A classroom experiment in monetary policy

Abstract

The authors propose a classroom experiment implementing a simple version of a New Keynesian model suitable for courses in intermediate macroeconomics and money and banking. Students play as either the central bank or members of the private sector. The central banker sets interest rates to meet twin objectives for inflation and the output gap or to meet only an inflation target. In both settings, private sector agents are concerned with correctly forecasting the inflation rate. The authors show that an experiment implementing this setup is feasible and yields results that enhance understanding of the New Keynesian model of monetary policy. They propose alternative versions where the central bank is replaced by a policy rule and provide suggestions for discussing the experimental results with students.

Keywords:

• experiment
• monetary policy
• undergraduate teaching

JEL codes:

• A22
• C90
• E52

Acknowledgments

The authors are grateful to Tyler Boston and Jason Ralston for research assistance and for the constructive feedback of two anonymous referees.

Notes

1 Contemporary textbooks by Chugh (2015), Mankiw (2016), Mishkin (2016), and Jones (2017) each include business cycle models with some incarnation of the expectations-augmented Phillips curve.

2 Resources for running the experiments described in this article, including z-Tree programs along with data and programs for reproducing our figures and statistics, are available at https://github.com/letsgoexploring/monetary-policy-game.

3 Link: http://people.whitman.edu/∼hazlett/econ/

4 In addition, as the game is played repeatedly, students can learn from past realizations of inflation to form more accurate forecasts of inflation in the current period.

5 In practice, central banks set the nominal interest rate. But because expected inflation in our framework is determined before the central bank sets policy, the Fisher equation implies that the central bank implicitly sets the real interest rate with its choice for the nominal interest rate. Therefore, for simplicity, we assume that the central bank directly sets the real rate.

6 See note 2 for a link to our GitHub repository that contains software and instructions for running the experiments with z-Tree.

7 The task faced by the private sector players constitutes a “learning-to-forecast” experimental design. See Hommes (2011) for a survey of other experiments using this approach.

8 As an alternative to the inflation targeting regime’s objective function, it might also be of interest to consider a price level targeting regime by changing the central bank’s loss function to $20-\frac{1}{T}{\sum }_{t=1}^T\left({\pi }_t-{\pi }^{*}\right),$ where T is the current period of the experiment. This regime allows for periods of inflation below (above) the target value provided that they are later matched by periods above (below) the target value, which works to better stabilize the level of prices, but can lead to greater volatility in the inflation rate.

9 Such knowledge facilitates computation of the rational expectations equilibrium. One could relax this assumption, for instance, by making the central bank’s targets for inflation and the output gap private information (known only to the central bank) that the private sector would have to learn over time.

10 We note further that while the quadratic loss functions admit negative point earnings, our program truncated the point formula at zero, so that no subject actually earns less than zero points.

11 For example, Givens (2012) estimates the coefficient on output in a New Keynesian Phillips Curve to be about 1.5 while Ireland (2004) calibrates the value to 0.1.

12 In a calibrated model, Woodford (2003b) obtains a weight of 0.048 on output gap fluctuations in the loss function but Givens (2012) estimates values between 0.0987 and 0.1351 while Givens and Salemi (2015) estimate values between 0 and 0.5667.

13 A disadvantage of the latter design is that inflation expectations may be widely varied, at least initially.

14 Our program chooses one set of demand shocks for all groups in a given session. Each session consists of between four and six groups. Thus, while there is some variation in demand shocks, there is not as much variation in these shocks as there are groups of players.

15 According to former Federal Reserve Chairman Ben S. Bernanke, inflation expectations are “anchored” if they are “relatively insensitive to incoming data” (Bernanke 2007, online).

16 Some central banks, like the Bank of Korea and the Reserve Bank of New Zealand, pursue strict inflation targeting policies while others, like the European central bank and the Bank of England, follow a hierarchical regime in which meeting an inflation target is a primary goal that takes priority over other stabilization objectives.

17 Under our parameterization for the normally-distributed demand shocks, 98.8 percent of all realizations will lie within the interval [−1.25,1.25], making the extreme values comparable to the uniform shock case, although the normally distributed errors will, of course, be more tightly concentrated around 0.