The Computationalist Theory of Economics - Introduction

Economic activity is computation.

Let me first note that this view is not new. While googling "Computationalist Theory of Economics" results in zero hits as I'm writing this, computationalism is a fairly common view among researchers in cognitive psychology and computer science. Furthermore, computational models are becoming entrenched, from physics, chemistry and biology to sociology, economics and finance. Thus both the idea that physical and social phenomena can be seen as essentially computational and the more pragmatic notion that they can be modeled by computation are gaining mind share, owing in part to the ubiquity of digital electronic computers.

But noting that research in this area is not particularly well-known in the general public, even those who are economically literate, and in the interest of sharing my thinking without being encumbered by existing literature that I haven't read, I will present things as I see, without much in the way of formal acknowledgement except where I see fit. This isn't to say other works haven't influenced me, but much of this thinking is a result of internal synthesis and a product of the zeitgeist. Many others likely stumbled upon this way of thinking after being exposed to pieces of it in unrelated context. It's not a coincidence that the area is named in different ways with a slightly different focus - complexity economics, agent-based computational economics, cognitive economics, just to name a few. Some economists claim that they have accounted for it all along, whereas others disagree with its usefulness. The taxonomy of knowledge is fuzzier than the rigid structure of attribution gives it credit for. I'll use this fuzziness as an excuse to avoid the mind-numbing tediousness that is proper scholarly attribution.

Before we restrict ourselves to economics, let's review why computation is a useful model for real-world phenomena:

Completeness

Insofar as real world phenomena deal with finite data, finite computation can fully describe all real world phenomena. While there's not as much evidence that reality is finite, reality as we perceive it, to the extent that cognition depends on computational mechanisms, likely is finite. Furthermore, even if cognition isn't finite computation, all concrete communication of cognition is essentially finite. In fact we lack the ability to explicitly reason about that which cannot be finitely described. Thus, we lose nothing of essence by considering all reality as computation.

Reducibility

All computation is reducible to simple models of computation, which allows for both theoretical mathematical analysis as well as practical computational exploration of phenomena.

Equivalence

There are multiple equivalent computational models, to name a few: turing machines, lambda calculus, recurrent neural networks, cellular automata and unrestricted grammars. Each model of computation is universal and is capable of emulating every other model of computation.


What is economy?

Now that we've established the mental framework in which to discuss economics, we can ask: what sort of computational system is an economy? What does it compute and how? In what ways can we map the concepts derived from traditional economics to this new framework?

The basic unit of this computational system is an economic agent, which is also a computational system. Each economic agent then computes bids (which are simply offers made to others to exchange one type of goods and/or services for another in some concrete amounts) based on his ability to produce, his preferences in regards to consumption, bids made by others and other information that can affect any of those. A market, then is a collection of related bids, or an environment in which bids meeting certain criteria are made. Whenever bids are matched transactions occur and the matched bids disappear. A collection of markets, combined with production systems (which can be seen as a special form of market - after all all production is transformation of one for or another and the ability to transform that production mechanisms give you, is a natural bid, based on which artificial bids are made) forms an economy. Note that neither the notion of a market, nor the notion of an economy, is precise. These are concepts.

I think it's fairly intuitive to see that an economy then is a computational system that consists of economic agents who are themselves computational systems, one that is largely self-organizing by way of computation of bids. How this self-organization comes about and the variety of different forms it takes, not to mention how and why specific organizations work are open questions that we'll explore.

The Solution

So I was in bed one night and as I was falling asleep, suddenly this idea for solving the credit crisis came to mind. After a few days of deliberation, here's what I ended up submitting to change.gov:

To whom it may concern,

I'd like to propose a new policy tool that bridges the gap between monetary and fiscal authorities, one that should not only help us deal with the ongoing credit crisis as it stands now, but should work as a calibration tool, much like the fed funds rate, under normal conditions.

We're in the midst of a credit crisis and facing the threat of deflation. Between Keynes, Friedman, Schwartz and even present Fed chairman Bernanke, whether you're a Monetarist or a Keynesian, the academic case for expanding the money supply and making sure that it gets to the economy (for those who care about wonky concepts such as money velocity) is well-established. In fact, I'm sure you're working right now on a fiscal stimulus package and ensuring that it has the desired effect on price stability and employment.

However, as both the popular backlash against and continued calls for various types of bailouts demonstrate, investing a large amount of money on short notice is a huge political challenge. How it gets invested has a large impact on many individuals' lives - there will be winners and losers, no matter the decision and the process invites uncertainty, speculation and rent-seeking behavior. While I have more faith in the team President-Elect Obama has assembled than I had in any administration in the past, the sheer difficulty of the task must not be underestimated. The political difficulty of engineering and implementing a fair solution, not to mention engineering the perception of fairness ensures that the incoming administration will walk a tight rope, despite the general mandate it enjoys. Most importantly, any specific fiscal stimulus is unlikely to become an enduring template for future administrations. The Obama administration may figure out a way to spend the money in a way that enhances America's long-term well-being and competitiveness, but there's no guarantee that future administrations will be as successful. Furthermore, it invites the question: if there's a way to spend money wisely, shouldn't we have already been doing it prior to the crisis? In short, while we all understand that printing money and spending it wisely will end the crisis, the uncertainty inherent in an ad hoc process is essentially what allowed the crisis to persist as long as it has.

That's probably enough introduction now - here's my humble proposal: let every taxpayer borrow from the Federal Reserve member banks (probably indirectly through other institutions; what's important is that the loan end up on the Fed's balance sheet) the amount necessary for them to pay federal taxes, at an interest rate set by the Federal Reserve. I will refer to this program as TLP for Tax Lending Program from here on.

Let's set aside the implementation details (what should the maximum allowable balance be; how should be we determine the initial interest rate; whether the Federal Reserve should issue bonds to fund this; how should it be administered, etc) for time being. To understand why TLP will work, it's necessary to understand why and how traditional tools of monetary policy work. Without getting into great detail, in essence, low target rates are maintained by expanding monetary base, which lowers banks' cost of funds and thus lowers the cost of borrowing for all users of credit. Thus credit and money velocity expand with a corresponding inflationary effect. However, when bank balance sheets are impaired, credit spreads are high and capital market liquidity for loans is low, lowering the risk-free rate does not have a corresponding effect on credit, as illiquid assets on banks' balance sheets have a higher rate of return (relative to their resale value) than new loans they can make at rates borrowers are willing to accept. Furthermore, faced with the possibility of deflation, credit spreads expand for all risky loans because each dollar is worth more in scenarios they default than in scenarios where they don't (this lack of comparability of the value of dollar is partially why credit spreads seem extreme relative to any reasonable probability of default). Thus, we're seeing a drop in credit availability, especially outside of the government subsidized residential mortgage market.

Theoretically, however, monetary stimulus has worked, in the sense that there's now one borrower who has more incentives than ever to borrow: The Unites States Federal Government. In fact, if the US Government was as sensitive to interest rates as private sector entities were, monetary policy would work by making fiscal stimulus seem economical. Unfortunately for Monetarists, the US Government is almost entirely insensitive to interest rates in the short term and is only moderately sensitive in the medium term, for political reasons I've mentioned earlier, as well as bureaucratic overhead. Federal budget cannot easily be changed according to the needs of monetary policy.

TLP largely avoids the shortcomings of existing mechanisms. First, it bypasses the banking intermediaries and supplies credit directly to individuals. Second, it works from the more flexible revenue side and can be fine-tuned without political negotiations. Third, it's less likely than existing mechanisms to lead to asset price distortion. Banks are largely homogenous in terms of the incentives they face and are much more likely to direct the funds towards a single asset class to produce a bubble. Furthermore, with TLP be in place, monetary stimulus will require a smaller adjustment in rates to produce the same effect, reducing any tendency of monetary stimulus to cause malinvestments. Fourth, it leaves no permanent imprint on the budget. Part of everyone's fear regarding fiscal stimulus is the worsening budget picture and the long term threat of inflation it implies. TLP, by temporally moving revenue to a future date, much like other monetary tools, can be used to mitigate deflation in the present without affecting the longer term picture. Furthermore, at a reasonable rate, it should be a small source of revenue.

One objection that could be raised is the idea that the US households are overleveraged and cannot take on additional debt. This concern, aside from being an alarmist half-truth, does not favor direct government spending over TLP. Ultimately, TLP transforms the meaning of money or government liabilities in general. Presently, money is implicitly backed by future tax revenue. Under TLP, money will be backed by past tax receivables as well as future tax revenue. Neither system can function without an underlying economy that supports the level of government spending and the idea that the US government can borrow against future tax revenue, but not against past tax receivables is illogical. If anything, households, being more sensitive to both total debt amounts as well as ongoing debt servicing, are more likely to be able to make sensible adjustments.

While I propose TLP as an enduring policy tool, not merely something that can sustain us through this credit crunch, there's one benefit that is specific to the current situation. Should TLP go into effect, much of the money borrowed should go towards paying off higher-interest rate existing loans, whether mortgage loans or credit card debt. Without considering the direct effect it has on the bottom line of banks, it gives secondary markets a huge amount of information regarding the future performance of those loans. Clearly, one won't take out a loan from the government that will be hard to discharge in a bankruptcy to pay off a more easily discharged loan, if he's considering bankruptcy. Conversely, large loans with high rates that are not being paid off essentially imply that the borrower either has other high interest-rate loans he needs to pay off first, has little income and did not pay much in the way of income tax, or is considering defaulting on the loan, all of which are negative signs. Thus, the introduction of TLP can add one-time clarity in the secondary market for assets backed by consumer loans.

My hope is that TLP will free the US government from being encumbered by concern for short-term monetary policy and allow them to focus on long-term economic issues, whether energy independence, sustainability, global competitiveness, income disparity or budget deficit. All fiscal actions ultimately have monetary implications, but some fiscal actions, without TLP, cannot have its monetary effect temporarily reversed or transferred to a future date by the Federal Reserve. This limitation restricts the government's ability to set its own long term economic agenda and act upon on them. If TLP had already been in place, for instance, I'm fairly certain that we could cut the budget deficit and repeal Bush tax cuts while not having those measures have a large deflationary effect. In fact, TLP offers an intriguing possibility of running a balanced budget or even budget surplus in the long run without facing the possibility of deflation.

It is ultimately with hope that TLP can be administered and used correctly, by the incoming administration, by the Federal Reserve and by the American people that I humbly make this proposal. Should you have any questions, please do not hesitate to contact me. Otherwise, good luck.

Sincerely,

Pricing Uncertainty

What do you all think?