The distributional consequences of trade: Evidence from the Grain Invasion

One of the classic insights from international economics is that there are winners and losers from trade. Recent research has pointed towards geography as a dimension along which the distributional consequences of international trade can occur (Topalova 2010, Autor et al. 2013, Dix Carneiro and Kovak 2018). We provide new theory and evidence on these distributional consequences of trade using one of the most influential trade shocks in history – the New World Grain Invasion that occurred in the second half of the 19th century following Britain’s repeal of the Corn Laws (Heblich et al. 2024). ¹

Existing research on this Grain Invasion has focused on economy-wide distributional effects across factors of production (land versus labour) or industries (agriculture versus manufacturing), as emphasised in the traditional Heckscher-Ohlin or specific-factors theories of international trade (Chepeliev and Irwin 2021, O’Rourke 1997, Sharp and Weisdorf 2013). In contrast, the key idea behind our approach is that there is a marked difference in agroclimatic conditions between the western and eastern parts of England and Wales, with western locations more suitable for grass (and hence the grazing of cattle and sheep) and eastern locations more suitable for the cultivation of cereal grains. This difference in initial patterns of specialisation meant that eastern locations were much more exposed to the trade shock from the Grain Invasion than western locations.

We make use of a newly created, spatially disaggregated dataset on population, employment by sector, property values, and poor law transfers for over 10,000 parishes in England and Wales from 1801 to 1901. We begin by showing that Britain experienced a large-scale trade shock from the New World Grain Invasion. Figure 1 displays quantities of wheat production, imports, and consumption (production plus imports) for the UK in 1831–1901 (measured in quarters). ² Imports are disaggregated by major sources of supply, as shown by the three-letter country codes.

Over the 25-year period of 1846–1871, imports of wheat grew more than fivefold, from 1,411,000 quarters to 8,952,000 quarters. Traditional suppliers in Prussia and Russia were replaced with new suppliers in Argentina, Canada, and the US, such that the share of wheat imports from the New World rose to 43%, from 17%. In the next three decades, from 1871 to 1901, imports of wheat increased by an additional 77%. With continuing reductions in transatlantic transport costs from the steamship and railway, the share of Britain’s wheat imports from the New World more than doubled, to 95% from 43%.

Figure 1 UK consumption, production, and imports of wheat, 1831–1901

Notes: Consumption, production, and imports of wheat (quantities) in millions of quarters for the UK based on its 1901 boundaries (England, Wales, Scotland, and Ireland). Consumption is the sum of domestic production and imports. Imports from different source countries indicated by the grey shading and white letters: ARG (Argentina), AUS (Australia), CAN (Canada), IND (India), ROW (Rest of World), RUS (Russia), and US (US).
Source: Sharp (2009).

To examine the distributional consequences of this large-scale trade shock, we exploit the marked difference in agroclimatic conditions between the western and eastern parts of England and Wales noted above. In particular, the warm ocean current of the North Atlantic Drift and the prevailing winds from the southwest generate greater cloud cover, more precipitation, and lower average temperatures in western areas. Many of the more mountainous areas of England and Wales are also concentrated in these western areas (such as the Welsh mountains and the Lake District), with a line of hills (the Pennines) running approximately down the middle of England. As a result, these more rugged western areas typically have thinner and more barren soils. For these combined reasons of climate, terrain, and soil, western locations are more suitable for grass (and hence the grazing of cattle and sheep), while eastern locations are more suitable for the cultivation of corn (historically mainly wheat, but also barley, maize, and rye).

This difference in agroclimatic conditions has been reflected in longstanding differences in agricultural land use between western and eastern regions. In his seminal mid-19th century study of the state of English agriculture, Caird (1852) drew a line approximately down the middle of England that separated the ‘grazing counties’ of the west from the ‘corn counties’ of the east.

As a check on this categorisation, Figure 2 superimposes the Caird line on a map of low-input, rain-fed wheat suitability from the United Nations Global Agro-Ecological Zones data. We find a close correspondence between the Caird line and wheat suitability as measured by the UN. Despite the clear differences between western and eastern areas, we also find some heterogeneity in wheat suitability within each of these areas, which we exploit using our spatially disaggregated parish-level data.

Figure 2 Caird’s western-grazing and eastern-corn counties and wheat suitability

Note: This figure displays the Low-Input, Rain-Fed Wheat Suitability index from the United Nations Global Agro-Ecological Zones; lighter shading corresponds to greater suitability for wheat cultivation.

We begin by providing descriptive evidence of a substantial reallocation of economic activity in England and Wales following the Grain Invasion: (i) within agriculture, from arable to pastoral farming; (ii) between agriculture and other sectors; and (iii) between high- and low-wheat-suitability locations. Between 1871 (the first year of the agricultural census) and 1901, acreage of cereal grains declined by 29%, while acreage of permanent pasture rose by 35%. Between 1851 and 1901, the share of agricultural labourers in employment declined by more than 20 percentage points in eastern locations with the highest levels of wheat suitability, compared to a fall of less than 10 percentage points in western locations with the lowest levels of wheat suitability.

To tighten the connection between this reallocation of economic activity and the Grain Invasion, we estimate event-study specifications using our exogenous measure of wheat suitability. We include parish and year fixed effects to control for time-invariant unobserved heterogeneity across locations and secular trends over time. We include interactions between observed parish characteristics and year dummies to control for other determinants of parish growth rates.

As shown in Figure 3, in the first half of the 19th century, we find no differences in trends between high- and low-wheat-suitability locations. In contrast, in the second half of the 19th century following the Grain Invasion, we find a relative economic decline in high-wheat-suitability locations compared to low-wheat-suitability locations. By the end of our sample period in 1901, we estimate a relative decline in population and property values of 18% and 24%, respectively.

Figure 3 Estimated treatment effects of wheat suitability for log population

Notes: The figure shows the estimated treatment effects from an event-study specification using interactions between years and an indicator variable that is 1 for parishes with above-median wheat suitability and 0 otherwise; vertical lines show 95% confidence intervals based on standard errors clustered by registration district. The specification conditions on parish and year fixed effects, and interactions between year and (i) travel time to the nearest market town; (ii) travel time to the nearest coalfield; (iii) distance to London and distance to Manchester; (iv) Easting and Northing of the parish centroid; (v) an indicator that is 1 for Wales; (vi) an indicator that is 1 for urban parishes based on the 1801 distribution of population densities.

We next develop a quantitative spatial model to account for these empirical findings and evaluate their implications for the distribution of real income. We model parishes as small open economies that face exogenous prices of traded goods determined on world markets. We consider three production sectors: agriculture, manufacturing, and services. Within agriculture, we distinguish between arable farming (cereal grains) and pastoral farming (grazing). Markets are competitive. Goods are produced using labour and land. Each parish is endowed with a continuum of land plots that are heterogeneous in terms of their productivities for alternative land uses. Land is owned by landlords who are geographically immobile. Workers are geographically mobile and have idiosyncratic preferences for each location.

We use the model to undertake a counterfactual for the Grain Invasion, which corresponds to an exogenous fall in the price of arable products on world markets. We follow an exact-hat algebra approach, in which we use the observed values of the model’s endogenous variables in an initial year to control for unobserved location characteristics. We undertake these counterfactuals using both the observed fall in the relative price of arable products from 1846 to 1871 after the repeal of the Corn Laws and its subsequent fall from 1871 to 1901 following the continuing integration of New World producers into global markets. These counterfactuals isolate the general equilibrium impact of the Grain Invasion, holding constant other determinants of economic activity, including the prices of other traded goods, productivities, and amenities.

In response to a fall in the price of arable products, the model predicts a reallocation of land from arable to pastoral use within the agricultural sector; a reallocation of employment away from the agricultural sector towards manufacturing and services; and a reallocation of population away from locations with higher initial shares of arable land use. We find that these reallocations in our counterfactuals (where only the price of arable products changes) are sizeable relative to the observed reallocations in the data (which include other shocks, such as changes in productivities and amenities across locations and over time).

From these counterfactual predictions of our model, we find that the Grain Invasion had substantial income distributional consequences across locations. As shown in Figure 4, the change in the rental rate ranges from reductions of 10% in locations with high wheat suitability to increases of 10% in locations with low wheat suitability. These results contrast with the predictions of the conventional Heckscher-Ohlin model with two goods, two factors, and labour immobility, in which a decline in the price of the land-intensive agricultural good necessarily reduces the rental rate and increases the wage.

Our framework generates much richer income distributional predictions for two main reasons. First, the decline in the price of arable products leads to a reallocation within agriculture from arable to pastoral farming. Since arable farming is land intensive relative to manufacturing and services but labour intensive relative to pastoral farming, this reallocation towards pastoral farming mitigates the reduction in the demand for land from the contraction of the agricultural sector as a whole.

Second, the decline in the price of arable products reduces labour demand in high-wheat-suitability locations and leads to a population redistribution away from these locations towards those with low wheat suitability. This population redistribution magnifies the decline in the price of land in high-wheat-suitability locations and increases the price of land in low-wheat-suitability locations.

Figure 4 Model counterfactuals for the distributional consequences of the Grain Invasion across locations, 1846–1901

Notes: Counterfactual for the observed decline in the relative price of arable products from 1846 to 1901 (19% decline). Grey circles correspond to parishes in our data. Black lines show the linear regression relationship. Vertical axes show either the difference or the log difference between the counterfactual value of a variable and its value in the initial equilibrium. Horizontal axis shows our exogenous measure of wheat suitability (normalised to lie between 0 and 1).

We also examine the aggregate effects of the Grain Invasion on worker expected utility. From 1846 to 1901, we find an increase in worker expected utility of 3.65%, which includes the impact of both trade liberalisation (the repeal of the Corn Laws) and changes in the terms of trade (the fall in the world price of arable products from the increased integration of the New World into global markets). Although this increase in expected utility is substantial relative to conventional estimates of the aggregate welfare gains from trade for a relatively closed country such as the US, it is more modest compared to estimates for smaller and more open countries, such as England and Wales. Additionally, while these conventional estimates are computed at the country level, our estimates of these aggregate welfare gains incorporate reallocation across locations within England and Wales.

Comparing the two sub-periods of 1846–1871 and 1871–1901, we find increases in expected worker utility of comparable magnitude of around 2%. This pattern of results suggests that both the initial repeal of the Corn Laws and the subsequent deepening integration of the New World into global markets contributed to the impact of the Grain Invasion. These findings are in line with the historical narrative, which emphasises the closing three decades of the 19th century as the period of the great agricultural depression, when the full effects of the New World Grain Invasion were felt. Nevertheless, these aggregate welfare gains for workers remain small relative to the distributional consequences across locations.

Source: cepr.org