1. RURAL MARKETS, NATURAL CAPITAL AND DYNAMIC POVERTY TRAPS IN EAST AFRICA Discussion of Prototype CLASSES* Model Presently Under Development: A Work in Progress *CLASSES = Crops, Livestock and Soil in Smallholder Economic Systems

2. Problem Statement Why do we see persistent rural poverty? And why does this appear to be both cause and consequence of depletion of natural capital, especially soils? What sorts of policy or project interventions can help smallholder farmers escape from persistent poverty and natural capital degradation?

3. CLASSES Objectives Provide a simulation tool that replicates household-level welfare and resource dynamics that are heterogeneous over time and across initial conditions Learn about system behavior, particularly in response to alternative possible policy or project interventions (e.g., ex ante impact assessment).

4. Implication of Objectives : Develop a dynamic model capable of simulating a “closely coupled” human and biophysical system Model should allow for stochastic exogenous factors (e.g. input and output prices, rainfall, etc.) It should display a disaggregated system in terms of specific enterprises observed in the system

5. Why System Dynamics Approach? It offers a tractable approach to studying a system in which most variables are endogenous and in which out-of-equilibrium behavior is of explicit interest. It is possible to integrate or nest micro-economic or biophysical models in a modular SD model. It does not impose an optimizing behaviour on the farmer based on the assumption of perfect information. Can integrate multiple data sources and types. It offers a simple way of visualizing the structure of feedback and causal loops that generate potentially complex dynamics. Relatively easy to communicate to policymakers and to use.

6. ASSET STOCK - Land quality (soil depth) - Land quantity - Livestock - Labour quality (educ.) - Labour quantity (family size) - Non-farm/monetary CONSUMPTION AND INVESTMENT OUTPUT REALIZATION - Tea - Home gardening - Field crops (maize, rice, beans, etc.) - Fodder - Intensive/extensive livestock - Trees - Non-farm/off-farm income EXOGENOUS FACTORS -Weather -Factor prices -Output prices -Health INPUT ALLOCATION ACROSS ENTERPRISES - Land - Labour - Livestock - Fertilizer State variables Control variables

7. CLASSES: A Simple Description As each period begins, the household allocates its productive assets across different enterprises based on expected returns conditional on exogenous conditions (e.g., expected rainfall, prices). Input allocation choices are based on their relative attractiveness from the farmer’s perspective, reflecting the system design. A production function then generates output of each enterprise based on stochastic realization of exogenous variables. Output affects next period asset stocks directly through impact on the biophysical system (e.g., soil depth, livestock reproduction). Households make consumption and investment decisions that then affect next period’s asset stocks. The cycle begins again.

8. The Temporal Dimension Decisions made at seasonal frequency Value of an activity is the sum of both current and future discounted returns, with all the future compressed into a single value function term. If (endogenous?) discount rate is high and/or current returns do not satisfy immediate needs (e.g., nutritional minima), investments become unattractive.

9. Input Allocation Within The System (Illustrative response to change in exogenous variables) Change in land shares by change in Fertilizer Feed Mean Price Price Wages Rainfall Tea - - ? - - ++ Maize & beans + + + - Fodder crops + ++ + -- Home gardens - - - +

10. CLASSES – Input Allocation Module

11. CLASSES- Crop Production Module

12. CLASSES- Human population dynamics module

13. CLASSES – (Early) Soils Module

14. CLASSES – Livestock Dynamics Module

15. Sample simulation results: land allocation Initial farm land area of 0.8 ha i.e. the farm we visited in Madzu in January 2003 Land allocation decisions are assumed to occur every year Simulate the effect of a change in maize prices on area planted in maize and tea

16. Simulated soil dynamics

18. Soil depth 100 95 90 85 80 110192837465564738291100 Time (Month) Soil depth : With small farm size with Maize Price increases cm Soil depth : With small farm size No Maize Price change cm

19. Group Discussion CLASSES remains very much a work in progress. We need feedback from the broader team on: (a)key phenomena and prospective interventions that will need to be incorporated but might be absent right now, (b)the general feedback and causal loop structure of the model, (c)how best to model some of the specific mechanisms of interest, especially those not yet in this model (e.g., labor allocation or investment choice), (d)presentational issues of relevance to making this accessible and attractive to end-users in government ministries, donor agencies and NGOs.