. An article, “Finally getting a grip on Sweden’s agricultural revolution. A reconstruction of arable land and food production in 1810 and 1870”, describes the method, sources, and literature in more detail.
2. The project's results
2.1 Results
According to the results on an aggregate national level, Sweden had 1.3 million hectares of arable land in 1810, rising to 2.9 million hectares in 1870. This is slightly less than expected, especially for 1810. The official statistics were overall better than researchers have previously understood. The arable area was 67% larger than the statistics from 1810 suggest, and 13% larger than that of 1870. Between 1810 and 1870, the arable land increased by 120% and the population grew by 75%. Agricultural production growth permanently outpacing population growth is the definition of the concept “the agricultural revolution”.
2.2 Argument on the quality of the results
In essence, the project method has proven to be robust. Apart from Lapland and the cities (which were excluded from the investigation), we have succeeded all over the country in producing estimates of the arable and meadow areas. The aim was to have at least 30 land register units (the level at which the “mantal” was set) and 15 maps in each sample. In some areas, however, we had to accept a slightly smaller sample due to lack of available source material.
Can the figures be trusted? A calculation of the statistical variation was made according to the bootstrap model. Bootstrap is based on the calculation of random samples and is used to explore the statistical uncertainty when the sample (in this case per district) is small and a normal distribution cannot be assumed. At the national aggregate level, a bootstrap calculation shows that the confidence interval (95%) is at 2.4% for 1810, and 2.0% for 1870. At the local level, however, the uncertainty is greater.
Three main factors are considered to contribute to the statistical variation:
I. The “mantal” was not only an assessment of the land register unit's arable land, but also of other resources such as meadows, pastures, forests, and fishing. Therefore, the ratio arable land per “mantal” varies largely within the country. Since the arable area is calculated per district, this variation at the national level is not a problem for the project. The variation between land register units within each district, on the other hand, contributes to an increased statistical sensitivity.
II. The size of the sample. The variability increases the smaller the sample.
III. The land register unit was the level at which the “mantal” was set. However, it was difficult to extract the material at this level. Instead, each map had to constitute an observation. The surveyors mapped everything from several villages to parts of a farm. When calculating the arable area of the district, maps that cover many land register units have a large impact. The presence of such maps will, in all probability, also increase the measure of the statistical uncertainty in a bootstrap test. If these maps could have been split up at the land register level, the results (i.e. arable area per district) would have remained the same while the statistical uncertainty would have been lower.
Checks were also made against other source series: 1. The National Geographical Survey exists for some counties from about 1870; and 2. The official statistics (BiSOS) for 1870 of areas based on detailed local inquiries. Together, these reference areas covered close to a third of Sweden. The correlation in arable land per “mantal” at the district level between the project and the National Geographical Survey was 0.97; and between the project and BiSOS 0.93. Aggregated figures on the arable land of the reference areas differ from the project by 0.3% and 1.3% respectively.
The results of these comparisons and the bootstrap analysis demonstrates that our figures have a certain margin of error at the local level, while the figures at the aggregate level (county level and national level) are solid.
3. How the infrastructure was used and what research was initiated with the help of the infrastructure
In connection with the project, two articles have been written which present the results for Swedish and international research communities. Further research includes Mats Morell's agrarian history project on the agrarian revolution in eastern Sweden, to which we submitted area data for relevant counties in advance.
The database closes a gap between Lennart Palm et al.’s reconstructed series of agricultural statistics from 1570, 1630, 1690, and 1750, and the reliable agricultural statistics of the 20th century.
4. Unforeseen technical and methodological problems, as well as deviations from the original plan
The project followed the plan from the application with one exception. According to the application, areas where there were National Geographical Survey maps as early as about 1870 would not be processed. We did this anyway, partly to produce a database based on a uniform material and working method for the whole country, and partly to be able to use the National Geographical Survey maps as reference material (see above).
The investigation revealed a variety of problems that have to do with the fact that agriculture, terminology, assessment traditions, and enclosure processes differ at the regional level. This includes the classification of land areas that cannot obviously be categorized as either meadow or field, for example “åkerslått”, “sval”, and “nybruk”, along the coast of the North (“Norrland”), which was divided between field and meadow in the final tables. Another problem was the reuse of older maps at enclosures (mainly in Skåne County). Close attention thus had to be paid to when the survey was actually done. In Dalarna County and in “Norrland”, the investigation was made more difficult by the fact that information on “mantal” was often missing in the maps. Other source materials then had to be consulted. In some cases, local tax measures needed to be converted into “mantal”.
The excerpting and review of the official agricultural statistics of 1870 (BiSOS N) and the population statistics of 1810–1870, respectively, which were used as a basis for the production and consumption calculations, was time-consuming. The inclusion of this data in the database makes central information available to the research community.
5. The integration of the work in the authority / organization, and how the infrastructure is to be maintained in the long term
The database is handed over to SND. The university has an agreement on long-term storage of data with SND.
6. Infrastructure accessibility and relation to requirements for open accessibility and Open Science
The database is freely available to researchers.
7. Possible international collaborations
International contacts and exchanges have been established through conference participation (see below).
8. Any publications that have been brought about as a result of the research carried out in connection with the infrastructure
8.1 Articles
Even though the project was an infrastructure project with no real requirement of analytical publications, it has nevertheless brought about two articles:
• “Finally getting a grip on Sweden’s agricultural revolution. A reconstruction of arable land and food production in 1810 and 1870” will shortly be submitted to an international peer-reviewed journal, presenting the source material and the method as well as the results at an aggregate national level.
• ”Att komma till ’möjligaste sannolikhet; om icke alldeles träffa sanningen’. Kring präststatistiken 1802–1820 – världens äldsta jordbruksstatistik”, submitted to Scandia (peer-reviewed). This article analyzes the methods and calculations used by the clergy – the body responsible for the collection of data 1802–1820; and puts the statistics in context.
8.2 Participation with papers at conferences on the project and the database:
• European Society for Environmental History (ESEH), Boundaries in / of Environmental History, Tallinn 21–25 August 2019.
• Rural History 2019, the fourth biennial conference of the European Rural History Organization (EURHO), Paris 10–13 September 2019.
• Svenska ekonomisk-historiska mötet, Uppsala 10–12 October 2019.
• Historical Land Registers in Europe: Existing Sources & Current State of Research, University of Innsbruck, 26–28 February 2020
8.3 The third task - contact with the public:
• The article "Dalslands åker and jordbruksproduktion 1810 och 1870" for ”Hembygden”, the periodical for ”Dalslands fornminnes och hembygdsförbund” (in print)
• A lecture at Rinkeby library, March 22, 2019.
9. Links to own web pages
None.
1.1 Purpose
The purpose of the project was to use the rich Swedish Land survey material to create a database of arable and meadow areas for the period around 1810 and 1870. The project tried to achieve a much more reliable estimate of the total agricultural output in the country than has previously been possible.
The starting point of the project was the shortcomings of the official agricultural statistics of the era, which was largely based on information from the farmers themselves. As they feared increased taxes, they tended to information that was lower than reality. Researchers have estimated that the actual arable area was 100% larger than reported in 1810 and for 15–30% larger than reported in 1870. However, local studies have cast doubt on such general upward adjustments. The degree of misrepresentation varies greatly within the country.
The Land Survey maps cover Sweden's agricultural area for a number of time periods, though only part of the material was related to the project's focus of 1810 and 1870. The map stock is available as a digital archive on the National Land Survey's website, which makes a macro study possible. The database is based on over 9,500 maps.
The methodological key that allows data from a variety of maps to be combined is the “mantal” of the farms: a measure used in the assessment of land. By combining data on arable and meadow areas per “mantal” –at a sufficiently large number of farms within a certain area, with the total number of “mantal” for the same area, the total arable and meadow area can be estimated.
Based on the arable area, agricultural output was calculated. These calculations needed to involve credible information from official statistics on cultivating systems, the proportion between different crops, and yields. Thereafter, the output was set in relation to the nutritional needs of the population, not least as a way of testing the plausibility of the results.
1.2 Development
The empirical work started in the spring of 2017 with work on the maps. The level of study was the “härad”, the local jurisdiction district, but a lack of data in northern Sweden made it necessary to study larger areas. A total of 231 areas were examined. In the winter of 2019, raw data became available for all parts of the country.
The project then went into a process of quality assurance to detect and fix errors, weaknesses, and uncertainties in the compiled data. An important element was map analysis where data from the project was linked to a GIS program, which provided an overview of regional patterns. Deviating districts were checked to find out if the difference was due to specific local conditions or to errors or pitfalls in the data collection phase.
In 2020, production and consumption (calories) calculations were carried out. They involved analyses of many agricultural historical issues such as yields, grain weights, grain as fodder, grain waste, and calorie contents. An article, “Finally getting a grip on Sweden’s agricultural revolution. A reconstruction of arable land and food production in 1810 and 1870”, describes the method, sources, and literature in more detail.
2. The project's results
2.1 Results
According to the results on an aggregate national level, Sweden had 1.3 million hectares of arable land in 1810, rising to 2.9 million hectares in 1870. This is slightly less than expected, especially for 1810. The official statistics were overall better than researchers have previously understood. The arable area was 67% larger than the statistics from 1810 suggest, and 13% larger than that of 1870. Between 1810 and 1870, the arable land increased by 120% and the population grew by 75%. Agricultural production growth permanently outpacing population growth is the definition of the concept “the agricultural revolution”.
2.2 Argument on the quality of the results
In essence, the project method has proven to be robust. Apart from Lapland and the cities (which were excluded from the investigation), we have succeeded all over the country in producing estimates of the arable and meadow areas. The aim was to have at least 30 land register units (the level at which the “mantal” was set) and 15 maps in each sample. In some areas, however, we had to accept a slightly smaller sample due to lack of available source material.
Can the figures be trusted? A calculation of the statistical variation was made according to the bootstrap model. Bootstrap is based on the calculation of random samples and is used to explore the statistical uncertainty when the sample (in this case per district) is small and a normal distribution cannot be assumed. At the national aggregate level, a bootstrap calculation shows that the confidence interval (95%) is at 2.4% for 1810, and 2.0% for 1870. At the local level, however, the uncertainty is greater.
Three main factors are considered to contribute to the statistical variation:
I. The “mantal” was not only an assessment of the land register unit's arable land, but also of other resources such as meadows, pastures, forests, and fishing. Therefore, the ratio arable land per “mantal” varies largely within the country. Since the arable area is calculated per district, this variation at the national level is not a problem for the project. The variation between land register units within each district, on the other hand, contributes to an increased statistical sensitivity.
II. The size of the sample. The variability increases the smaller the sample.
III. The land register unit was the level at which the “mantal” was set. However, it was difficult to extract the material at this level. Instead, each map had to constitute an observation. The surveyors mapped everything from several villages to parts of a farm. When calculating the arable area of the district, maps that cover many land register units have a large impact. The presence of such maps will, in all probability, also increase the measure of the statistical uncertainty in a bootstrap test. If these maps could have been split up at the land register level, the results (i.e. arable area per district) would have remained the same while the statistical uncertainty would have been lower.
Checks were also made against other source series: 1. The National Geographical Survey exists for some counties from about 1870; and 2. The official statistics (BiSOS) for 1870 of areas based on detailed local inquiries. Together, these reference areas covered close to a third of Sweden. The correlation in arable land per “mantal” at the district level between the project and the National Geographical Survey was 0.97; and between the project and BiSOS 0.93. Aggregated figures on the arable land of the reference areas differ from the project by 0.3% and 1.3% respectively.
The results of these comparisons and the bootstrap analysis demonstrates that our figures have a certain margin of error at the local level, while the figures at the aggregate level (county level and national level) are solid.
3. How the infrastructure was used and what research was initiated with the help of the infrastructure
In connection with the project, two articles have been written which present the results for Swedish and international research communities. Further research includes Mats Morell's agrarian history project on the agrarian revolution in eastern Sweden, to which we submitted area data for relevant counties in advance.
The database closes a gap between Lennart Palm et al.’s reconstructed series of agricultural statistics from 1570, 1630, 1690, and 1750, and the reliable agricultural statistics of the 20th century.
4. Unforeseen technical and methodological problems, as well as deviations from the original plan
The project followed the plan from the application with one exception. According to the application, areas where there were National Geographical Survey maps as early as about 1870 would not be processed. We did this anyway, partly to produce a database based on a uniform material and working method for the whole country, and partly to be able to use the National Geographical Survey maps as reference material (see above).
The investigation revealed a variety of problems that have to do with the fact that agriculture, terminology, assessment traditions, and enclosure processes differ at the regional level. This includes the classification of land areas that cannot obviously be categorized as either meadow or field, for example “åkerslått”, “sval”, and “nybruk”, along the coast of the North (“Norrland”), which was divided between field and meadow in the final tables. Another problem was the reuse of older maps at enclosures (mainly in Skåne County). Close attention thus had to be paid to when the survey was actually done. In Dalarna County and in “Norrland”, the investigation was made more difficult by the fact that information on “mantal” was often missing in the maps. Other source materials then had to be consulted. In some cases, local tax measures needed to be converted into “mantal”.
The excerpting and review of the official agricultural statistics of 1870 (BiSOS N) and the population statistics of 1810–1870, respectively, which were used as a basis for the production and consumption calculations, was time-consuming. The inclusion of this data in the database makes central information available to the research community.
5. The integration of the work in the authority / organization, and how the infrastructure is to be maintained in the long term
The database is handed over to SND. The university has an agreement on long-term storage of data with SND.
6. Infrastructure accessibility and relation to requirements for open accessibility and Open Science
The database is freely available to researchers.
7. Possible international collaborations
International contacts and exchanges have been established through conference participation (see below).
8. Any publications that have been brought about as a result of the research carried out in connection with the infrastructure
8.1 Articles
Even though the project was an infrastructure project with no real requirement of analytical publications, it has nevertheless brought about two articles:
• “Finally getting a grip on Sweden’s agricultural revolution. A reconstruction of arable land and food production in 1810 and 1870” will shortly be submitted to an international peer-reviewed journal, presenting the source material and the method as well as the results at an aggregate national level.
• ”Att komma till ’möjligaste sannolikhet; om icke alldeles träffa sanningen’. Kring präststatistiken 1802–1820 – världens äldsta jordbruksstatistik”, submitted to Scandia (peer-reviewed). This article analyzes the methods and calculations used by the clergy – the body responsible for the collection of data 1802–1820; and puts the statistics in context.
8.2 Participation with papers at conferences on the project and the database:
• European Society for Environmental History (ESEH), Boundaries in / of Environmental History, Tallinn 21–25 August 2019.
• Rural History 2019, the fourth biennial conference of the European Rural History Organization (EURHO), Paris 10–13 September 2019.
• Svenska ekonomisk-historiska mötet, Uppsala 10–12 October 2019.
• Historical Land Registers in Europe: Existing Sources & Current State of Research, University of Innsbruck, 26–28 February 2020
8.3 The third task - contact with the public:
• The article "Dalslands åker and jordbruksproduktion 1810 och 1870" for ”Hembygden”, the periodical for ”Dalslands fornminnes och hembygdsförbund” (in print)
• A lecture at Rinkeby library, March 22, 2019.
9. Links to own web pages
None.