Entrepreneurial Experiments -- Research and Questions

This post is to share some original research I have in process with readers of this blog. For the last few months, I've been preparing to conduct some new economic experiments, and would love to engage in a discussion with you & get your advice. Dan Houser, a professor at George Mason University, and a graduate student, Xiaofei Pan, are helping me prepare to conduct the experiments at the GMU ICES labs, hopefully this summer.

The experiments are for the study of work and entrepreneurial behavior under different policy regimes, including taxation, welfare, and social insurance. I'll share a link to the preliminary literature review / experimental design working paper in a future post. Questions on my mind are:

1. Will higher federal taxes reduce entrepreneurship and growth, or are those fears overblown?

2. What's the best way to design an experiment for work under different policy regimes? Once we nail that down, what is the best way to design the choice of riskless versus risky (entrepreneurial) work?

3. Is there research already in existence informing this topic?

I was inspired by Vernon Smith (wikipedia entry here) after I spoke with him at Chapman University during last November's Global Entrepreneurship Week (hat tip to the Leatherby Center there). I was surprised to learn that experimental economics hasn't pushed very deeply into growth policy questions. According to a 2003 paper by Sutter and Weck-Hanemann (2003), “Experimental studies on issues of taxation abound.” However, these generally explore tax evasion, whereas “studies on the effects of tax laws on labor supply are very rare." Tax evasion is really just a subset of income shifting in response to taxes, related to the subset that naturally interests the Kauffman Foundation, which is the degree to which higher taxes are a barrier to entrepreneurial risk-taking.

The study of labor supply's elasticity has evolved in recent years. A consensus exists, holding that the number hours worked (i.e. labor supply) is inelastic with respect to tax rates, especially for males (primary earners). However, Larry Lindsey suggested in a 1987 paper that the key elasticity is not labor per se, but income. Martin Feldstein explored the idea of "elasticity of taxable income" (ETI) thoroughly in the late 1990s. Clearly, this is where my interest in entrepreneurship comes in. If higher taxes act as a job-lock by keeping more workers tied to their riskless salaried job (with untaxed fringe benefits) instead of a risky startup, then there is a short-term loss, measurable in terms of deadweight loss, and also a long-term loss, the immeasurable reduction in potentially innovative activity.

So what does the ETI literature tell us about entrepreneurship? Glenn Hubbard, Austan Goolsbee, and Doug Holtz-Eakin have written on this, including chapters in the 2000 tome, Does Atlas Shrug?, edited by Joel Slemrod. Goolsbee describes the "new tax responsiveness" (NTR) literature in terms of a methodology using econometric analysis of empirical data from natural experiments. The term confused me when I first heard it, because these are not really experiments in the proper usage of the term. But economists are able to assess activity before and after an unanticipated (exogenous) event such as a major tax reform, they call it a natural experiment. In any case, the early estimates of ETI were above 1, meaning that income was highly sensitive to taxation to the extent that an X percent increase in the marginal tax rate led to an equal or greater than X percent reduction in taxable income.

There may be more detailed papers about ETI and entrepreneurship, but the most useful survey of ETI I have found is a recent chapter (ungated here) by Seth Giertz in the 2009 AEI book Tax Policy: Lessons from the 2000s. Giertz reports that "recent studies report [ETI] estimates closer to 0.4." Estimates are narrowing, while at the same time we are learning that the ETI is not constant over time nor across earners. It seems the ETI rises with income itself -- the rich are so sensitive. Some studies that are able to focus on higher-income earners find estimated ETI slightly over 1.

Nonetheless, the NTR literature is limited to empirical econometrics, with all the pros and cons it entails. Giertz identifies a handful of issues that complicate ETI estimation: exogenous shifts in income distribution, mean reversion, endogeneity of the tax rate, institution factors in the tax system (shifting base, enforcement), transitory responses and income shifting, and transfers between economic agents. I know enough econometrics to worry that the empirical ETI debate is unlikely to be resolved since the results will be so sensitive to the inclusion or exclusion of other variables. It is worth noting that Giertz' paper makes no mention of insights made by experimental econ. But it seems to me that experiments can isolate ETI behavior and control for all the noise that muddies the signal in empirical work. It sure can't hurt.

Experiments and Taxes. A pioneering study was published by C.W. Swenson in 1988, updated and improved by M.A. Sillamaa in 1999, which confirmed Lindbeck’s theorem that higher marginal labor taxes will depress work effort even when all income is redistributed. Sillamaa reports that a marginal labor tax rate rising from 12 to 87 percent (across five discrete rates) - with revenue fully redistributed as a demogrant - is associated with a decline in total work intensity of 21 participants, from 7734 tasks to 6824. But by focusing narrowly on the Lindbeck theorem, this research sheds little light on richer, and more realistic, aspects of fiscal policy. Full revenue redistribution, in particular, is a very strong assumption.

And yet, the Sillamaa study establishes a firm foundation for future research using an experimental framework that includes (1) tedious work, (2) real money payments based directly on work effort, and (3) leisure opportunities as a valid participant activity. This framework could be extended to richer policy environments which would yield valuable insights into work behavior.

More recent experimental studies have explored the Laffer curve (Sutter and Weck-Hanemann 2003; Levy-Garboua, Mascalet, and Montmarquette 2008). But these studies use a 2-player game design where one player is the worker and the other is the tax authority. The results inform psychological motivations and rules of fairness, showing for example that a “veil of ignorance” about which player will fill role A or B leads to a greater sense of work obligation and a subsequently lower work elasticity. These papers treat tax revenue as a total loss to the worker, an assumption of zero redistribution. This is also a very strong assumption, but in the extreme opposite direction of the earlier papers.

The experimental framework I am planning on uses elements from both of these branches. The innovation I have in mind is structuring games so that all players are worker-taxpayers facing an identical environment, a faceless governing authority that can set taxes and welfare policies exogenously. A big question on my mind is whether work effort will change when the redistribution rate changes. If we redistribute a portion of tax revenues equally (so that all players get the same split), then I assume the free rider problem will loom larger as the rate of redistribution rises from 0 to 100 percent of tax revenues. But maybe social pressure in small groups will cause work to intensify with high redistribution rates.

Well, that's a pretty long preview. I'm excited about the research, and really happy to use the blog to share it with you. My hope is that experiments will offer a clean room approach to identifying ETI, which the empirical approach will have difficulty resolving alone. I can only imagine one topic of greater importance for economics, which is the precarious state of the global financial system (though I'm confident Bob Litan has that covered). The decades ahead will see developed countries struggling with the limits of taxation (here, here), and the more light we can shed on that topic, the better.

Comments

[John Bailey]

Conducting Entrepreneurial Experiments, http://www.growthology.org/growthology/2009/05/entrepreneurial-experiments.html, is very exciting as well as very important.

In thinking about it over the last few days, I have come up with some suggestions. The first would be to consider developing a framework that could be deployed/used over the Internet so that other experimenters could use it to test new hypotheses and/or replicate your experiments.

I think that the focus on effort/risk/reward is a very good idea. In thinking about how to implement an experiment, I have some suggestions. Participants could “play” one or more games by which they could earn real rewards for some games (work) while playing others for fun (leisure). Ideally the games would be designed so that success at the game(s) grows over time. The games could vary from relatively boring “work” to more exciting (Samuel Adams Beer Ad, “Do what you like and you’ll never work a day in your life”). The participants could choose a “no lose” option (work), a “risk” option (entrepreneur), an “investment” option (investor). Investors could pay with their own earnings or borrow, with the borrowings due back at the end of the experiment. Those choosing the “risk” and “investment” options could lose but they could also gain value for their “business”. Participants would have to pay to play the game(s) (tax). Of course, the taxes could be pure costs or redistributed in a variety of ways. Specific games could be terminated (break) (job loss) with those players receiving (or not) payments while their games were “fixed.” Players with broken games could have the option of other games at which their rewards would likely be less. Some games could be “fixed” while others would not return. Players could be isolated or aware of the actions/results of others.

One alternative would for the rewards to be U.S. currency. Another would be to earn points that could then be turned towards the purchase of real items. The latter would allow for an inflation/deflation effect as the “price” of items could vary during the experiment.

Another area of experimentation that I think would be very valuable is economic calculation/pricing. The goal would be to simulate a market economy in which participants “discover” the prices that tend toward optimal production and then respond as external events modify the value of their output as well as the costs of the inputs. To me this would be really cool but I haven’t yet come up with a straight forward way to create this kind of experimental environment.

In a somewhat related matter, Virginia Postrel wrote an article http://www.theatlantic.com/doc/print/200812/financial-bubbles concerning financial bubbles that quotes Charles Noussair, a professor at Tilburg University, in the Netherlands as saying that financial bubbles are a consistent result of experiments by Vernon Smith and others.

I have just finished writing a draft paper, “Downturn and Recovery” that has a variety of theories that could potentially add to your ideas for experiments.

Good luck. I think that this is exciting and especially important now.

John Bailey
(706) 252-2474
jbailey@sprintrome.com

[Noreen]

I would like to see these videos reviewed.

http://www.youtube.com/watch?v=fIqyCpCPrvU&feature=related

http://www.youtube.com/watch?v=YsB_rnzBA08&feature=channel

http://www.youtube.com/watch?v=Mw7LtVwDCbs&feature=channel_page

[louis vuitton]

The comments seem to be talking all around the question but not addressing the heart of the entreprenurial debate. The issue should be: How to get state capital around the financial logjam and into the hands of the entrepreneur.

[replica watches]

The comments seem to be talking all around the question but not addressing the heart of the entreprenurial debate. The issue should be: How to get state capital around the financial logjam and into the hands of the entrepreneur.