Agriculture in economic development: primary engine of growth or chicken and egg?

Two polar views regarding the centrality of agriculture's role in the process of economic growth are prominent in the literature of economic development. At one pole, a substantial literature argues that agricultural development is necessary for overall economic transformation of a country. The contribution of agriculture in food, raw materials, and financial surplus (including foreign exchange) to invest is essential for the process of industrialization in its early stages, during which by definition, the industrial sector is small (Johnston 1970). At the other pole is the view that economies can always bypass this process of agricultural development and instead invest to build an industrial base. This latter view, popular in the 1950s, also has recent adherents. The question of which view is correct remains open. In this paper, we address the reasons for the lack of resolution of the debate and explore an alternative method for moving forward.

The Findings of Econometric Approaches Have Not Established Causality

It is straightforward to tell a story of why agricultural growth is a necessary condition for a country's economic development in the poorest areas of the world. There the share of the population in agriculture, as well as the share of food in consumption are so high that income generation, or new income streams in the terminology of Schultz (1964), have to come from agriculture if they are to make any substantial national impact. Yet, it appears some countries have managed to grow without a flourishing agricultural sector.

A natural way to try to assess agriculture as a cause of growth is through econometric investigation of cross-sectional data for a panel of countries, or possibly regions within a country. However, this approach is fraught with difficulties that have so far precluded definitive findings. Most notably, the criteria of statistical significance have not provided answers as durable as the confidence intervals on estimated coefficients would lead one to expect. A prime example is a critical review of World Bank economic research on the effectiveness of development assistance (Banerjee et al. 2006). Using country cross-sectional regressions, Burnside and Dollar (1997) found that aid stimulated economic growth, but only conditional on an indicator of good government. This statistically significant finding had been widely touted by Bank management as support for their efforts to make their assistance conditional on satisfaction of good governance criteria in client countries. But with later attempts to replicate the Burnside and Dollar findings with two years of additional data and some alternative specifications, it became clear that the original finding was far from robust, and indeed, it disappeared.

With respect to agriculture in relation to overall Gross Domestic Product (GDP) growth, in a cross-sectional panel of 52 developing countries, Gardner (2005) found no significant evidence of agriculture leading overall economic growth. But in a more sophisticated analysis using Granger-causality tests on very similar data, Tiffin and Irz (2006) found "overwhelming evidence that supports the conclusion that agricultural value-added is the causal variable" (2006). Based on econometric work by Sumarto and Suryadi (2003) on the topic of agricultural growth as related to poverty for Indonesia, Timmer (2005) concludes, "Roughly two-thirds of the reduction in poverty observed during the period of fastest growth in manufactured exports was due to growth in agricultural output at the provincial level." Yet Fane and Warr (2003), in a general equilibrium model of the same economy, conclude, "Contrary to the assumptions of many commentators, the poor do much better if a given amount of GDP growth is produced by technical progress in services or in manufacturing than if it is owing to technical progress in agriculture." As is apparent from broader assessments of this issue, such as those conducted by Valdes and Foster (2005) and Timmer (2005), the results of econometric analyses are inconclusive and even contradictory with one another. Timmer asks, "What are we to make of all this confusion?"

Our view is that economists will simply have to face the fact that econometric studies of country data will not be able to establish causality.

Popperian Approach: Focus on Refutations of Bold Conjectures

The argument that statistical association does not prove causation is a special case of the more general principle of the weakness of confirmations. No amount of confirmation can establish a universal truth, as argued over two centuries ago by David Hume. An alternative approach is Karl Popper's (1965): that the only valid empirical test of a hypothesis is a refutation of it. Popper's approach makes use of the logical strength of refuting instances--one exception can disprove a theory as a universal hypothesis, whereas countless confirmations cannot establish its universality.

To apply this idea one formulates the hypothesis in such a way that it is refutable, and then looks for refuting instances. So, to test the hypothesis that agricultural growth is necessary for general economic growth, it is sufficient to find instances where general growth is achieved without agricultural growth. This leads to a case study approach. It can be viewed as informal non-parametric econometrics. We do not look at the significance of estimated parameters in cross-country regressions. Rather, we look in detail at each country and inquire whether the behavior of variables over time is consistent with the hypothesis under consideration. In the remainder of this paper, we consider in this light both polar claims with respect to agriculture's role in economic development. To do so, we outline four country cases: England (1650-1850), the United States (1800-2000), South Korea since World War II, and People's Republic of China, pre- and post the 1979 agricultural reforms.

Agriculture in English Economic Growth

England was the first country to industrialize, and agriculture had a well-documented role in the process. While economic historians debate the scope of "an agricultural revolution" and its precise dates--starting as early as 1650 for some, there is consensus that the revolution unfolded over the span of a century or more, and its core was "an increase in cereal yields per acre that is the amount of grain that could be produced from a given area of land sown with a particular crop" (Overton 1998). Agriculture's contribution to GDP, estimated at around 43% in 1700, declined to 10% by the 1880s. Roughly 75% of the English population was dependent on agriculture in 1700. But by the late 19th century, the urban population predominated. The industrial revolution, characterized by the increasing application of power-driven machinery (instead of human labor) to manufacturing, started in the mid-18th century and continued into the 19th century.

From around 1770s to 1840s, while money wages went up, real agricultural wages declined with sharp and sustained rises in wheat prices. The main factors pushing prices up were the unprecedented rise in population, especially the increasing non-agricultural population (64% of total by 1801), the Napoleonic Wars (1793-1815), and the continued protection afforded by the Corn Laws (1815-1846). By the 1850s, small farms were a minority. There was increased agricultural investment, which gradually included the purchase of industrially manufactured farm implements. The expansion of export and re-export trades generated wealth that was re-invested in land by the successful merchant class. Thus, by the early 19th century, a virtuous circle of agricultural and industrial integration and expansion had set in.

[FIGURE 1 OMITTED]

The quantity of food exported from agriculture to the urban sector increased by 265% from 1701 to 1820. "The industrial revolution proceeded without a large increase in the import of food and raw materials" (O'Brien 1977). Thanks to the agricultural revolution, neither the Malthusian fears over excess population nor the Ricardian concerns over diminishing returns materialized to choke off the industrial revolution. The polar view that industrial expansion can bypass agricultural development gets no support from England's industrializing experience and in the view of economic historians could not have succeeded without agriculture's contribution.

Agriculture in the U.S. Economic Growth

Until 1830, over 90% of the U.S. population resided on farms, and neither the data available nor historical narratives indicate substantial productivity growth in agriculture. A common view among historians describes "eighteenth and early nineteenth century farmers in New England as trapped by poor husbandry in chronically low-yield, subsistence agriculture" (Rothenberg 1995). Agriculture is implausible as an engine of economic transformation until later in the 19th century.

Figure 1 shows the rise in real GDP per person in the United States over a span of two centuries. Dividing 1800-2000 into four fifty-year periods, the rate of growth of real GDP per person was 0.7% annually during 1800-1850, and 1.7, 2.0, and 2.2% during the succeeding three periods. The growth rate almost tripled during 1850-2000 as compared to 1800-1850. Both the transformation of agriculture and the acceleration of overall economic growth came after 1850. The period after the Civil War (1861-1865) was when we see both accelerated GDP growth and a sharp increase in the non-farm share of the United States population. What was the contribution of agriculture?