Recent Posts

Sunday, July 7, 2019

Can Fiscal Policy Prevent Recessions?

The focus of my upcoming book is on recession forecasting, and not policy responses towards recessions. However, I expect that this is a subject of interest to many of my readers, so I will offer a brief outline of some of the literature.

(Note: this is an unedited first draft of a section from my manuscript book on recessions. For a typical standalone article, it is too long. However, I do not want to spend time stripping out information that should appear in a book. I could have split it into multiple parts, but I did not feel there was a natural splitting point. And yes, I went a bit nuts with footnotes.)

In this section, I am going to sidestep the issue of what neoclassical models say about fiscal policy. This is because I am returning to discuss neoclassical models in the next volume, and I think we need to dig into their rather awkward mathematics in order to properly discuss them. From a practical perspective, I do not think the operational differences between post-Keynesian thinking and the “consensus” neoclassical view is that large in this context – although the way the arguments are presented would be quite different.

Complicating matters, I will be referring to aggregate demand management – which is arguably a consensus approach and could therefore be described as “mainstream.” The preferred framework for aggregate demand management among “mainstream” economists is via the use of interest rate policy, but the sympathy for fiscal policy rises and falls. I am using “aggregate demand management” as that is the most familiar wording, but perhaps a better phrasing for what I mean is “non-targeted stimulus.”

 Aggregate Demand Management

The use of aggregate demand management is not purely a neoclassical view, for example, post-Keynesian authors do suggest such approaches. As an example, in the article “Fiscal Policy in a Stock-Flow Consistent Model,” Wynne Godley and Marc Lavoie[i] argue that fiscal policy can achieve (in theory) all the objectives that neoclassicals claim can be achieved via monetary policy. In the paper, they propose “fiscal policy rules” that are analogous to central bank reaction functions in neoclassical models. In the paper, they note that there could be practical issues with following such rules. Interest rate policy is inherently an aggregate demand management tool, and it is largely only the MMT wing of post-Keynesianism that questions the effectiveness of monetary policy.

The question arises: what is aggregate demand management? It is what results when we approach the question of stabilising economic aggregates using models that themselves are constructed via aggregated sectors. Given the intractability of breaking up economic models into a multitude of actors (although the agent-based literature is moving in that direction), this ends up covering almost all analysis based on macroeconomic models.[ii] For example, the discussion of the fiscal multiplier in {another section of the book} is an example of such an aggregated model, and aggregate demand management.

 In the case of interest rate policy, the policy rate is a single rate that covers the entire currency bloc. (This is an issue for the euro area, where countries with divergent economic trends is covered by the same rate, but many other countries face regional economic divergences.) The effect of the policy rate is transmitted to the entire risk-free curve, which is used to price private sector debts. As such, interest rate policy cannot target individual regions or most sectors. If we widen “monetary policy” to include quantitative credit controls, some notion of regional targeting would be available. (Currently, it is possible for central banks to target residential mortgage rates to a certain extent by intervening into mortgage-backed securities markets. If those securities were segregated by region, regional control might be achievable. It would be more difficult to target industrial sectors under current institutional arrangements, particularly given that crafty corporate treasurers would arbitrage cross-sector funding cost differentials.[iii]) However, quantitative credit controls have fallen out of favour, and are probably not thought of as being part of “monetary policy” at present, even though the central bank would have administered such controls historically.[iv]

Fiscal policy can be either targeted or non-targeted, but in practice, discretionary fiscal policy tends to lean towards non-targeted aggregate demand management.

  • A broad-based personal tax cut or fiscal transfer is an attempt to “get money into the pockets of consumers.” Since the transfer/tax cut are normally designed to hit a wide number of voters, the effects are not targeted at areas of weakness.
  • Tax cuts for business are typically aimed at measures to boost investment: accelerated depreciation allowances, etc. Once again, they are typically broad-based, and so aiming at an aggregate.
  • Some measures were targeted at industries; in the aftermath of the Financial Crisis, there were measures aimed at home renovations and to increase trade-ins of old cars for new ones. These measures are in a half-way zone: they hit sectors of interest, but not regionally targeted.
  • The main “targeted” fiscal interventions are the “automatic stabiliser” policies of the welfare state: unemployment insurance, and to a lesser extent, welfare payments. (I believe that in most countries, the spending on “welfare” is slower moving than the cycle, as it takes time to qualify.) The money is flowing only to households where they have lost work, and so it is targeted only at areas where unemployment is rising. On the flip side, taxes are generally imposed mainly via income taxes and value-added taxes (outside the United States), and so people losing jobs and spending less also reduces the aggregate tax bill in a targeted manner. (In an online conversation, Eric Lonergan raised the concern that automatic stabilisers are not considered “fiscal policy” in some quarters. I would argue that this point of view makes sense from the perspective of central bankers (for example) – since discussing changes to welfare state programmes is normally outside of their domain – but not in the context that I am using here. The parameters of welfare state programmes are fiscal in nature and set by policymakers.)

The use of the terms “targeted” and “non-targeted” suggests my bias: I am in the camp that prefers targeted measures. There is a long line of critiques of non-targeted measures, which I will return to shortly. It would be unfair to argue that economists who discuss “non-targeted” fiscal policy do not understand the advantage of targeting. They would likely respond that it is difficult to expand such targeted measures, so they do not emphasise them. Furthermore, the belief that policymakers can optimally steer the economy – an offshoot of the optimal control theory fad that took hold in the 1960s – has not completely died in some quarters.

Other than the Job Guarantee proposed by Modern Monetary Theory, there are not too many additional avenues to expand the list of possible targeted welfare state programmes. Although increasing the generosity of welfare state programmes in many developed countries seems politically and economically feasible, we should not expect such steps to abolish the business cycle completely. One quick justification of that assertion is that welfare state programmes were generally stronger in the 1950s-1970s, and recessions were arguably more frequent in many of the developed countries. I will return to this point further below, in my discussion of the Job Guarantee proposal.

There is a long history of criticism of aggregate demand management. Hyman Minsky discussed problems with the old Keynesian policies in the 1960s; the essays reprinted in Ending Poverty: Jobs, Not Welfare[v] provide an overview of his thinking. From a theoretical perspective, his complaint was that we need to disaggregate the labour force, and aggregate demand policies ended up stimulate the higher-wage members of the work force (labourers who were either more skilled or fit the biases of employers). The increase in demand did not trickle down to the most disadvantaged, and in-demand workers faced large demand. This helped create an inflationary bias in the economy – in demand workers were able to push through wage increases – even though aggregate employment was relatively soft (since there was relatively high unemployment among low demand workers). I would note that this explanation of “stagflation” – the combination of high unemployment and inflation, defying the Phillips Curve – appears to have very little to do with the consensus neoclassical narrative about stagflation. (I will not pursue this discussion, as I wish to defer the discussion of inflation to another book.)

For a more recent take on the subject, Pavlina R. Tcherneva wrote “Reorienting Fiscal Policy: A Critical Assessment of Fiscal Fine-Tuning” in 2013.[vi] She outlines the criticism of aggregate demand management policies from the perspective of Modern Monetary Theory, which continues the arguments made by Hyman Minsky. Once again, the argument is that the labour market needs to be disaggregated.
Chart: Long Term Unemployment (US)

The figure above depicts one of the empirical problems she observes in the report: the secular increase in long-term unemployment as a share of the total. The percentage of the total unemployed that have been unemployed for at least 27 weeks jumps in each recession, and the minimum reached in each cycle has risen during the recessions after the 1970s. That is, there is a decreasing inability for a segment in the unemployed to escape that fate; dynamism in the labour is fading. The argument is that policy is tilted against such people: they are exhorted to find jobs, yet no effort is made to ensure that jobs exist for them to take.

The argument from MMT is that policy needs to be skewed towards the disadvantaged to reduce the divergences in the labour market. This is presumed to have less of an inflationary effect, since there is less stimulus of the higher wage cohorts. The preferred MMT policy proposal - creating a Job Guarantee that ensures that all workers have access to a job at a fixed wage – is precisely targeted at this cohort. The Job Guarantee wage acts as a de facto minimum wage, but its level will not greatly influence renumeration at higher incomes, much as is the case currently with minimum wage laws.

The Tcherneva article explains how this critique of policy is tied to the MMT critique of the world view provided by neoclassical models. Since I am deferring the subject, the reader may wish to pursue the analysis therein.

 Reactive, not Proactive?

Although I believe that the Job Guarantee is a better policy framework, it is not enough to prevent recessions. Since private sector wages will be set at a markup over the Job Guarantee wage, anyone losing a private sector job would most likely have an after-tax income reduction when they transition to the Job Guarantee. As such, consumer demand will fall, as can private investment. As a result, output will drop, even if there is technically no increase in unemployment. (Since the Job Guarantee is a form of employment, people in the programme would not be “unemployed” under standard definitions. However, under the assumption that popular unemployment insurance schemes are also left in place,[vii] there would likely be a rise in the measured unemployment rate.)

In other words, the Job Guarantee will soften the blow of a recession for workers but will not prevent them from happening. Firms with shaky business plans will still be weeded out.

I would note that if one attaches great significance to expectations and their interaction with the business cycle, that downcast assessment might need to be amended. One could argue that the expectation of a strong auto-stabiliser will cause firms to rationally not cut back on investments in a coordinated fashion. This prevents recessions, other than fluctuations caused by “extreme bad luck” with respect to pesky “external shocks.” However, such beliefs about the self-reinforcing nature of expectations do not seem to be borne out by observed data (since recessions do in fact occur in welfare states). As a result, I will not pursue that argument further.

 Sahm Rule as a Proactive Response?

In the article “Direct Stimulus Payments to Individuals” (initially described here), Claudia Sahm proposed what others have referred to as “the Sahm Rule.” The rule can be summarised as a direct payment to individuals is triggered when the unemployment rate rises beyond a threshold (the 3-month moving average rising by 0.5% over its minimum over the pervious year was the threshold proposed for the United States).

 (To be clear, Sahm’s policy proposal is not being suggested as a method to prevent recessions. Since the unemployment rate would have already risen by 0.5%, a recession may have already occurred. However, one could imagine some econometric tweaking to the rule that could result in a slightly faster reaction time, and hypothetically prevent recessions. Since nobody is proposing such a rule right now, I am discussing the Sahm proposal as the closest approximation of such a hypothetical proactive rule.)

Sahm’s arguments are based on analysis of consumer behaviour in response to stimulus payments in previous recessions. Consumers were more likely to spend out of one-time grants of cash than they would for repeating small payments. As a result, she advocates a single payment, with only a possibility of payments a year later (assuming continued weakness). The suggested size was a payment that results in a total cost of 0.7% of GDP.

Since I believe that the use of fiscal policy is preferable to relying on monetary policy to stimulate the economy[viii], I have some sympathies for the policy. The positive factors I see are as follows.

  • Relatively easy to set up from a political perspective, but I think the politics are a net negative once we factor in the reaction I expect from politicians and political commentators (discussed below). This is aided by the fact that it is being sold as an automated decision rule – which is the preference of “mainstream” economists (although not my personal preference, but I know that I would be in the minority).
  • Since the premise is that the transfer is equal for citizens, it would help flatten income inequality and ensure that there is a higher propensity to consume out of it. (A tax cut presumes one is paying taxes, and the amount that can be cut is going to be in some sense proportional to taxes you pay. As a result, tax cuts can end up being slanted towards households with a lower propensity to spend the proceeds, lowering the effective multiplier.)
  • Since the payment is not tied to the amount of money paid in unemployment benefits (etc.), the amount can be larger than what is implied by job losses. This means we could have a larger effective response than what is possible solely from automatic stabilisers.

Unfortunately, I see several negative factors, which means I would not make a major push in the policy’s favour.

  • Recessions now seem to happen only once per decade. Building an administrative infrastructure that sends out cheques once every ten years is perhaps not the best use of governmental resources. As Sahm notes, the issue is that governments are in the business of taking in tax payments from “everyone,” but currently most governments do not make universal payments. We cannot just reverse the income tax flows, since adults who are below the income tax threshold pay no taxes, nor do they even need to file a tax return (in Canada, at least). Unless tax payments are being made by withholding, the tax authorities do not even know your current address – they only would find out when you file. (This lack of information may not be true for all developed countries; so, the importance of that point varies.) As such, the government needs to build a secure system to make such payments, which then needs to be updated as information technology systems keep evolving. From a practical perspective, it would look like a prime example of wasteful government practices. As someone whose introduction to politics was the rise of the pro-market Reform Party (a prairie populist party), I believe that such a programme would be a sitting duck for similar populists.[ix]
  • Most of the payments will be sent to households that have not yet experienced job losses (unless it is a rapid-moving depression), and so the stimulus will likely be hitting regions that are not yet weak. It seems difficult to worry about stagflation at the time of writing, but this would be unwelcomed in an environment with more inflationary pressures.
  • The bias towards rules-based policy by mainstream economists makes them too easily dismiss the fact that elected politicians can do a good job every so often. The immediate response by fiscal authorities to the Financial Crisis was broadly correct; at most, one could argue that stimulus plans could have been larger. By not blindly applying a rule, expenditures can be targeted at the obvious points of weakness. If they instead rely on the rule, they may fool themselves into believing that they need do nothing else to respond.
  • Fiscal conservatives will fiscal conservative. Although a stimulus package of 0.7% of GDP is not that big a package to throw at a typical recession, that is not how it would be reported in the business press. For the current American economy, the usual tactic would be to argue that the government is spending $147,000,000,000! Since fiscal deficits will be blowing out by even larger amounts, there would be immediate calls to cut spending on social programmes in order to avoid a “financial crisis” caused by “bond vigilantes.” This highly predictable reaction blows up the entire premise of rules-based policy, since the cuts to social programmes is likely to cripple the recovery, as the countries that pushed austerity policies in the post-2010 period discovered the hard way.

 Forward-Looking (Discretionary) Policy?

If we wanted to prevent recessions with recessions, we need policymaking to be more forward-looking. I am not too enthusiastic about the prospects for such a policy framework.

 If we accept my premise that recessions are hard to forecast, it is obviously going to be difficult to forestall them with fiscal policy (whether we attempt to apply a rule or use policymaker discretion). For example, I launched my website six years earlier than this text was written. I do not attempt to forecast the economy, but I do comment on the current situation. For almost the entire period I have been writing, I felt that it was almost always necessary to put is disclaimers about potential recession risks. To be fair to my own forecasting abilities, I was not forecasting a recession, rather I was noting that there were plausible stories about recession risks. If one examines financial market data, we did see periodic episodes that might be viewed as incipient recession scares (the inversion in yield curves that started in 2018 being the strongest such episode). Since pro-active policy cannot wait until a recession is clearly underway, I would argue that in the counter-factual world of discretionary pro-active recession prevention, we would have had at least one round of fiscal stimulus in that six year period – even though no recession occurred (putting aside the uncertain status of the latest scare, which is coincident with my writing of the manuscript).

If we couple this to the Minsky/Tcherneva critiques of aggregate demand management, we could see that this might have impart an inflationary bias to policy. Admittedly, growth in the developed economies have been so sluggish that the inflation risks would have been quite small. (For example, the tax cuts from the Trump administration did little to raise inflation – yet.)

Correspondingly, my view is that we should set our sights slightly lower than preventing recessions. Instead, many developed countries should improve the safety net to better shelter workers from economic volatility (the business sector is very good of taking care of itself) and rely on the instincts of politicians to react to downturns in a sensible fashion. If central banks wish to attempt to prevent recessions with monetary policy, they are welcome to try. However, I am unconvinced about the plausibility of such efforts, which I will discuss in Volume II, where I will look at monetary policy in greater depth.


[i] This appears as Chapter 9 in The Stock-Flow Consistent Approach: Selected Writings of Wynne Godley (edited by Marc Lavoie, Gennaro Zezza), Palgrave Macmillan (2012). ISBN: 978-0-230-29311-3.

[ii] The “micro-foundations approach” of neoclassical models are really aggregates, even though they might be based on a mathematical formalism based on optimising agents. Some models effectively reduce to a single agent (the “representative agent”), while others have a multiple classes of agents. However, these agents are still not tied to a particular geographical location (for example), and are effectively just a way to embed extra parameters into a sectoral behavioural function.

[iii] I would like to thank other Twitter users (including Eric Lonergan and Matthew C. Klein) for raising the point of residential mortgages when I discussed this topic on that platform.

[iv] Using credit controls to target regions is probably going to be more effective to slow down overheating regions than boost weak ones. If a region or industry is facing falling incomes, banks are not going to rush to lend to them, even if their lending cap is increased. This is because credit controls put limits on lending aggregates, but there is no obvious mechanism to force banks to lend to particular entities.

[v] Ending Poverty: Jobs, Not Welfare, by Hyman P. Minsky, Levy Economics Institute, 2013. ISBN: 978-1-936192-30-4.

[vi] “Reorienting Fiscal Policy: A Critical Assessment of Fiscal Fine-Tuning,” by Pavlina R. Tcherneva, Levy Economics Institute Working Paper Number 772, 2013. URL:

[vii] Unemployment insurance seems redundant if a Job Guarantee exists. However, unemployment insurance schemes typically have benefits tied to previous wages. For example, at the time of writing, payments in Canada are 55% of insurable earnings, with a cap: Meanwhile, not being required to work on a Job Guarantee job to get benefits makes job searching easier. The path of least political resistance is to leave both systems in place.

[viii] To be clear, my views about monetary policy are not symmetric. Rapid rate hikes have a proven ability to cause stress in the private sector, and presumably induce a recession. This means that central banks do have the ability to enforce a cap on inflation – if we are willing to accept ugly economic outcomes. Instead, I question the ability of central banks to induce an acceleration in inflation but cutting rates.

[ix] Although I am not a proponent of Universal Basic Income policies, if they did exist, they would require the infrastructure to allow such counter-cyclical payments.

(c) Brian Romanchuk 2019


  1. Would this be better called: can automatic stabilisers prevent recessions?

    1. That’s a big focus, true. But the Sahm rule might not qualify as “automatic”; it’s a rule for policymakers to apply stimulus. (From a controls systems engineering point of view, a feedback rule like that would be considered automatic - the main journal is called the “IEEE Transactions on Automatic Control.” However, that’s not how “automatic stabiliser” is typically thought of in economics.) My coverage of using discretionary fiscal policy to prevent recessions is short, but it is also considered.

      The problem with using discretionary policy to prevent recessions is that you need to forecast them. To what extent my book is original research and not a survey (which is what the book mainly is), my argument is that recessions are hard to forecast. However, those arguments are found elsewhere in the manuscript (and will be most developed in Volume II, after I finish my survey of the existing literature).

  2. Brian, Is there any chance that you might consider using a Kaleckian slant i.e. in using profit rates as an indicator of where the economy is heading or aggregate output rates?

    1. I have a section on the Kalecki profit equation, and generic post-Keynesian business cycle models. I think some preliminary versions were already posted here, but I think they got rewritten later.

      Profits have a “positive feedback loop” relationship with fixed investment, and that’s at the core of my arguments about the non-forecastsbility of recessions. I think I’m essentially restating existing “Keynesian” (including people like Kalecki) views, but I may or may not have a different spin on the topic (which is probably the result of my control systems background).

  3. Brian, weren't the last two significant recessions basically caused by "Wall Street" excesses? Governments these days seem to have gotten more adept at prolonging business cycle extremes but it seems where they may be missing out is in the appropriate regulations area particularly regarding investment banking.


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.