Climate Stress-Test Model

Model structure

CST26 extends the ENSOID26 causal-loop model to both phases of the El Niño–Southern Oscillation. The dry chain is unchanged from ENSOID26. An El Niño signal (positive ONI) drives drought, fire and haze, which cut output in five exposed sectors. The wet chain mirrors it. A La Niña signal (negative ONI), amplified by tropical cyclone activity, consecutive La Niña years and upstream watershed degradation, drives extreme rain, floods and landslides. These cut output in food crops and horticulture, estate crops, fisheries, mining and transport, and destroy housing and public assets. Both chains feed the same second wave (food prices, household demand, health, tourism, trade, production cost, relief and reconstruction) and the same five macro outcomes. With adaptation set to None and the mode set to El Niño, CST26 reproduces ENSOID26 to machine precision.

The mode switch selects the El Niño chain, the La Niña chain or the full cycle. The full cycle runs an El Niño year followed by a La Niña year, the usual transition seen in 1997–2000, 2015–18 and 2023–26. Damage, growth loss and the deficit are cumulative over the two years. Inflation, jobs and poverty are reported for the peak year. The current account is the weaker year.

Two measures are reported separately. Total economic damage is a welfare and cost concept comparable to the World Bank's accounting of the 2015 fires (1.9 percent of GDP). The GDP growth impact is a smaller flow concept, because much of the fire and flood cost is asset and health loss rather than lost output.

Wet-phase calibration

La Niña damage is scaled to a reference ONI of −1.5, the strength of the 2010–11 event. At that strength and moderate cyclone activity, flood and landslide asset damage is set near 0.45 percent of GDP. This sits above the ordinary-year average of about 0.1 to 0.3 percent of GDP and above the 0.29 percent GDP loss estimated for the single 2025 Sumatra disaster. Mining loss follows the 6 percent production shortfall Adaro reported in the record-rain year of 2010. Food price pressure in wet years comes mostly from perishables such as chili and shallots, so the import buffer works at half strength. Wet years may raise hydropower output, but no quantified evidence was found, so the model sets that gain to zero.

Feedback loops

R1, demand and employment. Weaker household demand reduces employment, which lowers income and further weakens demand.
R2, fiscal strain. Relief, reconstruction and lost revenue widen the deficit and draw down fiscal headroom.
B1, safety net. Relief and subsidies cushion demand and pull the poverty headcount back down.
B2, reconstruction. Public rebuilding after floods adds demand and offsets part of the growth loss.

Cross-model links

Nine links carry the dry phase into the wet phase in Both mode. L1 raises flood runoff and landslide risk in proportion to the fire index, and peat rewetting and reforestation weaken it. L2 lets drought-depleted reservoirs absorb part of the first floods, the one balancing link. L3 moves half of the drought-driven palm oil loss into the wet year. L4 raises wet-season food price sensitivity when dry-year stocks are drawn down. L5 carries 35 percent of dry-year inflation into the wet year. L6 cuts wet-year relief capacity as the dry-year deficit grows. L7 raises the poverty response for households scarred by the dry year. L8 keeps 30 percent of dry-year job losses unrecovered. L9 raises the flood health burden after haze exposure. The compound effect equals the cycle's damage minus the sum of the two phases run alone.

Adaptation layer

Fifteen levers in four approaches (sectoral, economy-wide, nature-based and engineering) each run from 0 to 100 percent of an achievable potential. Each lever reduces specific loss channels by a maximum share at full implementation, and overlapping levers combine multiplicatively so reductions never exceed 100 percent. Adaptive social protection reduces the poverty and demand responses, and disaster risk finance speeds recovery and moves reconstruction off the emergency budget. Avoided damage is the net damage without adaptation minus the net damage with it. It is split across approaches in proportion to each approach's stand-alone effect.

The benefit–cost ratio annualizes avoided event damage at the chosen number of stress events per decade and adds ordinary-year losses the same measures avoid (about 0.15 percent of GDP a year, weighted to floods). An optional switch adds co-benefits in line with the triple dividend framing. Costs are annualized public and private outlays at full implementation. Benefits count only the losses this model represents, so ratios are conservative for measures such as resilient infrastructure whose main return is service continuity.

Key parameters

Parameter	Value	Basis
Nominal GDP base	≈ IDR 22,000 tn / $1.37 tn	2024 nominal, BPS
CPI food and beverage weight	25 percent	BPS CPI basket (2022 base)
BI inflation target	2.5 ± 1 percent	PMK 31/2024
Food-crop damage at ONI +2.0	≈ 10 percent of GVA	ISEAS 2023; USDA-FAS
Fire and haze damage at ONI +2.4	≈ 1.9 percent of GDP	World Bank 2016 ($16.1 bn)
Flood asset damage at ONI −1.5	≈ 0.45 percent of GDP	Scaled from Bappenas, BNPB and CELIOS event data
Mining loss at ONI −1.5	5 percent of exposed output	Adaro 2010 (94 percent of target)
Import buffer for wet-year perishables	0.5 × lever	Chili and shallot price spikes, 2010
Ordinary-year hydromet losses	≈ $2.1 bn per year	Bappenas (Rp 22.8 tn per year); World Bank 2011 (0.3 percent of GDP, all disasters)
Early warning effect on flood damage	up to 25 percent	GCA 2019 (24-hour warning cuts damage 30 percent)
Flood control and drainage cost	$1.2 bn per year at 100 percent	Order of magnitude from the NCICD master plan
Mangrove restoration cost	≈ $3,900 per ha	World Bank 2022
Output multipliers (type II)	1.4 – 1.8	IO-table range
Cross-model links, R1 gain, lever effects	see text	Illustrative

Selected sources

• Aldrian, E. and R.D. Susanto, 2003, "Identification of three dominant rainfall regions within Indonesia and their relationship to sea surface temperature," International Journal of Climatology, 23 (12), pp. 1435–1452.
• Callahan, C.W. and J.S. Mankin, 2023, "Persistent effect of El Niño on global economic growth," Science, 380 (6649), pp. 1064–1069.
• Cashin, P., K. Mohaddes, and M. Raissi, 2017, "Fair weather or foul? The macroeconomic effects of El Niño," Journal of International Economics, 106, pp. 37–54.
• Glauber, A. and I. Gunawan et al., 2016, The Cost of Fire: An Economic Analysis of Indonesia's 2015 Fire Crisis, World Bank, Jakarta.
• Global Commission on Adaptation, 2019, Adapt Now: A Global Call for Leadership on Climate Resilience, GCA and WRI, Rotterdam and Washington, DC.
• Menéndez, P., I.J. Losada, S. Torres-Ortega, S. Narayan, and M.W. Beck, 2020, "The global flood protection benefits of mangroves," Scientific Reports, 10, 4404.
• World Bank, 2022, The Economics of Large-scale Mangrove Conservation and Restoration in Indonesia, World Bank, Jakarta.
• World Bank and GFDRR, 2011, Indonesia: Advancing a National Disaster Risk Financing Strategy, World Bank, Washington, DC.
• World Resources Institute, 2025, Strengthening the Investment Case for Climate Adaptation: A Triple Dividend Approach, WRI, Washington, DC.
• Kiely, L. et al., 2021, "Assessing costs of Indonesian fires and the benefits of restoring peatland," Nature Communications, 12, 7044. Cited from a secondary summary.
• Adaro Energy, 2011, Fourth Quarter 2010 Quarterly Activities Report, Jakarta.
• Bappenas, 2021, Pembangunan Berketahanan Iklim (Climate Resilient Development) policy materials, as reported by LCDI and Bisnis.com. Cited from secondary reporting.
• BNPB disaster statistics for 2021, 2022 and 2024, and the 2025 Sumatra flood assessments, as reported by ANTARA, January 2023 to March 2026.
• CELIOS, 2025, estimate of national GDP loss from the Sumatra floods, as reported by Fortune Indonesia, December 5, 2025. Cited from secondary reporting.
• Bank Indonesia, 2025, "Inflation target and CPI disaggregation," bi.go.id; BPS-Statistics Indonesia.

For research and scenario discussion only. Coefficients are illustrative and should be recalibrated against an input–output table, event-level loss data and project-level adaptation costs before use in published work. Figures marked as secondary reporting need confirmation against the primary source. Not investment, fiscal or policy advice.