How MARTIN shapes monetary policy in a commodity-exporting economy
Deep Analysis of MARTIN and Australian Macroeconomic Modeling
The Reserve Bank of Australia uses MARTIN (MAcroeconomic Relationships for Targeting INflation) as its primary forecasting and policy analysis tool. This page examines how the model addresses Australia's unique economic structure: heavy mining sector dependence, close trade ties with China, high household mortgage debt, and exchange rate sensitivity to commodity prices.
On this page, we review and present the economic models used by the Reserve Bank of Australia to derive opinions on monetary policy decisions. We provide current input parameters for the MARTIN model to derive theoretical model-based cash rate targets and compare these to market expectations.
MARTIN—MAcroeconomic Relationships for Targeting INflation—serves as the Reserve Bank of Australia's primary tool for economic forecasting and policy analysis. Introduced in 2016 to replace the older AUS-M model, MARTIN represents the RBA's accumulated understanding of how Australia's economy operates and responds to policy changes.
The model takes a middle path between pure statistical forecasting and rigid theoretical models. Where simple statistical models extrapolate historical patterns without economic logic, and DSGE models impose tight theoretical constraints that may not fit Australian data well, MARTIN balances empirical flexibility with economic theory. This "macroeconometric" approach mirrors what central banks like the Federal Reserve (FRB/US) and Bank of Canada (LENS) use for operational forecasting.
Understanding MARTIN matters because it directly shapes monetary policy. When the RBA Board reads in the Statement on Monetary Policy that inflation will likely fall to 2.5% by late 2025, that projection reflects MARTIN's assessment based on current economic conditions, global developments, and assumed policy paths. The model doesn't make decisions—the Board does—but it structures how policymakers think about economic trade-offs.
The RBA can't simply borrow the Federal Reserve's or European Central Bank's models and adapt them with Australian data. Australia's economic structure differs fundamentally in ways that require specific modeling choices:
Mining dominance and terms-of-trade shocks: Mining accounts for roughly 10% of Australian GDP but generates far larger swings in national income when commodity prices move. The 2011-2012 iron ore boom, driven by Chinese infrastructure spending, pushed Australia's terms of trade to record highs—export prices rose 70% relative to import prices over three years. This windfall boosted national income substantially even as non-mining sectors struggled with an appreciating Australian dollar. Standard models that aggregate all production into a single sector couldn't capture these divergent dynamics. MARTIN explicitly separates mining from non-mining investment and consumption, tracking how commodity price shocks propagate differently across sectors.
Variable-rate mortgage exposure: Roughly 65% of Australian mortgages carry variable interest rates, compared to under 10% in the United States where thirty-year fixed mortgages dominate. When the RBA hiked the cash rate from 0.1% to 4.35% during 2022-2023, mortgage payments jumped immediately for most borrowers—adding thousands of dollars annually to debt service costs. This creates much faster monetary policy transmission to consumption than in countries where households lock in rates for decades. MARTIN's household sector incorporates this sensitivity through detailed mortgage market modeling, distinguishing between borrowers facing imminent payment increases and those with fixed-rate buffers.
China trade dependence: China accounts for roughly 35% of Australian exports—primarily iron ore, coal, and LNG. When Chinese steel production slows or construction activity softens, Australian mining revenues fall immediately. This dependence far exceeds what typical trade models capture, requiring explicit China linkages in MARTIN. The model treats Chinese GDP growth and fixed asset investment as key external drivers, recognizing that Chinese policy decisions often matter more for Australian growth than domestic factors.
Exchange rate-commodity price nexus: The Australian dollar functions partly as a commodity currency—appreciating when iron ore and coal prices rise, depreciating when they fall. This creates automatic stabilizing adjustments: commodity price booms strengthen the dollar, making imports cheaper and moderating inflation; price slumps weaken the dollar, supporting non-mining exporters. MARTIN models this endogenous exchange rate response explicitly, rather than treating currency movements as independent shocks.
MARTIN is the Reserve Bank of Australia's primary macroeconomic model for forecasting and policy analysis. Unlike purely theoretical DSGE models, MARTIN follows a macroeconometric approach that balances economic theory with empirical relationships fitted to Australian data. This methodology aligns with similar models at other central banks, including the Federal Reserve's FRB/US and the Bank of Canada's LENS model.
Relative RMSE vs Consensus (1-year ahead):
International Comparison: MARTIN's performance comparable to FRB/US and LENS for domestic variables, superior for trade-related forecasts due to China integration.
Conditional vs Unconditional: MARTIN shows 25-30% RMSE improvement when incorporating near-term staff judgement, similar to other central bank models.
Structural Flexibility: MARTIN's reduced-form equations can capture data patterns that pure DSGE models might miss, while maintaining theoretical foundations through long-run equilibrium relationships.
Australian-Specific Features:
Error Correction Framework: Most equations distinguish between short-run dynamics and long-run equilibrium relationships, allowing the model to capture both cyclical adjustment and structural trends.
GDP Growth: 1-quarter ahead RMSE of ~0.4pp, comparable to consensus forecasts
Inflation: 1-year ahead trimmed mean inflation RMSE of ~0.6pp
Unemployment: 1-year ahead unemployment rate RMSE of ~0.3pp
Conditional Forecasting: Model performance significantly improves when incorporating near-term staff judgement and high-frequency indicators
MARTIN disaggregates the Australian economy into sectors that respond differently to policy changes and external shocks. This structure reflects the reality that Australia's mining boom-bust cycles, housing market sensitivity, and China trade dependence create heterogeneous responses that aggregate models miss entirely.
Australian household behavior centers on housing wealth dynamics more than in most advanced economies. With roughly 70% of households owning their homes and carrying variable-rate mortgages, interest rate changes transmit to consumption rapidly through two channels: direct cash flow effects via mortgage payments, and indirect wealth effects as house prices adjust.
When the RBA hiked the cash rate from 0.1% to 4.35% during 2022-2023, households with $500,000 mortgages saw monthly payments increase by over $1,500. This immediate hit to disposable income forced consumption cuts—retail sales growth slowed sharply within two quarters. MARTIN models this sensitivity by linking consumption to both current income and housing wealth, with elasticities calibrated to Australian mortgage market data showing much stronger transmission than in the United States or Europe.
The wealth channel operates more gradually. Housing price declines following rate increases reduce household net worth, prompting precautionary saving increases even among mortgage-free homeowners who don't face cash flow pressure. This wealth effect builds over 4-6 quarters as price adjustments work through the market, creating persistence in consumption responses beyond the immediate mortgage payment shock.
MARTIN's explicit mining-nonmining investment decomposition addresses Australia's distinctive economic structure. Mining investment—ranging from 2% to 9% of GDP over 2000-2020—responds primarily to global commodity prices and Chinese demand rather than domestic interest rates. When iron ore prices surged from $50 to $200 per ton during 2020-2021, mining firms expanded capacity aggressively despite rising interest rates, because project returns depended on global steel demand, not Australian financing costs.
Non-mining business investment follows more conventional interest rate sensitivity. Manufacturing, construction, and service sector firms face standard cost-of-capital trade-offs where higher rates reduce net present values of expansion projects. But these firms also contend with exchange rate effects: commodity price booms appreciate the Australian dollar, eroding non-mining competitiveness and depressing investment even when domestic demand remains strong. The 2011-2013 period exemplified this tension—strong mining investment coincided with weak manufacturing as the dollar traded above parity with the U.S. dollar.
China's dominance in Australian export markets—absorbing 35-40% of total exports—creates external dependencies that domestic monetary policy cannot offset. When Chinese property construction slowed during 2021-2022, Australian iron ore demand weakened immediately, reducing mining sector income regardless of RBA rate settings. MARTIN treats Chinese GDP growth and fixed asset investment as exogenous drivers, acknowledging that Australian policymakers must respond to Chinese demand shocks rather than influence them.
This dependence operates through multiple channels beyond direct export demand. Commodity price movements driven by Chinese activity affect the exchange rate, with iron ore and coal price fluctuations explaining roughly 30-40% of Australian dollar variation. When Chinese steel production surges, iron ore prices and the Australian dollar both strengthen, creating complex adjustments across mining (benefits from higher prices) and non-mining sectors (hurt by stronger currency). MARTIN captures these cross-sector spillovers, helping the RBA assess whether aggregate demand is strengthening uniformly or diverging regionally.
The flexible exchange rate provides automatic stabilization for an economy heavily exposed to commodity price volatility. When iron ore and coal prices rise, the Australian dollar typically appreciates 10-15%, dampening inflation by reducing import costs even as mining incomes surge. Conversely, commodity price collapses weaken the currency, supporting non-mining exporters and cushioning the broader economy.
MARTIN models this endogenous exchange rate response through uncovered interest parity conditions modified for Australia's commodity currency status. The relationship isn't mechanical—during risk-off episodes, safe-haven flows can appreciate the dollar even when commodity prices fall, as occurred during the 2008 crisis and COVID-19 pandemic. The model incorporates time-varying risk premia to capture these deviations from standard interest parity, improving forecast accuracy during periods of financial stress.
MARTIN represents the Australian economy through interconnected behavioral blocks, with monetary policy transmission occurring through multiple channels:
$$GDP_t = C_t + I_t + G_t + (X_t - M_t)$$
Where each component is determined by sector-specific behavioral equations with explicit Australian features
Consumption Function:
$$\Delta c_t = \alpha_1 \Delta y_t + \alpha_2 \Delta w_t + \alpha_3 \Delta h_t + \epsilon_t$$
Key Feature: High sensitivity to housing wealth and interest rate changes due to variable rate mortgage dominance
Investment Decomposition:
$$I_t = I_{mining,t} + I_{non-mining,t}$$
Employment: Okun's law relationship with output gap, modified for Australia's flexible labour market
Interest Rate Channel: Cash rate → Mortgage rates → Housing investment & consumption
Exchange Rate Channel: Cash rate → AUD/USD → Export competitiveness & import prices
Asset Price Channel: Cash rate → Equity prices & housing values → Wealth effects
Credit Channel: Cash rate → Bank funding costs → Credit supply conditions
Australia's housing market occupies an unusually prominent role in monetary policy transmission compared to other advanced economies, creating both opportunities and challenges for the RBA. This centrality stems from three structural features: variable-rate mortgage dominance, high household leverage, and housing's outsized wealth share.
Roughly 65-70% of Australian mortgages carry variable interest rates, contrasting sharply with the United States where 90% of mortgages lock in fixed rates for 30 years. This difference fundamentally alters monetary policy transmission speed. When the RBA adjusts the cash rate, Australian mortgage payments adjust within 1-3 months as lenders reprice variable loans. American households with fixed mortgages experience no direct payment changes, weakening and delaying policy effects.
The 2022-2023 tightening cycle illustrated this transmission mechanism dramatically. As the cash rate rose from 0.1% to 4.35%, households with $500,000 mortgages saw monthly payments increase from roughly $1,670 to $3,200—an extra $1,530 monthly, or $18,360 annually. For median-income households earning $90,000, this represented a 20% hit to disposable income. Consumption growth decelerated sharply within two quarters as households prioritized debt service over discretionary spending.
MARTIN models this transmission channel explicitly through mortgage payment cash flow effects on disposable income. The model distinguishes between mortgagor and outright owner households, as only the former face direct payment shocks. With roughly 35% of households carrying mortgages and median loan-to-value ratios around 50%, a 100 basis point rate increase reduces aggregate household disposable income by approximately 0.5%, with effects concentrated among younger, higher-leveraged borrowers.
Housing represents 55-60% of Australian household wealth, substantially higher than the 35-40% share typical in European economies. This concentration magnifies wealth effects when house prices adjust to interest rate changes. Empirical estimates suggest Australian households increase consumption by 2-4 cents per dollar of housing wealth gains—a modest marginal propensity to consume from wealth, but applied to massive housing wealth swings.
The 2012-2017 Sydney housing boom exemplified these dynamics. Sydney house prices rose 70%, adding roughly $300,000 to median homeowner wealth. Even at a 3% marginal propensity to consume, this generated $9,000 annual consumption increases per household—boosting retail sales, restaurant spending, and discretionary purchases. When Sydney prices corrected 15% during 2018-2019, the reverse wealth effect contributed to consumption weakness despite stable employment and rising wages.
MARTIN incorporates housing wealth effects through direct wealth terms in consumption functions calibrated to Australian data. The model captures heterogeneity across age groups—older outright owners show stronger wealth effects than younger mortgagors constrained by debt service—and across regions, as Sydney and Melbourne house price movements dominate aggregate wealth fluctuations given their population concentration.
Residential construction accounts for 5-6% of Australian GDP but generates employment multipliers exceeding unity through input linkages. A housing construction boom increases demand for building materials, tradespeople, architects, conveyancers, and real estate agents. When dwelling investment surged to 6.5% of GDP during 2015-2017, construction employment rose to 9.3% of total employment, supporting demand beyond the direct project spending.
Interest rate effects on construction operate through housing affordability and price expectations. Rate increases reduce buyer purchasing power and dampen price growth expectations, discouraging speculative building. The 2022-2023 rate hikes pushed dwelling approvals down 35% from peak levels as developers responded to weakening pre-sale demand. MARTIN models residential investment as responding to both current housing demand (affected by rates, income, population growth) and forward-looking price expectations, with construction lags of 12-18 months from approval to completion creating persistence in the investment cycle.
Mortgage Market Composition (2024):
$\Delta IH_t = \beta_1 \Delta HPI_t + \beta_2 \Delta r_t + \beta_3 \Delta Y_t + \beta_4 \Delta POP_t + u_t$
Where: HPI = House Price Index, r = Real mortgage rate, Y = Real household income, POP = Population growth
Interest Rate Elasticity: -0.7 for housing investment (higher than US due to variable rate dominance)
Wealth Effect: 0.04 (4 cents additional consumption per $1 housing wealth increase)
Regional Heterogeneity: Sydney/Melbourne elasticities ~20% higher than national average
Credit Constraints: APRA serviceability buffer (3% above current rate) incorporated in demand equations
MARTIN incorporates key APRA macroprudential tools:
Model Impact: These constraints create non-linearities in housing demand response, particularly during tightening cycles.
Australia digs up and sells a lot of valuable stuff to other countries. This is such a big part of our economy that MARTIN has to pay special attention to it:
When commodity prices go up:
RBA's Challenge: Balancing the good effects (more export income) with potential bad effects (higher inflation from energy costs)
Export Share Composition (2024):
$TOT_t = \frac{P_{exports,t}}{P_{imports,t}} = \frac{\sum w_i P_{i,t}}{\sum v_j P_{j,t}}$
Where iron ore and coal weights ($w_i$) dominate export price index movements
Direct Effects:
Exchange Rate Channel:
Fiscal Effects:
Short-term (0-6 months): Price shocks immediately affect export revenues and exchange rate
Medium-term (6-18 months): Investment decisions and employment adjustments in resource sectors
Long-term (18+ months): Structural reallocation between mining and non-mining sectors
Asymmetric Effects: Price falls create faster adjustment than price rises due to sunk cost nature of mining investment
China is like Australia's biggest customer. About 1 in every 3 dollars Australia earns from selling stuff overseas comes from China. This makes Australia's economy quite connected to what happens in China.
China makes about half the world's steel, and they need Australian iron ore to do it. When China builds more buildings and infrastructure, Australia benefits.
China burns lots of coal for electricity and steel making. Australian coal helps power China's economy (though this is changing as China goes greener).
Chinese students study at Australian universities and tourists visit Australia. This brings money into the economy through services, not just mining.
Chinese companies and individuals invest in Australian businesses, property, and infrastructure projects.
When China's economy:
The Challenge: The RBA has to manage Australia's economy based partly on what's happening in another country!
Trade Dependence (2024):
$\Delta GDP_{AUS,t} = \gamma_1 \Delta GDP_{CHN,t-1} + \gamma_2 \Delta PMI_{CHN,t} + \gamma_3 \Delta STEEL_{CHN,t} + \varepsilon_t$
Estimated spillover elasticity: 1% China growth → 0.3-0.4% Australian growth
Manufacturing PMI Relationship:
Steel Production Intensity:
Property Market Linkage:
Decarbonization Impact:
Geopolitical Risks:
Model Adjustments: MARTIN incorporates time-varying China sensitivity parameters to reflect evolving economic relationship and structural changes in Chinese demand patterns.
Based on current economic data, here's what the MARTIN model thinks will happen to Australia's economy over the next few years:
2025: 2.1% growth
2026: 2.3% growth
What it means: The economy will grow slowly but steadily - not too fast to cause inflation, not too slow to cause unemployment
End 2025: 2.6%
Mid 2026: 2.5%
What it means: Prices will rise at a reasonable pace - right in the RBA's target range of 2-3%
End 2025: 4.3%
End 2026: 4.1%
What it means: Unemployment will rise a little as the economy slows, but should come back down
Trend: Moderate weakness
Drivers: Lower interest rates, China slowdown
What it means: Overseas holidays might be more expensive, but exports become more competitive
These are just predictions! The actual economy might be different because:
GDP Growth: 2.1% (2025), 2.3% (2026) - below trend reflecting global uncertainty and domestic consumption weakness
Unemployment Rate: Rising to 4.3% by end-2025 as demand-supply imbalances correct, returning to 4.1% by end-2026
Trimmed Mean Inflation: 2.6% (end-2025), 2.5% (mid-2026) - within target band throughout forecast horizon
Cash Rate Path: Model-consistent with additional 25bp of easing through 2025, terminal rate ~3.5%
MARTIN model forecast charts will load here
Interactive visualization of key economic projections
GDP Growth (±2 std dev): 0.5% to 3.7% range for 2025
Inflation (±2 std dev): 1.8% to 3.4% range for end-2025
Unemployment (±2 std dev): 3.7% to 4.9% range for end-2025
Confidence Intervals: Based on historical forecast errors and model residual volatility
Baseline Assumptions:
Upside Scenario (25% probability):
Downside Scenario (25% probability):
Most central banks around the world use computer models to help make decisions. Let's see how Australia's MARTIN model compares to what other countries use:
Style: Practical and flexible
Focus: Housing market, mining, China trade
Strength: Captures Australia's unique economy well
Philosophy: "Use what works best for Australia"
Style: Large and detailed
Focus: Financial markets, consumption, business cycles
Strength: Very comprehensive and well-tested
Philosophy: "Cover everything in detail"
Style: Theory-based (DSGE)
Focus: Multiple countries, currency union dynamics
Strength: Handles complex multi-country interactions
Philosophy: "Theory should guide everything"
Style: Similar to MARTIN
Focus: Commodities, housing, trade
Strength: Good for commodity-dependent economies
Philosophy: "Adapt to country-specific features"
| Model | Type | Equations | Estimation | Key Features |
|---|---|---|---|---|
| MARTIN (RBA) | Macroeconometric | 30+ behavioral | Error correction | Housing, commodities, China linkages |
| FRB/US (Fed) | Hybrid macro | 284 equations | Mixed estimation | Financial frictions, forward-looking |
| NAWM (ECB) | DSGE | Multi-country | Bayesian | Currency union, spillovers |
| LENS (BoC) | Semi-structural | 25+ equations | Cointegration | Commodities, small open economy |
Small Open Economy Focus: MARTIN's structure explicitly recognizes Australia as a price-taker in most global markets, unlike closed-economy focused models.
Housing Market Sophistication: Variable rate mortgage dominance creates faster policy transmission than fixed-rate economies, requiring specialized modeling.
Commodity Sector Disaggregation: Mining/non-mining investment split captures resource boom-bust cycles that aggregate investment equations miss.
China Integration: Explicit China GDP and PMI variables in export demand equations provide superior forecasting for Australia's largest trading partner.
Model Estimation: R/Python hybrid approach using error correction methodology
Real-time Updates: Automated data ingestion with quarterly re-estimation cycles
Scenario Analysis: Monte Carlo simulation for uncertainty quantification
Forecast Combination: MARTIN combined with BVAR and nowcasting models for optimal predictions
Revision Analysis: Model robust to typical ABS data revisions (±0.1pp GDP impact)
Real-time Performance: Out-of-sample testing using historical real-time datasets
Cross-validation: Rolling window estimation to assess parameter stability
International Benchmarking: Regular comparison with IMF, OECD, and market forecasts
If you're interested in learning more about how the RBA works and makes decisions, here are some great places to start:
The RBA's own simple explanations of how monetary policy works, perfect for beginners
Published 4 times a year, explains the RBA's thinking and forecasts
Regular speeches by RBA officials explaining current economic conditions
readrba R Package: GitHub Repository - Access RBA data programmatically
Model Replication: Key MARTIN equations and estimation code available in RBA Research Discussion Paper appendices
Data Sources: Full specification of data sources and transformations in technical documentation
External Review: Pagan and Wilcox (2016) review of RBA forecasting framework that motivated MARTIN development
Forecasting Performance: Regular evaluation in RBA Statement on Monetary Policy forecast error analysis
International Benchmarking: Comparison with other central bank models in academic literature
Policy Impact Assessment: Usage in RBA policy analysis documented in Board meeting minutes
Economic models are not crystal balls! MARTIN and other models help us understand how the economy works, but they can't predict the future perfectly. Unexpected events (like COVID-19), changes in human behavior, or shifts in global conditions can make actual outcomes different from what models predict.
This website is for educational purposes only and should not be used as financial advice. Always consult with qualified professionals before making investment decisions.
Model Limitations: MARTIN represents the current understanding of Australian economic relationships as of 2024. Structural breaks, parameter instability, and model specification uncertainty may affect forecast accuracy. This analysis is for educational purposes and does not constitute investment advice.
Data Accuracy: While we strive for accuracy, data sources may contain errors or be subject to revision. Users should verify critical information with official RBA sources before making decisions.
No Affiliation: This website is not affiliated with or endorsed by the Reserve Bank of Australia. All analysis represents independent research and interpretation.