Winner of the New Statesman SPERI Prize in Political Economy 2016


Wednesday 27 August 2014

Filling the gap: monetary policy or tax cuts or government spending

Suppose there is a shortfall in aggregate demand associated with a rise in involuntary unemployment in a simple closed economy with no capital. Do we try and raise private consumption (C) or government consumption (G)? If the former, why do we prefer to use monetary policy rather than tax cuts?

If consumers have stable preferences over privately and publicly produced goods, then ideally we want to keep the ratio C/G at its optimal level. So if the aggregate demand gap is caused by a sudden fall in C, we will want to do something to raise C. As real interest rates are the price of current versus future consumption, the obvious first best policy is to set nominal interest rates to achieve the real interest rate that gets C to a value that eliminates the consumption shortfall. That is the basic intuition behind the modern preference to use monetary policy as the stabilisation instrument of choice: part of what I have called the consensus assignment.

In classical or real business cycle models this happens by magic. It normally goes by the term ‘price flexibility’, but it is magic because it is rarely explained how a lack of aggregate demand gets translated into lower real interest rates. In the real world, the magicians are central banks. Note that I have not mentioned anything about implementation lags associated with monetary or fiscal policies, which is one of the reasons you will find in the textbooks for the consensus assignment. My reason for preferring monetary policy is more intrinsic than that.

What happens if the aggregate demand shortfall occurs because ‘supply’ increases through technical progress? Once again the first best policy is to lower interest rates to increase consumption, but we would also want to raise public consumption to keep the optimal C/G ratio.

Finally consider a more difficult shock - a ‘cost-push’ shock to the Phillips curve that raises inflation for a given level of output and aggregate demand. We know that we want policy to reduce output (to create a negative demand gap) to partially reduce inflation, assuming that both the output gap and inflation are costly. However it is less obvious in this case that monetary policy is first best. However, as Fabian Eser, Campbell Leith and I showed in this paper, it still is. It turns out we can complicate the model in some ways (but not others) and the result that we use just monetary policy to maximise social welfare still holds.  

If we return to the case of a demand gap caused by a fall in consumption, suppose we cannot use monetary policy because nominal rates are stuck at zero. As we want to increase private consumption, the obvious alternative to try is a tax cut. If we had access to a lump sum tax (a tax that is independent of income, like the poll tax), and if consumers responded to a tax cut, then this would work pretty well too. There are two problems: Ricardian Equivalence, and there are no lump sum taxes.

If Ricardian Equivalence held completely tax cuts would be totally ineffective at stimulating consumption, but the consistent evidence is that Ricardian Equivalence does not hold. But this evidence does suggest that at least half and perhaps more of any tax cut would be saved, which means that tax cuts would have to be relatively large in money terms compared to the consumption gap. It also adds a degree of uncertainty to their effectiveness. If there is some financial limit on the size of any stimulus package (as often seems to be the case), this puts tax changes that rely on income effects at a severe disadvantage. Even if financial limits are not present, the relative ineffectiveness of tax cuts in stimulating consumption is a problem for another reason.

Lump sum taxes do not exist, so some distortionary tax (a tax that influences incentives) has to be used. This means that a tax cut violates tax smoothing. This is the idea that the best policy is to keep tax distortions constant. A tax rate of 30% is better than a tax rate of 10% in odd years, and 50% in even years. So filling the consumption gap with a cut in the income tax rate (to be followed by increases in that rate) has a cost. The more tax cuts are saved, the bigger the cost. It is highly unlikely that this cost will be sufficient to stop us trying to fill the consumption gap, because unemployment costs are far greater than uneven tax distortions. However there are costs, unlike the first best of changing real interest rates.

In contrast, using public spending to fill any demand gap is much more straightforward, as its impact on demand and employment is more predictable. But it too has a cost: we get the C to G balance wrong (too much G compared to C). Chris House has a recent post on tax cuts versus government spending as alternative means of fiscal stimulus. (Noah Smith wrote a subsequent post and Chris responded.) The proposition he wants to put forward is that government spending should only be used as a stimulus measure if its social benefits outweigh its social costs. I’m not sure that is a very helpful way of thinking about it. Far better, in my view, is to accept that the demand gap must be plugged (because the costs of not doing so are very large), and then work out the way of doing that which leads to the lowest collateral damage. That might well be an increase in G rather than a tax cut. It will almost certainly be so if there is a financial limit on the size of the stimulus.

The same reasoning can and should be applied to unconventional monetary policy, but that has to be another post.



35 comments:

  1. In real business cycle models the implicit mechanism should be the following. A shortfall in aggregate demand reduces prices today. As the shock is temporary, prices are expected to rise back to the initial level in the future. Inflation expectations rise, thus reducing the real rate. A possible alternative to the use of the nominal rate at the zero lower bound is to use consumption taxes. A lower VAT tax today and a higher tomorrow should work in the same manner as an interest rate cut.

    ReplyDelete
    Replies
    1. The mechanism you assume takes it as given that the monetary authority has some form of fixed nominal target. The mechanism you describe will not work with an inflation target. My point is simply that the RBC model presumes an appropriate monetary policy, but they never say that.

      Delete
    2. You are right, such a mechanism would work with a fixed money supply, for instance. Anyway in the RBC model the unit of account is the consumption good itself and there is no monetary policy. Suppose people start discounting the future less, which should produce a fall in "aggregate demand" today. The real rate falls. If there is capital investment will rise. If there is no capital, and my intuition is right, consumption and output should actually decline: in this sense we could say that aggregate demand shortfalls cause a recession also in the RBC model. Nonetheless, I doubt aggregate demand is well defined in RBC models.

      Delete
    3. Probably, for this mechanisms to work the shock should hit differently leisure and consumption. People would prefer to consume less and have more leisure. Is this aggregate demand or aggregate supply? I think it is impossible to say.

      Delete
  2. Hi Simon,

    You write that the cost of government stimulus is that the balance between government and private spending is thrown out of whack. For the benefit of the economically illiterate, why is this bad? What happens when government is spending too much relative to private consumers?

    ReplyDelete
    Replies
    1. We have too many hospitals, and not enough restaurants. And if you think that does not sound too bad, then that probably means you are living under a government that has a lower desired G/C than you do.

      Delete
    2. Wait. The western world has chosen progressive government. Doesn't this imply a choice for higher G during recessions (as more people land on the safety net) at the same time C is low? Doesn't the choice for progressive taxation also imply lower-than-expected taxes when incomes are lower than expected? Therefore, doesn't preference for progressive government imply preference for variable G/C? If so, then the following assumption is wrong "...consumers have stable preferences over privately and publicly produced goods...".

      I'm no economist, so I apologize if these questions are naive. Thanks for the interesting discussion.

      Delete
    3. I like questions from non-economists, because I want to make any post that I do not label as 'for economists' as accessible as I can. G, government consumption, is things like schools, defence etc, rather than things like like unemployment benefits. In other words G is government spending on stuff, rather than income transfers.

      You are right that what economists call 'automatic stabilisers' - increasing transfers - will operate in a recession: the issue here is whether to go further, by raising G. You could of course raise transfers like welfare benefits. These would work more like taxes - some would be saved, and there might be incentive effects.

      Delete
    4. I see. So questions about the relative merits of automatic stabilizers vs monetary policy differ from the questions you are focussed on here. Thanks again!

      Delete
  3. "simple closed economy "

    This seems to render any observations moot. The whole point of US, UK, Japanese, Swiss, and Swedish monetary interventions has been to lower (or cap) the respective country's exchange rate, to let foreign consumers provide extra demand. It's simply beggar-thy-neighbour with a little more obfuscation to keep criticism to a minimum.

    ReplyDelete
  4. It seems to me the Ricardian equivalence issue should apply equall to government spending increases. An increase in G implies a future tax increase, and thus motivates current savings. Under conditions giving complete Ricardian equivalence, a $1 per capita increase in G implies (with zero interest) a $1 per capita tax increase after the recession and I must therefore save $1 to pay my coming bill -- that is, there is a $1 per capita decrease in C. (That's putting aside Keynesian multiplication of C, but that too should be equivalent in both cases.)

    Is that wrong? Or does the asymmetry arise only as we consider the reasons for incomplete Ricardian equivalence? Why?

    Much the same seems to hold for the tax distortions point.

    ReplyDelete
    Replies
    1. It is wrong. Even if the increase in G is funded by taxes, because the G increase is temporary and consumers smooth the impact of any tax change, there will be a positive impact on demand. But if higher G is paid for by lower G in the future, there is no problem - just G/C too high today, and too low in the future

      Delete
    2. Ricardian equivalence works through consumption smoothing. If taxes are raised by 1$ per capita, you will reduce your income but under RE you will smooth out this reduction over your lifetime. Say you expect to live another 50 years, that means you will spend 1/50=0.02 (two pennies) less this year and every year in the future as a result of the tax raise.

      Delete
    3. Thank you, Simon, that is crystal clear -- I had added an assumption about G smoothing which need not be true. But then, of course, there is likely a loss from fluctuating G, with a clear exception when you are talking about government investment that pays for itself, or accelerated maintenance. That is specifically NOT the kind of spending that Noah Smith and Chris House were debating.

      Delete
    4. On reflection, if we are allowing unfettered departures from desired G/C paths, there is no reason for classic Ricardian equivalence in the tax case, either. A tax cut could cause expectations of lower G rather than a future tax increase. And if the lower G is on pure public goods the lack of which can't be compensated by smoothing private consumption (e.g., national defense) there is no reason to alter private consumption at all in response.

      So the answer to whether my initial point was wrong ,or the asymmetry is rooted in conditions that preclude Ricardian equivalence, is, perhaps, "Both".

      Delete
  5. I think you will find that the queries you mention, were resolved in Prof Bill Mitchell's post:- "The deficit terrorists have found a new hero. Not! Posted on Thursday, July 29, 2010 by bill".

    His reference to the paper by "Gauti B. Eggertsson – What Fiscal Policy is Effective at Zero Interest Rates?" is worth a read. Particularly parts 5; 6 and the Conclusions.

    "There are two general lessons to be drawn from this paper. The first is that insufficient demand is the main problem once the zero bound is binding, and policy should first and foremost focus on ways in which the government can increase spending."

    " ... the effect of tax cuts and government spending is fundamentally different at zero nominal interest rates than under normal circumstances."

    ReplyDelete
    Replies
    1. Maybe you could give us the short version why tax cuts don't work at all?

      I believe the answer will have to do with the marginal propensity to consume. Tax cuts and helicopter drops are better than nothing at all. Government spending is even better.

      "" ... the effect of tax cuts and government spending is fundamentally different at zero nominal interest rates than under normal circumstances.""

      I don't see why this is the case. I'd say they are different once the output gap is filled and they cause inflation rather than when the economy is suffering from a lack of aggregate demand.

      Delete
    2. This comment has been removed by the author.

      Delete
    3. I believe you were replying to my queries. So let me say first, thank you. And second, what I am after here is above all theoretical clarity -- for myself, but even more so for students whom I am trying to teach to think in tractable old-Keynesian models about some things which really don't come into clear focus in my own mind without dynamic models. I believe Simon is way ahead of me on the latter mission, which is why I push for more.

      I don't think the Eggerttson mechanisms are examples of either of the sources of asymmetry in this post. I believe, for example, his negative multiplier on labor tax cuts results from an increase in labor supply coupled with a variant on the "paradox of toil" he developed more fully in papers with Krugman and Mehrotra. I don't see any way that is like the Ricardian or distortion points above.

      Delete
    4. Peter et al, I think that increased government spending and tax cuts both work to add spending power to the private sector. I think the point is that tax cuts are slightly less effective due to the propensity for households to save some of it, creating a leakage from the money-go-round. Hence the government spend scores better.

      The high multiplier for government spending that is not a substitute for current private sector spending is significant. For instance a new runway here; a high speed train there. Or a national pothole filling project. ATB Acorn

      Delete
  6. Market Fiscalist27 August 2014 at 15:34

    "As real interest rates are the price of current versus future consumption, the obvious first best policy is to set nominal interest rates to achieve the real interest rate that gets C to a value that eliminates the consumption shortfall"

    Having read Nick Rowe's blog for a while I'm not sure I agree with this. Rather I would say the first best policy is to raise people's expectations of future NGDP by being prepared to increase the quantity of base money until those expectations are met. This may have the effect of raising the interest rates rather rather than lowering them as spending plans may increase both in the present AND in the future ue to higher income expectations.

    On Noah Smith and Chris House: I took Chris's point to be that any direct govt investment needs to be justified on a cost/benefit basis independent of any boost to GDP it provided via a multiplier effect. If a bridge costs $100 but only provided $90 of benefits then it is a bad investment even in a recession. Any multiplier effect can be achieved just as well by giving the money directly to the bridge builders (or those with similar savings/consumptions patterns). The hard bit of course is figuring out the real benefits of govt spending.



    ReplyDelete
    Replies
    1. On Nick: To get the first best you have to raise C today. If you do that by promising higher than ideal inflation tomorrow, or higher than ideal C tomorrow, that is a cost, so it cannot be as good as getting the right real rate today.

      On Chris: Suppose government investment of 100 gets rid of the gap, then that is a benefit. Any additional social benefits of this investment are on top. I think Chris tries to combine the two which is just confusing. Talking of multipliers is also confusing, which is why I did not.

      Delete
    2. Market Fiscalist27 August 2014 at 16:25

      "To get the first best you have to raise C today. If you do that by promising higher than ideal inflation tomorrow, or higher than ideal C tomorrow, that is a cost, so it cannot be as good as getting the right real rate today."

      I agree that the aim is to increase C today. In some ways saying "C can be increased by lowering interest rates" and "C can be increased by increasing the qty of base money" are 2 ways of saying the same thing as OMO involves both.

      However assuming that the use of OMO to purchase assets for new money is successful at boosting C then this will tend to cause interest rates to rise . It seems more logical to say "it was the increase in base money that caused AD to increase and subsequently interest rates to increase" than "lowering interest rates boosted C, and this then led to interest rates rising". This second version seems to ignore the fundamental role that the money supply change (or more accurately NGDP expectations driven by this change) has both on C and interest rates (and of course this becomes crucial when discussing the ZLB and monetary policy).

      Delete
    3. I agree with your last line, which is why I think my way of talking about it makes much more sense. Central banks set interest rates, and that is what everyone sees. They do not see a change in base money, and then change their expectations. Base money changes AD through interest rates. Why do you think it logical to think otherwise?

      Delete
    4. Good question.

      I think possibly because I am imagining a theoretical world where the CB doesn't case about interest rates - it just uses assets purchases to adjust the monetary base in order to manage AD and lets the money market set interest rates. In this case I think it is clear that base money is not changing AD via interest rates.

      Of course CBs don't do this but rather use interest rates as the main monetary policy tool. This gives the impression that interest rates are controlling AD via an inter temporal effect. This is partially true. But I think the bigger effect is that to achieve its rates target the CB has to buy assets for new money and its this new money entering the economy that drives most of the changes in AD, just as in the version above where the CB doesn't care about rates and just adjusts the base directly to control AD directly.

      Delete
  7. Good clear post, Simon.

    Presumably, the initial shock in the first paragraph is a drop in the rate of subjective time preference? That would indeed require (assuming no investment in the model) an equivalent drop in the real interest rate for first best. And that drop in the real rate would, for given expected inflation, a drop in the nominal rate, and an increased demand for money, so the central bank should increase the supply of money, to prevent a recession.

    "I’m not sure that is a very helpful way of thinking about it. Far better, in my view, is to accept that the demand gap must be plugged (because the costs of not doing so are very large), and then work out the way of doing that which leads to the lowest collateral damage."

    I think there ought to be some sort of envelope theorem which says these are all equivalent. If you are forced to cut tax rates to prevent a recession, this should reduce the marginal cost of public funds, and so increased G would meet the standard micro cost-benefit test?

    I look forward to your subsequent post, on unconventional monetary policy. Do central banks really ever run out of assets to buy?

    ReplyDelete
    Replies
    1. I'm a bit more liberal in the consumption functions I entertain, so the initial C drop could be a desire to hold more precautionary savings, but I do not think this is important.

      When I said Chris's method was not a helpful way, I did not mean to imply it was wrong. If the cost benefit (CB) is done properly, you should get the same result. However CB analysis normally assumes full employment, which is why I think my way of looking at it is more straightforward.

      Apart from length, what stopped me writing about QE was trying to answer this question: when the central bank buys assets, and changes longer term rates or asset prices for a given path of future short rates, what distortions is it creating? If it is not creating negative distortions, why isn't QE permanent?

      Delete
    2. Simon: "when the central bank buys assets, and changes longer term rates or asset prices for a given path of future short rates, what distortions is it creating?"

      Good question.

      (Weird thing is, just after writing my above comment I was thinking vaguely similar thoughts myself! My initial answer: it's not so much the term-structure that is being changed, but the liquidity-structure (if that's a word). When the central bank buys more and more assets, it is moving along the liquidity spectrum from the more liquid to the less liquid, and compressing the liquidity premium all along the curve.

      "If it is not creating negative distortions, why isn't QE permanent?"

      Initial response: well, Milton Friedman's optimum quantity of money argument, taken to its extreme, says it should be permanent. We should target such a low rate of inflation, and low nominal interest rate, that the central bank's balance sheet would need to be very very large permanently, to prevent deficient aggregate demand. From that perspective, QE is removing a distortion!

      Delete
  8. No capital..you mean no assets, no debt, just income, consumption and production..but no saving/net +ve investment....just taxation that impacts distribution of expenditure via redistribution of income...a stable state with a shock..and some wage rigidity..everything usually spent in every period...

    Where did involuntary employment come from? Are you saying people stopped consuming all their income and have unconsumed income and therefore production adjusted...so it is just a case of getting people to spend...?

    Where does the government expenditure come from? Additional taxation to redistribute unconsumed income?

    ReplyDelete
    Replies
    1. No, government debt and money are allowed, just not physical capital. However I think the points I make would also apply if you allowed for physical capital.

      Delete
  9. "If Ricardian Equivalence held completely tax cuts would be totally ineffective at stimulating consumption"

    blasphemer

    ReplyDelete
  10. One problem with this analysis is that it deals only with the economy as an aggregate. There are really two types of economic actors out there: the vast majority who are living on the margin and a handful of wealthy who have more money than they can spend or invest.

    Interest rates and tax cuts have little or no effect on the former group. They have no assets and no prospects, so they can't borrow from banks. They have little income, even in the aggregate, so even a large tax cut will have little impact on their spending.

    Interest rates an tax cuts could have an effect on the latter group, but only in a robust economy. If there are no investment opportunities because most people are broke and have stagnant incomes, the wealthy will not spend money on goods or services. Instead, they will buy symbolic goods which at the inflated prices caused by the ongoing capital glut.

    This division and its effects are quite real and have been documented extensively. It gibes with everyday experience. (I've used it to make a fair bit of money dealing in symbolic goods which is sort of ridiculous. I actually do have marketable skills, but they don't pay as well.) It has even been field tested and validated as anyone studying our last great depression could tell you.

    I find the fact that what was once common knowledge is now considered worthy of extensive argument as rather fascinating. Has anyone done some serious research on just how the repeatedly discredited phlogiston theory managed to replace the well validated science of chemistry?

    ReplyDelete
  11. I don’t agree with Simon’s advocacy of interest rate adjustments to boost C. Interest rate adjustments are distortionary: they influence PARTICULAR forms of C, that is C based on borrowing and lending, rather than current consumption items.

    Ergo it makes more sense to boost C by helicopter drops, i.e. tax cuts / increasing the state pension & benefits, etc (maybe just temporarily).

    Also I don’t agree that in Simon’s “magic” perfectly functioning free market that it’s primarily interest rate adjustments that boost demand. Indeed, as he says, “it is rarely explained how a lack of aggregate demand gets translated into lower real interest rates.” In a perfect market, one mechanism that boosts demand is the Pigou effect. That’s simply the fact that in a perfect market and given deficient demand, all prices fall, which raises the REAL VALUE of base money (which comes to the same thing as helicopter drops). Though of course the Pigou effect is hopeless in the real world given “sticky downwards” wages.

    ReplyDelete
    Replies
    1. Ralph. I disagree about what you say on distortions: in the basic consumption model it just influences the consumption/saving decision, where saving can be positive or negative.

      I have also written about why the Pigou effect does not work under inflation targeting:

      http://mainlymacro.blogspot.co.uk/2013/08/why-pigou-effect-does-not-get-you-out.html

      Delete
    2. Simon, Thanks for your response. Your argument in the article you refer to, as I understand it, is thus. The Pigou effect raises the value of the stock of base money in real terms, which causes extra demand and more employment and faster inflation. Plus the faster inflation brings lower real interest rates, which solves the problem.

      My answer to that is that problem is solved (i.e. full employment is attained) BEFORE those lower real interest rates have a chance to get going.

      Delete

Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.