The claim is that fluctuations in national income around Yf are mainly caused by the interaction between the multiplier (which amplifies an initial change in AD into a larger change in equilibrium national income) and the accelerator (which links induced investment to the rate of change of output). To evaluate, I distinguish the mechanism's role in PROPAGATION — amplifying and sustaining fluctuations once they have begun — from its role as an IMPULSE — the original disturbance.
The multiplier-accelerator mechanism. Start from a modest positive AD shock. The multiplier turns the initial ΔAD into k⋅ΔAD worth of extra national income through successive rounds of induced consumption. The accelerator then feeds this back into investment: Itinduced=v⋅(Yt−Yt−1), so the newly higher ΔY triggers a wave of capital-widening investment. The extra I is another AD injection, re-fired through the multiplier. In the boom phase, this is a positive feedback loop — AD rises, Y rises, I rises, AD rises further. In the deceleration phase, as ΔY shrinks, induced I collapses (the accelerator's asymmetry: a mere slowdown is enough to zero out induced investment) and the loop runs in reverse — I falls, AD falls, Y falls, induced I falls further. The mechanism naturally generates oscillations.
Case supporting the view. Three strong empirical and theoretical arguments support the claim. First, investment is the most volatile component of AD. GDP peak-to-trough swings in postwar OECD data are a few per cent; investment peak-to-trough swings are typically 20-40%. Much of this volatility fits the accelerator pattern — induced I collapsing in advance of actual output declines, exactly as the model predicts. Second, cycles are stylised, not random. Business-cycle data show recurring expansion-contraction patterns of roughly similar duration (3-7 years for Juglar cycles, matching capital-widening lead times) rather than purely random walks. A mechanical oscillator like the multiplier-accelerator interaction fits this stylisation naturally. Third, the mechanism explains co-movement. Output, investment, employment and consumption all tend to move together across the cycle, which is exactly what the multiplier-accelerator feedback predicts: an initial shock is amplified through consumption (multiplier) and investment (accelerator) in parallel.
Case against — alternative causes of fluctuations. The view overstates the mechanism's role because at least four other factors are important impulses that the multiplier-accelerator model takes as exogenous.
Monetary and credit shocks. Interest-rate cycles directly change the cost of capital and shift autonomous investment independently of ΔY. A central-bank tightening — not any accelerator response — can tip the economy into recession. Credit crunches (2008-9 global financial crisis) collapse I by shutting off the lending channel rather than by changing ΔY.
External shocks. Oil-price shocks (1973-4, 1979-80, 2022), pandemics (2020), natural disasters, geopolitical conflicts and foreign-economy recessions all hit AD or SRAS exogenously. These impulses are not generated by the multiplier-accelerator loop; they are shocks to which the loop then responds.
Expectations and animal spirits. Keynes's own insight was that fluctuations in autonomous investment and consumption often reflect waves of optimism and pessimism not derivable from the data. The accelerator captures the mechanical part; expectations capture the autonomous part, and the autonomous part dominates at business-cycle turning points.
Policy-induced cycles. Procyclical fiscal policy (expansionary in booms, contractionary in busts), inappropriate monetary tightening, exchange-rate pegs breaking under pressure — all generate fluctuations that have nothing to do with the multiplier-accelerator interaction.
Qualifications on the mechanism itself. The accelerator in its simple form assumes a constant capital-output ratio v and no spare capacity. In practice, firms with underutilised capital respond to extra output growth by using it up, not by investing. The accelerator effect is therefore MUTED when the economy has slack — which, ironically, is exactly when the mechanism would be most needed to amplify a recovery. The multiplier itself can be small in open, high-tax economies (see worked example cfm-9), so the amplification stage is also weaker in some settings than in others. Both ingredients of the amplifier are state-dependent.
Reasoned judgement. The multiplier-accelerator interaction is a central propagation mechanism — it explains how modest initial shocks are amplified into much larger output fluctuations, how investment is empirically so much more volatile than GDP, and why cycles have stylised rather than random features. It is not, however, the main source of impulses. Business-cycle impulses come overwhelmingly from monetary and credit shocks, external shocks, policy errors and shifts in expectations about autonomous demand. The strongest formulation is therefore: the multiplier-accelerator interaction explains most of the AMPLITUDE and PERSISTENCE of fluctuations, while monetary, external and expectational shocks explain most of their TIMING and DIRECTION. On this reading the view is half-right — correct about propagation, overstated about causation — and the policy implication is that cycles are stabilised not primarily by trying to interrupt the multiplier-accelerator loop (hard to do without undermining the recovery mechanism itself) but by stabilising the impulses: predictable monetary policy, countercyclical fiscal rules, and policies that anchor expectations about future AD.