Table 2 Overview of variables and associated assumptions

Controls

Age of farm under present management in 2002 (logged)

Production is expected to be curvilinear with respect to age^a

Descendant households

Descendant households are expected to invest in higher production than their predecessors

Area

Area under rice, ha logged

Growth in production is assumed to be largely extensive. Due to diminishing returns, the correlation is expected to be curvilinear

Improved variety

Used improved variety in 2002

Use of improved variety is expected to increase production, since improved varieties are higher yielding than traditional varieties

Used improved variety in 2008

Used improved variety at both considered time points

Stopped using improved variety between the two considered time points

Started using improved variety between the two considered time points

Land type

Used lowland at both considered time points

Use of lowland is expected to positively affect production, since yields are expected to increase

Stopped using lowland between the two considered time points

Started using lowland between the two considered time points

Fertilizer

Used fertilizer at both considered time points

Use of seed fertilizer technology is expected to positively affect production by increasing yields

Decreased or stopped using fertilizer between the two considered time points

Started or increased using fertilizer between the two considered time points

Tractor ploughing

Used tractor plough at both considered time points

Use of tractor ploughing is expected to positively affect production by facilitating an area expansion (extensive growth)

Stopped using tractor plough between the two considered time points

Started using tractor plough between the two considered time points

Commercialization

Sold rice at both considered time points

Sale of rice is expected to positively affect production, since the availability of market outlets is expected to encourage surplus production

Stopped or decreased selling rice between the two considered time points

Started or increased selling rice between the two considered time points

Policy

Government expenditure on agriculture and rural development during the period (lagged and logged)

Increasing government expenditure is expected to positively affect production by providing an enabling environment

Import of rice in relation to gross domestic production during the period (lagged and logged)

Increased import dependence is expected to negatively affect domestic production, since imported rice will drive down local market prices and thereby undermine production incentives

Gross domestic product (GDP) per capita during the period (lagged and logged)

We assume a high elasticity of rice production to economic growth

Distributional dimensions

Proxy for village elite membership

Smallholder-friendly development would decrease the importance of elite membership

Gender of farm manager in 2002

Discrimination of women is expected to negatively affect production

Weather

Drought in E and S. Africa in 2002, village dummy

Control for and estimate of the effects of drought

Floods in W. Africa in 2008, village dummy

Control for and estimate of the effects of flood