Using climate models to improve Indonesian food security

Abstract

El Niño Southern Oscillation (ENSO) events exert significant influence on Southeast Asian rice output and markets. This paper measures ENSO effects on Indonesia's national and regional rice production and on world rice prices, using the August Niño 3.4 sea surface temperature anomaly (SSTA) to gauge climate variability. It shows that each degree Celsius change in the August SSTA produces a 1,318,000 metric ton effect on output and a $21/metric ton change in the world price for lower quality rice. Of the inter-annual production changes due to SSTA variation, 90% occur within 12 provinces, notably Java and South Sulawesi. New data and models offer opportunities to understand the agricultural effects of ENSO events, to reach early consensus on coming ENSO effects, and to use forecasting to improve agencies' and individuals' capacity to mitigate climate effects on food security. We propose that Indonesia hold an ‘ENSO summit’ each September to analyse the food-security implications of upcoming climate events.

Acknowledgments

Funding from the National Science Foundation, Stanford's Office for Technology Licensing and the Agency for International Development (via Development Alternatives' Indonesian Food Policy Support Project) is gratefully acknowledged. The authors thank David Battisti, Steven Block, Ashley Dean, Matthew Evans, James Gingerich, Carl Gotsch, Jack Molyneaux, Scott Pearson, Anne Peck, Ning Pribadi, Peter Rosner, James Sanchirico, Karen Seto, Kacek Suryanto, Lester Taylor, Peter Timmer and Roger von Haefen for suggestions and helpful assistance. The usual disclaimers apply.

Notes

In this essay, ‘paddy’ (gabah) refers to unhusked rice, and ‘rice’ refers to milled white rice (beras). Typical milling ratios result in about 66 kg of milled rice from 100 kg of paddy.

The Niño 3.4 SSTA index is one of the leading measures of Central Pacific sea surface temperatures used in climate models. It is composed of temperature measurements from 170 degrees West to 120 degrees West and from 5 degrees North to 5 degrees South. The Niño 3.4 data are maintained by the National Oceanic and Atmospheric Administration (NOAA) and are updated on a weekly basis (NOAA 2003). All future references to SSTA in this article refer to the August value of the Niño 3.4 data series, unless specifically noted otherwise.

All yield, area, production and area damaged data used in this article come from Badan Pusat Statistik (various years), Produksi Tanaman Padi dan Palawija di Indonesia [Production of Paddy and Secondary Crops in Indonesia], Jakarta. Unless otherwise noted, all area data are for harvested area.

The most interesting crop-year equation from the earlier research is:

Both time and time-squared are used in all equations. Use of the same regressors across equations permits the sum of the trimester coefficients to equal those for the entire crop year. By the same logic, only three of the four equations, i.e. the three trimesters and the total, contain ‘real’ statistical information, since any one of the equations could be obtained by knowing the other three. In the sets of equations that follow, however, we present all four in tables 1 –4 for ease of interpretation.

The Dickey–Fuller statistic was used to test whether the errors from the regression of production on t and t-squared were stationary or non-stationary. The test statistic, Z(t), was −5.38, which allowed us to reject the null hypothesis of non-stationarity at a 99% confidence level. The deviations from the t and t-squared equation would also have been an appropriate dependent variable to regress against the August SSTA. We ran these regressions, and the results were virtually identical with those for equation (1). The latter, however, is easier to manipulate econometrically and also easier to explain to policy makers.

This approach is sometimes referred to as a ‘jackknife’ procedure to evaluate skill (Kleinbaum, Kupper and Muller 1988).

We have intentionally kept the regressors the same across tables 1 –4 to permit addition across trimesters. For the May–August trimester, however, there is one other effect that should be noted. This trimester is also being affected by the ‘new’ ENSO conditions that are developing. If the SSTA variable for both the previous and current year are included in the estimating equation for area harvested, both coefficients are statistically significant; they are almost identical in magnitude; they are of different signs; and the adjusted R² is unaffected. While the ‘current’ August SSTA obviously provides limited lead time forecasting purposes, including this variable in table 4 after the fact may provide a additional explanations of how climate affects paddy area. These explanations may be particularly useful given that official estimates are avialable only in June of the following year.

Private sources indicate that Indonesian rice imports peaked in 1998 rather than 1999, at a volume of slightly more than 6,000 tmt (Slayton 2002).

A tiny futures market for rough rice (paddy) currently exists at Chicago's Board of Trade. It is used primarily for domestic marketing in the US. This market's history has been disjointed, and it has never been a significant feature in contracts for rice being traded internationally.

Official Bangkok price data are from USDA (2003).

The mean and standard deviations of the August 3.4 SSTA over that time period are0.04 and 0.91, respectively.