“tax changes have very large effects on output. Our baseline specification suggests that an exogenous tax increase of one percent of GDP lowers real GDP by roughly three percent. Our many robustness checks for the most part point to a slightly smaller decline, but one that is still well over two percent…”
In the above statement, the ‘exogenous’ in bold is of utmost importance in explaining the reason for why there is such a stark constrast in multipliers. The problem in the model is that the tax cut occurs in a non-systematic, random point, with no reference to the business cycle. This means that
i) we can attribute any output effects solely to the tax cut.
ii) Expectations are more unlikely to adjust, that is, people are unlikely to behave in a ricardian fashion and to save the entire reduction in taxes
The problem of course, is that the current stimulus package can be seen as endogenous, in that it is a direct response to insufficient demand, and is a ‘countercylical’ response. Mankiw explains this as a tax cut designed to offset the demand dampening effects of an event X.
We know that the unfolding of event X on aggregate demand is incomplete, and that any output effects due to the tax cut are likely to be mitigated by the output effect of event X. An event X that comes to mind is the credit crunch, which is leading to a ‘flight to quality’, with excess holdings of money and government debt, and a corresponding rise in precautionary savings. Thus, given that the credit crunch is leading to a rise in the saving rate, it is likely that a tax cut is going to be largely saved (In Macroblog, it notes that the saving rate has gone from –0.1% in 06 to 2.8% in November 08). In any case, the ongoing dampening effects of the financial crisis and the aftermath of current lowering of production in response to lowering demand is likely to offset the direct output effect from the tax cut.
It is interesting to note that in Romer’s paper she makes a reference to tax-based investment incentives,
“For tax-based investment incentives, we used the rule of thumb that the output effects correspond to one-fourth of the effects of an increase in government spending with the same immediate revenue effects. This implies a fairly small effect from a given short-term revenue cost of the incentives. But, because much of the lost revenue is recovered in the long run, it implies a fairly substantial short-run impact for a given long-run revenue loss…"
This makes it clear that Romer believes that dynamic scoring means that ‘bang for buck’, the amount of short-term stimulus added per dollar of debt- is effectively high. In other words, they believe that tax cuts are more effective at recovering revenue in the long-term, and so in light of the effective dollars spent (that is not financed via recovered revenue), the short-run effects seem worthwhile.
By ‘same immediate revenue effects’ I assume that Romer believes that the static deficit increase is the same, whether it be through an increase in G or a reduction in T. Why, according to Romer, would the incentive effects of investment tax-cuts lead to a smaller long-term revenue cost than government spending? In other words, she believes that over time, dynamic estimates of the deficit will be lower for the reduction in T, compared to an equivalent increase in G. This is in stark contrast to Krugman’s estimate in Dynamic Scoring, where he shows clearly that the ‘bang for buck’, the amount of stimulus per dollar of future deficit, is much larger for Government spending than for tax cuts. The difference is of course due to Krugman’s neglect of incentive effects of tax cuts. There is little doubt that investment tax cuts, such as a cut in capital gains tax, company tax, tax rebates for innovation, a cut in payroll tax, do provide some impetus for capital investment, or perhaps greater employment. In fact, Mankiw, in discussion of why the Romers estimate a tax multiplier can reach a maximum of 3, says
“…tax cuts produce a bigger boost in investment demand. This might work through changing relative prices in a direction favourable to capital investment--a mechanism absent in the textbook Keynesian model. Suppose, for example, that tax cuts are not lump-sum but instead take the form of cuts in payroll taxes (as suggested by Bils and Klenow). This tax cut would reduce the cost of labour and, if labour and capital are complements, increase the demand for capital goods. Thus, the tax cut stimulates demand not only by increasing disposable income and consumption spending (the textbook Keynesian channel) but also by incentivizing more investment spending.” (Greg Mankiw)
Mankiw’s opinion of current is to provide a permanent reduction in the payroll tax coupled with a gradual, permanent increase in the gasoline tax. This will equal each other in net present value, so in other words, it ties the budget to be neutral over the course of the business cycle. The idea of payroll taxes, according to Bils and Klenow, has 3 benefits:
1) Like previous stimulus efforts, it has the standard demand side impact (same as cutting checks). But it also stimulates employment directly by reducing the tax penalties for working and for hiring workers. Related, it works under all business cycle models (even including those obeying Ricardian Equivalence).
(2) It targets domestic production better than sending out checks (or a sales tax cut).
(3) It targets lower income households, due to the cap on social security taxes. These households may respond more in both their consumption and employment decisions.
Thus it is obvious that there are many factors behind tax cuts that are not included in simple Keynesian theory. It has supply-side effects, as it targets domestic production via lowering costs of production, and it increases the incentive to raise employment, and capital. Plus, given that payroll taxes are paid by all employers, low and high-income, it will attempt to reduce a proportional tax, thus relieving a greater burden for low-income employers and increasing income equality. Such incentive effects are long-term, and will not really affect the direct 1st round output effects. Given the fact that tax-cuts are more effective in the long-run, and that incentive effects are very hard to analyse using past examples (due to the endogenity problem, that is, it is hard to isolate incentive effects of increased output and revenue recovery from other events).
Thus there is ample reason to believe that tax-cuts are not very stimulating, and in many cases will likely be saved via Ricardian Equivalence measures. In fact, Ricardian Equivalence tends to favour government spending rather than tax cuts, with an emphasis on whether they be temporary or permanent. Though there is no doubt that any tax cut, to have effect, must be permanent, it is impractical to make tax-cuts permanent because, once the economy returns to normal, you cannot renege on the policy and increase those taxes again to be budget neutral. On the other hand, one can argue that Government spending may not be temporary, as many projects may require ongoing funding commitments, and certain political lobby groups may force the Government to continue subsidising their programs. Thus there is a chance that, on either sides of the G vs T debate, there may be a permanent increase in Government involvement, which makes it harder for fiscal policy to be budget neutral in the long-run.
Before I go through a couple more arguments, I would like to surmise my findings above. I have stated that there are real problems in determining multipliers, especially in relation to government spending multipliers vs tax cuts. Simple Keynesian theory is blurred by theories of incentive effects of tax cuts, which tend to arbitarily increase tax multipliers. Thus there are numerous problems in estimating the reaction of households to deficit financed tax cuts/spending increases, and it is this problem of estimation that is causing much angst among economic commentators at present. The only consensus in the multiplier debate is that multipliers are always greater in a less-than full employed economy.
Another argument proposed by Gauti Eggertsson is that tax-cuts may in fact have negative multiplier effects at the 0 lower bound! The reason is simple, Gauti believes that at the 0 lower bound, any supply-side effects will lead to deflationary pressure. This deflationary pressure cannot be counteracted by lower nominal interest rates- as interest rates are fixed at the 0 lower bound. The economy plunges into a deflationary spiral, increasing the output gap, mitigating any output effects from the tax cut. The effect of aggregate supply was covered in my previous post on supply shocks, and I talked about how ‘vice becomes virtue’, that policies to reduce aggregate supply rather than increase it, can be more helpful in preventing a deflationary spiral.
Lastly, I would like to reflect, again, on the endogenity problem of tax cuts. There is an intuitive argument, mentioned previously, that the endogenous event at current is a natural increase in ‘precautionary savings’ and a system-wide deleveraging effect. The need to pay off debt may lead many households to save much of their tax cuts, especially if they believe that the cuts are only a temporary increase in disposable income (thus leading to ricardian effects). This provides an excellent reason for why Romer’s exogenous tax change study lead to a tax multiplier of 3, but the Romer-Bernstein analysis of the Obama package gives a tax multiplier approaching 1. However, Mankiw expresses some distaste for this idea, as he distinguishes between the idea of increases in precuationary savings leading to an increase in the APC (Average propensity to Consume, with differing implications for the MPC (Marginal propensity to consume). Mankiw’s idea stems from his model of the consumption function. His hypothesis is that consumption will be depressed more in lower income groups, a fair assumption as they tend to have larger debt to income ratios, and are at more risk of losing a job (with less insurance as well!).
The above graph is very simplified, but it shows clearly that there is an autonomous fall in consumption, but with an emphasis on lower income groups. Thus, a tax cut, which enables a rise in disposable income, will, on a marginal dollar, be consumed more by a low-income household than a high-income household. This provides a nice rebuttal to the ‘precautionary savings’ fallacy, which implies tax cuts completely ineffective.There are numerous other arguments which claim that G > T, or T > G, but I can no longer dwell on them here. I tend to favour government spending multipliers being greater, simply because most Macroeconomic models predict this result, and Permanent income, Ricardian Equivalence models predict the same result. I also feel that tax cuts are more effective in a period of full employment, but with under-employment, incentive effects may not have the desired effect, especially due to the fact that investment is constrained in a recession. In fact, I am of the belief that the greater output effects of government spending are more likely to kick-start investment than tax cuts.
Thus fiscal stimulus is necessary…there is no question! But the structure of the fiscal stimulus, whether it is via tax cuts or government spending, is still open to debate. My feeling is that this question will be resolved using this crisis as a precedent, and a consensus will be reached on the efficacy of fiscal and monetary policy in depression-like conditions. No doubt such a consensus will be a hallmark in the history of economic thought, and I for one hope it will bring together a unity resembling the harder sciences.
No comments:
Post a Comment