In this essay I propose that the central bank be freed from its role of using interest rate policy to support aggregate demand. Instead, a truly variable public spending program is suggested to regulate aggregate demand. The program should be running constantly in order to minimize the time delay of fiscal responses. The amount of spending is variable and can be automatically computed from the realized nominal GDP so as to target a fixed growth rate for the nominal GDP. In order to gain popular support and avoid the pitfalls of traditional Keynesian stimulus programs, I propose that an electronic national market be set up to give voters direct control over where such stimulus spending is applied.

The recent financial turmoil points to an urgent need for new macroeconomic tools to stabilize the national economy. In general, a certain degree of economic volatility is both unavoidable and desirable; as businesses compete, new technologies develop, and consumer demands shift. But when an excessive debt-to-income ratio is established in the economy, a dramatic downswing of aggregate demand feeds on itself and the economy's ability to correct the condition on its own is limited. While this phenomenon and the need to stimulate the aggregate demand are well understood, recent developments suggest that the policy responses are unsatisfactory. Indeed the talk of a return to the 1930s-style depression is alarming and reflects deep anxiety about the effectiveness of traditional policy responses. In this essay, I first review the criticisms of the traditional macroeconomic policy tools and then propose a new way for regulating aggregate demand and stabilizing the national economy.

In recent decades more emphasis has been placed on the monetary policy conducted by the central bank and transmitted through the financial markets to regulate the macroeconomy. The central bank's stated goal is to ensure price stability. In addition, the Federal Reserve is also tasked with formulating policy that is compatible with full employment. The central bank's policy tool is the supply of money. While it mostly controls money supply through the setting of interest rates, sometimes quantitative control of money supply has also been tried to uncertain effect.

The central target of price stability is more art than science. As the mix of goods and services constantly shifts, based on supply and demand, the constitution of a reference basket for price comparison is a problematic task in itself. While a degree of price stability is necessary for rational transactions based on money and for planning at the microeconomic level, there are many problems with targeting the composite price level itself:

1. Technological change can have dramatic effects on the mix of products and services. For example, measured by computing power alone, the price of computers has deflated greatly, yet it is hard to take this into account in terms of impact on the price index.
2. Large spikes in the price of non-discretionary items such as oil or food could induce a large drop in demand for discretionary items in the short term and therefore their long-term effect on the general price level is course-dependent on the economy.
3. Individual price volatility is necessary as a market signal for supply and demand. Such volatility can be dramatic yet completely normal for the functioning of the economy. On the other hand, even if the composite price level is stable, individual firms may still need to cope with volatile input or output pricing.
4. Price deflation combined with robust demand could be a sign of economic health and rising standards of living instead of a cause of alarm.
5. Foreign exchange rate volatility is another source of fluctuation that is outside the control of central banks supporting floating currencies.

Therefore, the justification for targeting price level is more based on the theory of "sound" money — that is, the trust in the unit of transaction, even though price fluctuation is both normal and necessary for efficient allocation of resources. In the face of practical difficulties in regulating real prices central banks have shifted towards an emphasis on the importance of "inflation expectation," a necessarily subjective and psychological matter. And for this they increasingly look to the financial markets, especially the government bonds market, for guidance. Like other markets, however, bond pricing is also subject to supply and demand. In the US, the Treasury debts are very much in demand by other central banks as instruments of foreign exchange reserves. Bond market trading is also heavily influenced by speculation on the future moves of the central bank, therefore it could cause circular logic and reinforce perceived central bank intentions instead of transmitting signals from the real economy. While it is undeniable that monetary policy has been effective in maintaining long-term price stability, it is possible that directly targeting the price level could also be the cause of short-term economic instabilities:

1. As in any control loop, delay could skew the intended effect of intervention. Since the effect of interest rate policy on aggregate demand is indirect, there is significant delay. And if the timing of a policy change turns out to be inappropriate the effect would likely be destabilizing.
2. Interest rates have a more direct effect on the pricing of capital assets than on aggregate demand. When aggregate demand is affected through asset prices, boom and bust cycles become inevitable. In addition, such cycles could become progressively amplified as policy responses to ameliorate a bust lead to a bigger new boom.
3. To the degree such a policy succeeds in the short term, it encourages more speculation in asset prices and is self-defeating in the long run.
4. As the central bank conducts its policy through the financial markets, the financial sector becomes hyper-developed as the central bank's role in macroeconomic regulation becomes progressively accentuated. Large amounts of resources, especially human talents, are dedicated to developing financial instruments. While the credit market plays a critical role in resource allocation for the real economy, the financial intermediations start to dominate the real economy with confused incentives. Excessive speculation built on financial leverage does not lead to a healthy accumulation of physical and human capital in the real economy -- instead the incentive is for short-term fictitious profits that overly benefit the intermediation agents and leads to gross misallocation of resources in the real economy.

And, finally (but imminently relevant to our current crisis), there is a limit to the effectiveness of monetary policy alone in generating aggregate demand in times of great uncertainty. As has been amply explained by Keynes, private actors in the economy can be driven by excessive fear and low interest rates alone may not be able to sustain credit creation. Viewed alternatively, private actors may take actions that are individually rational, based on their projection of how other actors may be reasonably expected to behave, but taken in aggregate such actions could drive the economy far below its productive potential, in other words, creating a depression.

Managing aggregate demand through public sector spending is much more direct and easy to understand. While Keynes had long ago established the need for fiscal policy response to prevent financial crises, active state management of aggregate demand has fallen out of favor recently because of the practical difficulties in implementing effective Keynesian stimuli.

Direct government spending is necessarily a political decision and governments must decide how much to spend and which sectors should benefit from the stimulus. Even if such decisions can be arrived at rationally, the delay involved in arriving at a decision may well mean that the response becomes unnecessary for normal inventory-cycle-driven recessions yet may well be too late or insufficient to address a true financial crisis to minimize damage to the real economy.

While it is easy to convince people of the need for additional government spending in times of recession, withdrawing such spending during normal times is hard. Since only changes in public spending have the desired effects -- progressively applying more and more government spending becomes the political reality if short-term stimuli cannot be withdrawn and become institutionalized. Over time this leads to an over-sized public sector to the detriment of private sector capital formation and long-term growth of the economy.

Instead of increased spending by the government, tax cuts are another way to sustain aggregate demand by stimulating private demand. However here again it is a political process to decide who benefits from tax cuts and implementation is necessarily time consuming. It is also unclear how efficiently an across-the-board tax reduction would translate into increased demand since the propensity to spend irregular tax incentives varies among taxpayers. In developed economies that run large trade deficits, there is an additional concern that consumer spending has significant import leakage and does not have the maximum GDP multiplier effect.

So while the effect of public sector spending on aggregate demand is not in doubt, its long-term effect on the economy has been called into question because of the checkered record of increasing state management of the economy. In addition, many social safety nets have been established since the Great Depression, and public spending in proportion to the total economy has increased greatly, which by itself has a stabilizing effect on total demand, therefore the desire for further management is reduced. However, I would argue that abandoning a standing fiscal control to stabilize aggregate demand because of past failures is like throwing out the baby with the bathwater, especially since the monetary policy tool in its stead is problematic, as we discussed earlier.

While the need for maintaining aggregate demand is not in dispute, the policy action to best achieve such effect is, especially in light of our current policy failures. However a great financial crisis also gives us a chance to rethink policy choices in a fundamental way.

First, let us agree on the following principles:

1. Financial systems and private actions should remain the predominant actors in allocation of resources. Policy action should augment and stabilize aggregate private demand but not replace it to any great degree.
2. The main goal of any policy response is to rationalize private sector decisions away from the destructive kind of self-interest.
3. Any response needs to be automatically actionable so as to establish the desired expectation in the private sector.
4. Such policy should be compatible with both long-term price stability and full employment.

I would argue that in light of our requirements the obvious target to regulate in order to stabilize the macro economy is the nominal GDP:

1. Nominal GDP is the most straightforward tally of aggregate economic activity. What we want to stabilize is the nominal national income, to which GDP should be equal. While private sector GDP cannot be directly controlled, total GDP is controllable through a variable public spending component. A regulated nominal GDP does not in itself presume and constrain how resources should be allocated among economic sectors, nor does it constrain choices between consumption and investment.
2. Assuming long-term price stability and full employment, trend growth rate of nominal GDP should be equal to the sum of the growth rate of the working population and productivity. Therefore a numerical target can be set that is compatible with both price stability and full employment, within a margin of error. A stably growing nominal GDP is therefore both clear as a target and compatible with long-term policy goals.
3. Debt-to-income ratio is the determining factor in extending credit. During economic downturns uncertainty about solvency of borrowers could push private demands into a downward spiral (the fear factor). Stabilizing nominal GDP means stabilizing aggregate income and should lead to rational financial projections based on a known macroeconomic condition. The central bank is freed from managing the aggregate demand, especially from the bottom, and can focus on regulating the health of the banking system to achieve desirable debt levels with a stably growing nominal national income in mind. Ideally the average short-term risk-free real interest rate should be close to zero, while the long-term rate should be set by balancing the desire for current consumption with the need for investment.
4. Real demand and real GDP should be allowed to fluctuate. Real demand must respond to real supply. Since we need the market to function in order to maximize efficiency, controlling real demand is undesirable. For example, in times of supply bottlenecks of oil, we should let the real economy adjust to the change through market pricing and resultant adjustments in real demand for oil. Controlling real demand during such times is both undesirable and unachievable. Another possibility is that the economy may enjoy above-trend productivity growth for an extended period of time, and during such time real GDP may grow faster than nominal GDP and there may be some price deflation. However since it is combined with stably growing aggregate demand, price deflation under such conditions should not be a cause for financial instability. Indeed the national economy may well be enjoying very robust health during such periods with substantial lifts in standards of living.
5. Despite short-term fluctuations of the real economy, in the long term it will grow like the nominal GDP with the difference in growth rates being the rate of inflation. Assuming financial markets function rationally when the growth in aggregate demand is not in doubt, full employment should be the rational outcome except during transitional periods when there are large adjustments to changes in private sector demand and supply, such as the extremes of inventory cycles or an oil shock or a large technological breakthrough. Assuming full employment is achieved and therefore real GDP is growing at least as fast as population growth, the only cause for long-term inflation would be extended periods of subpar productivity growth. However if we only make moderate assumptions about productivity growth (about 1-2%) for the developed economies the deviation in price level should be well tolerated and compatible with stable inflation expectations.

It is possible that withdrawing the variable public spending alone may not be sufficient to counter a bout of high nominal GDP growth since there is a floor to such spending (it cannot be negative unless further spending cuts or tax raises are taken, which is politically difficult or time consuming to implement). However the traditional monetary policy tool is still available to prevent inflation from spiraling out of control. The fact that the central bank is relieved from supporting aggregate demand should give it much more freedom to conduct policy and regulate the financial system.

It is not hard to convince economists of the need to support aggregate demand to prevent economies from staying below their long-term potential for an extended period of time. Generating variable demand through the public sector to compensate for short-term private sector shortfalls is the most direct response. Supporting nominal GDP growth at a near constant rate is a natural outcome of such policies. The reason such policies are not implemented lies in the difficulty in implementing truly variable and responsive fiscal stimuli. Instead central banks have been relied upon to regulate aggregate demand through monetary policies. However, stimulating demand through low interest rates during economic downturns leads to asset price bubbles, which cause financial system instabilities.

A critical goal of public spending to support aggregate demand is to influence private participants (lenders and borrowers, producers and consumers) to make rational decisions. To achieve this it must be fully established in the expectation of private players. Therefore such spending must be automatic and responsive to private sector fluctuations. Naturally there is the question of what to spend the monies on and how much to spend. A prolonged policy debate on each spending item could be both inefficient and ineffective (by taking away the certainty from private expectations).

There must also be accountability to tax payers and the citizenry in general -- to reduce perceived waste, to respond to voter priorities and to address anxieties about long-term fiscal impacts.

The concerns about fiscal stimuli are both real and valid. They must be addressed for any fiscal policy tool to succeed. Fortunately new technologies are creating new possibilities to conduct such a policy successfully. In particular, Internet-based commerce has proven amazingly effective at creating a truly national electronic marketplace to match demand and supply among millions of participants. Financial security trading and auction markets like EBay are just two of the most successful examples. A national electronic marketplace can also provide a platform to give the citizenry unprecedented control over public spending. Therefore I propose that:

1. A fiscal policy for a variable stimulus to the economy should be established through the force of law. The aggregate amount is automatically computed from the difference between the realized nominal GDP and the expected nominal growth. The national Treasury should be authorized to automatically borrow or draw from a reserve fund (these actions are outside of the regular budget) to create the necessary amount in an aggregate spending account.
2. An electronic national marketplace would be established, whereby qualified entities would supply proposals for projects that require funding. Such entities would be vetted for their ability to carry out their projects as proposed if funding is provided, although such vetting is not critical since the marketplace itself could provide effective policing over time. It may be necessary to vet the proposals in terms of speediness and the multiplier effect of the spending.
3. Each taxpayer or voter is given a pro rata share of the aggregate Treasury spending account to "spend" in the electronic marketplace. One can choose to contribute one's share to any number of spending proposals. He or she should be able to access new proposals, review prior year results and the qualifications of proposers. Development of online communities can be facilitated to encourage exchange of information and ideas. To remove the effect of voter apathy, unused shares would expire after a certain amount of time and be reallocated to active voters on the marketplace.

The technological feasibility of such a system is beyond doubt. The direct control provided to the taxpayers or voters in general would provide the ultimate accountability. Over time many public and private entities could establish their reputations and competitiveness in winning bids for project funding. Certain active voters could also establish their reputations and considerable influence over others. In other words, the market will evolve and mature and become the best possible way to match supply and demand within the legal parameters of variable public spending. The advantage of giving voters control over a variable budget instead of tax rebates is that there is no doubt that the propensity to spend "other people's money" is much higher than to spend a tax rebate. Also, unlike tax rebates, the voting credit can be easily reallocated to ensure that the spending target is hit.

While the amount of the spending is unknown ahead of time, the response is swift and can also be continuous. That is, the aggregate spending account can be constantly recharged at varying rates in response to private sector conditions. During good years the variable spending could be minimal since it is formulaically determined, therefore it is truly counter-cyclical.

People familiar with the concept of control engineering should recognize that the proposed mechanism is a basic form of the feedback control loop. Its effectiveness can therefore be investigated from several aspects:

1. The critical determinant that the loop will function stably is the time delay in applying the stimulus with respect to the rate of fluctuation in private sector activities. Any change in spending will necessarily take some time, even in the scheme proposed here. One could argue this point in two ways: there are already some built-in mechanisms such as unemployment insurance that naturally dampens the fluctuation in private demand; and secondly it is hard to see how one can do any better than the proposed standby responses.
2. Since the variable spending cannot go negative, it is biased. The variable response, within the limits set, should be equal to A+c(T-G), where A is the bias (the average stimulus for a "normal" year), T is the GDP target, and G is the measured nominal GDP. The multiplicative factor c determines the strength of the response. It does not have to be 1, though it should probably be no less than the inverse of the estimated GDP multiplier if convergence is desired. The larger it is, the more sensitive the control loop. Unless the time delay is negligible, too strong a response (overcompensation) could cause instability. In control loop lingo, the bandwidth of the loop (larger bandwidth means more robust control) is jointly determined by the magnitude and the time delay of the stimuli.
3. The loop performance can also be evaluated through the tracking errors. There are two kinds of errors to consider:
   1. If the upper limit of the variable response is set too low it is possible that the nominal growth target is never achieved and the response will be stuck at its allowed maximum. If such is indeed the result even with fairly large allowed response with respect to the GDP, deeper structural problems in the economy are implicated: perhaps a very large disparity in income distribution depresses private demand even in the presence of a stimulus; or normally budgeted public spending may be too low with respect to taxation to be overcome with the variable response (viewed alternatively taxes may be too high with respect to public spending).
   2. On the other hand, even if an average nominal growth rate can be successfully achieved there may still be fluctuations in short-term realized nominal GDP. However, as long as the loop is stable (the loop remains a negative feedback loop) the regulated fluctuation should be smaller than the natural ones. There are also natural fluctuations in the nominal GDP that should not be causes for alarm, e.g. fluctuation due to timing of large orders of capital goods such as commercial airplanes, yet a formula-based response may be generated. Due to the nature of the closed-loop control, any spurious response will self correct over time (over stimulation in one period will naturally cause stimulus to recede in the next). If spurious response is undesirable, the input (measured realized GDP) may need to be smoothed at the price of reduced sensitivity.

A detailed analysis of the control loop dynamics is necessarily highly technical but can be readily achieved through standard techniques.

It is possible to adjust the control parameters over time. This allows introduction of the proposed policy in phases so that initially only a portion of private-sector fluctuation is compensated for (either parameter c above is set to be significantly less than 1, or the limit to the response is set to a low value, or both). Even a partial policy response will have a desirable stabilizing effect. Phasing could allow time to learn and experiment. Trade-offs like these being discussed will be necessary to actually implement the control mechanism to best match expectations.

Details of control loop dynamics aside, the importance of establishing private sector expectation should be emphasized. Success is achievable so long as there is sufficient and justifiable confidence in its working. The beauty of a closed feedback control loop is such that despite errors being generated in the control process (such as errors in measurement of GDP, estimation of voter participation rate, estimation of magnitude or time delay of stimulus) the system will over time compensate and work as intended. That is, the proposed mechanism is robust so long as the stability criteria is met.

The long-term fiscal impact of a variable spending budget could be offset by subtracting the mean from the regular budget. Reducing the regular budget might be politically difficult. However, if the regular budget is simply held below GDP growth for some years after the variable spending is introduced, the long-term impact of the variable budget to the national debt could be neutralized. Of course the fiscal impact could also be managed through a reserve fund, though the discipline required might be hard to maintain.

Since voters have direct control over how to spend the variable budget there need not be undue constraints on what spending proposals are eligible for listing in the marketplace. For example, a proposal for personal elective surgery could be perfectly acceptable if a few voters could be interested to fund it. On the other hand, the stimulative effect of proposals may not be high on the priority list of voters — purchasing imported electronics would be much less effective than purchasing domestically provided services — so it may be necessary for the listing agency to review and act as a gatekeeper on this. However such reviews can be done simply most of the time and when necessary (for very large spending proposals that have won sufficient support, for example) can be done more extensively.

Local governments are natural candidates to propose projects for funding. Because of their high propensity to spend their marginal income they should be allowed to list currently budgeted and ongoing projects. This will provide a large pool of well conceived projects for voters to choose from with minimum delay to the additional spending. Funding provided to already budgeted projects would free up funds to support other local government spending. This is especially true in the US: most states are constitutionally bound to not run operational deficits, yet many states depend on progressive income taxes that can be very volatile. Providing assistance during a downturn is therefore counter-cyclical and stabilizing.

It is debatable whether taxpayers alone or the entire citizenry should get the right to direct spending in the proposed marketplace. Since the public debt is the burden of all citizens, it may be constitutionally required to give all citizens equal rights — subject to expiration due to voluntary non-usage.

It may be more efficient to rotate voting rights among citizens instead of always giving them pro rata shares of the spending at the same time. It is much more likely that a voter will exercise due diligence in spending $1000 rather than $10. By rotating the rights, each voter would have more spending power and more incentive to properly investigate the spending proposals.

It is not clear what proportion of voters will choose to actively participate at any given time. Even if only a fraction of eligible voters participate (and those who do will also wield more influence over the spending due to reallocation of expired rights), the process will still give the proposed mechanism more legitimacy than alternatives. Indeed, each sizable spending proposal will have to be scrutinized and approved by many more eyes than any legislative budget item has ever encountered.

It is debatable whether the voting rights should be tradable. If allowed, they will likely trade at a fairly steep discount to the voting power they represent. Over time such practices have the potential to distance voters from their sense of direct responsibility for national spending decisions and deprive the policy of the necessary political support to survive. So I would argue for disallowing trading of rights even if a certain amount of vote buying may be unavoidable. Voting rights transfers to trusted proxies (friends and relatives), on the other hand, may be conducive to giving the widest possible range of voters a sense of participation, and could be allowed if not explicitly condoned.

A desirable complement to such a policy would be the elimination of "pork barrel" spending. Pet projects favored by individual legislators would now be directly proposed to voters in the marketplace so their existence as earmarks in regular budgets would no longer be justified.

The proposed policy response is equally applicable in non-democratic economies, such as China. Indeed China may have the most to gain from a policy tool like this since its consumer spending propensity is lower and saving preference stronger. The marketplace concept proposed here could prove acceptable to the ruling party as a starting point for the populace to learn about and exercise democratic powers in a most concrete manner yet not be directly threatening to its current political structure.

While the need to stabilize aggregate demand is well known, the right policy tool to achieve the goal has so far been elusive. Here I proposed a direct regulation of the nominal GDP with a variable public spending account controlled directly by voters. Only recently has technology become available for such a policy response to be practical. Such a policy has the potential to directly affect aggregate demand and stabilize the economy, while at the same time give the citizens a direct say and therefore a sense of responsibility in the conduct of national affairs. In the long run such a policy may also prove optimal, given the constraints, in resource allocation because it is conducted in a public marketplace.