DSGE policy tool adapted by Bangko Sentral yields biased insights 340.0

DSGE is the newest type of policy analysis tool adapted by the Bangko Sentral(BSP). It complements the existing set of econometric models it uses as aids in policy making, particularly inflation targeting.

Its quite-a-mouthful and technical long name is Dynamic Stochastic General Equilibrium.

As with the Corden and Neary framework that BSP analysts used in showing evidence of Dutch disease in the Philippines, DSGE is a pioneering effort to use a leading edge modeling technique to simulate the Philippine economy.

The image at left is of the December 2009 working paper; a subsequent Powerpoint presentation in October 2010 showed the progress of the Team’s work.

Unfortunately, even as the developed models indicate that the authors have the math, I believe they have not exhibited the venturesome mind that would have made the adapted tool yield the best insights, its declared primary goal.

In fact, it yields insights that follow a neoclassical bias that have made the Philippines path diverge from its more venturesome peers in Asia. The Philippines did not converge with the rest of the neoclassical world as the theory assumes that follows from diminishing returns.

This post provides three levels of critiques on why DSGE leads to biased insights: (a) general and technical issues within the neoclassical, (b) critiques from new institutional and evolutionary economics and (c) specific issues to move the BSP-adapted DSGE model closer to Philippine reality.

The goals of the DSGE as limited tool for the Bangko Sentral

The BSP and the National Economic Development Authority – Philippine Institute of Development Studies (NEDA-PIDS) are two government institutions that use econometric models to simulate the economy.

This blog post is about the BSP’s pioneering effort to adapt a DSGE model for the Philippines that I paraphrase from the working paper as to its purpose and areas of inquiry:

This paper is a pioneering effort in the Philippines to use the DSGE framework for [a] small and open economy, capturing the key characteristics of the Philippine economy.

We solve the model to assess what its implications are with respect to the responses of key variables to internal and external shocks, the way alternative policy regimes can change and how key variables respond to such internal and external changes. The principal goal is insight.

What is DSGE (with some brief comments)?

From Wikipedia, I paraphrase the following technical definition:

Dynamic stochastic general equilibrium modeling (DSGE) is a branch of applied general equilibrium theory that … attempts to explain aggregate economic phenomena, such as economic growth, business cycles, and the effects of monetary and fiscal policy, on the basis of macroeconomic models derived from microeconomic [foundations].

More popularly, my own English definition and brief comments begins from its keywords:

1. DSGE is dynamic because it attempts to consider changes over time within the constraints of neoclassical math modeling. To simplify the math, these constraints are often uni-directional and a single rate of change that do not reflect the reality of messy change in the real world.

In the BSP-adapted DSGE model, an example of this are the calibrated parameters θ for inter-temporal elasticity of substitution. They are constants that assume a fixed rate of change.

2. DSGE is stochastic because it is random and probabilistic instead of structural and deterministic as the older econometric models have been criticized.

While it uses Monte Carlo simulation with random numbers and thousands of displacements (repetition), it still uses as prior assumption a given probability distribution. While not necessarily Normal with the usual standard deviation (SD) proportions, I think, the expected values as calibrated or inferred will still be proximate to those of independent variables in the deterministic econometric models.

The stochastic process that can be used in the model, with the time dynamics, do not capture the truly random change in the real world even in the technology field alone.

Note: See below on other hidden assumptions behind the Cobb-Douglas production function.

For example, the calibrated Taylor coefficients ρ used in the monetary policy (Equation 58) of the BSP-adapted DSGE model attributes 90% of this period’s policy rate to the prior and less than 10% to inflation and output gaps. This calibration setting seems to be for a very short-term use.

Furthermore, this confirms my intuition from the previous linked first above based on sparse published data that unemployment is hardly a consideration in setting policy rates unlike in developed countries that the BSP-adapted DSGE model emulates.

It also confirms that the Monetary Board accepts that our labor economy is stuck within rigidities (the sticky wage assumption is not that important given the residual parameter) as is typical with many emerging markets (and Spain!).

3. DSGE is a form of general equilibrium. This implies neoclassical assumptions that bias the insights like perfect information, full employment, price -clearing, free and ubiquitous technology, and one type of labor with one price (stickiness assumption included).

I have written about corrections made by later generations of Nobel Prize winning economists in a blog post, Economics and Crises – an Epic History of Self-Healing on February 2011, on the hidden assumptions of the original Marshallian supply and demand equilibrium thesis (from 1890).

In a sense, insights from the model can only be assumed to hold for a differential time if they hold at all. Note that German social market economy and the success of our neighbors like Japan, South Korea, Taiwan and Singapore are not based on the neoclassical tradition but on more structural, evolutionary approaches.

I present more (a) general and technical as well as (b) specific model-related on the Philippine DSGE below. It may well work for a specific purpose like for testing specific policies like those on inflation targeting on specific shocks but, as it is designed, but may not work at the macro level.

SYNTHESiST and the Bangko Sentral

SYNTHESiST is not an economist or a central banker. I volunteer to follow the Bangko Sentral in behalf of innovators, in general, who have the vested interest for a stable peso to facilitate the conduct of long gestation innovation projects. My posts in SYNTHESiST on the BSP Watch are archived in Innovator Peso above.

SYNTHESiST is an engineer and management practitioner. Therefore, my critiques take the perspective from applied microeconomics; from what is useful from industry and management rather than macroeconomic perspective.

I was of the generation that first studied Operations Research in engineering college and implemented it for simple problems – more akin to the older MEM and SEM models especially of the structuralist variety.

I just have enough of the math to understand, translate to English and, I hope, make sensible critiques to the models to English but not enough to create.

Innovation comes from a virtuous interaction of new ideas and new tools

Still, I believe that much can be learned from analyzing DSGE as new policy analysis tool. On a very early SYNTHESiST post on the sources of innovation, Kuhn and Galison Defines Virtuous Cycle from New Knowledge and New Tools in March 2009, I opined that new insights come from new ideas and new tools that, working interactively, create scientific revolution.

Thomas Kuhn is well known for his book (Kuhn 1962) on scientific revolutions and for inventing the concept of paradigm shift. Peter Galison is not as well known with his Image and Logic (1997) that tells the story of how ideas hinted at by theoretical physicists were proven when new tools of material science – from the microscope to the collider in their times – were invented to correctly measure phenomena to prove the ideas.

This blog post, therefore, attempts to review the BSP-adapted DSGE to see if, as a tool of economic science, it will yield the appropriate economic insights even in the specific policy are of inflation targeting as intended by the BSP for an emerging market like the Philippines.

Three critiques of the BSP-adapted DSGE model

This post provides three levels of critiques on why DSGE leads to biased insights that are discussed further below:

• General and technical issues within the neoclassical,
• Specific issues to move the BSP-adapted DSGE model closer to Philippine reality.
• Critiques from new institutional and evolutionary economics and

General and technical and BSP-adapted DSGE model-specific comments

The working paper authors have addressed the most important general and technical issues on DSGE like the identification problem and the Lucas critique that are within the ambit of the neoclassical in the design of its adapted DSGE.

For example, the authors have referred to landmark citations like Smets and Wouter (2003) to incorporate enhancements like frictions to improve the neoclassical model.

The references below list four other general and technical critiques that apply to the BSP-adapted DSGE:

• Kirman (1992) and Rizvi (1994) question the use of representative agents [“atomistic”] behaving as average individuals in microfoundations for a global economy (Note: The use of first-order microfoundations was originally to correct for the Lucas critique.),
• Tovar (2008) highlights a host of issues but I appreciate his critique pertaining to emerging markets like the Philippines, and
• Schorfheide (2011) lists five possible areas of mis-specifications including issues on low frequency priors that may not provide enough basis for posterior insights. Schorfheide does list the work being done to fix these mis-specifications and concedes that much progress has been done and the procedures, already taught in PhD courses, as already standard.

On these issues of concern within the neoclassical area and given the limited scope in inflation targeting and goal as providing insight, I conclude that the DSGE can be useful.

The cautionary tale though is that DSGE model users must know the strengths and weaknesses inherent in the new tool.

Specific and Philippines-centered critiques of the BSP-adapted DSGE model

From the schematic diagram of the BSP-adapted DSGE model, there are three missing vectors that need to be considered to make the BSP-adapted DSGE model move closer to Philippine reality. These three already have falsifiable time series data (in the Karl Popper sense) given in the image below:

• OFW and Migrants with their remittances
• Structural unemployment and underemployment
• For its limited purpose in inflation targeting, a dummy variable on BSP monetary sterilization – the Special Deposit Account (SDA)

The table below shows the substantial size and remarkable growth of these agents/transactions.

OFW and Migrants. Members of the BSP Policy group have already shown evidence of Dutch disease in the Philippines as I posted about in the first link above. Because behavior drivers are different and the flows are significant, exported labor must be reflected as coming from Households.

Structural unemployment and underemployment. Standard neoclassical economics have placed a link between inflation and unemployment via NAIRU and the Phillips rule. Even as labor rigidities are accepted as making the unemployment an intractable problem, the model must include at least a dummy variable – rather than blindly accept an assumption of full employment – possibly in the Household sector, i.e. the CES equation.

Sterilization of remittances. The growing SDA account from 2006 and its existence as the biggest liability account requires that an SDA transaction ought to be added between the BSP and the banking sector. In addition, the SDA are investments by banks and not reserve requirements thus they behave and are incentivized differently.

Making the three adjustments makes the BSP-adapted DSGE model more realistic and more specific to the Philippines.

On this issue, I conclude that these three are necessary changes unlike the first two sections above – where the question is more of technicality and theoretical perspective – for the goal of insights to be realistic.

Without these changes, the steady-state -> shock-> steady-state becomes a mere exercise of analysis using neoclassical theory.

Critiques from economic history and new institutional economics and from evolutionary economics

From economic history and new institutional economics, Douglass North (2008) questions mathematical modeling itself and I paraphrase: There is a hidden assumption of ergodicity in an emerging market reality of path-dependence.

Starting from Heiner (1983) on The Origin of Predictable Behavior, North argues that people create institutions to simplify their understanding of the world. Individual agents represented mathematically cannot capture the complexity of the real world.

From evolutionary economics, Schumpeter (1911) and later modified by Nelson and Winters (1982) argue that the inter-temporal dynamics approach used by the DSGE is not realistic. Starting and ending in steady-state -> shocks -> steady-state is not how change happens in the world.

The DSGE’s Monte Carlo process is still just random number generation in a calibrated probability distribution. It does not reflect for example the stop-and-start, fast-and-slow process of real technology innovation for example.

I add two technical concerns, myself, from outside the neoclassical tradition:

First, the use of inferred priors that pass goodness-of-fit tests on time series data assume that the past will repeat itself. In their secular linearity, this approach does not consider the rise of innovators and entrepreneurs Thomas Edison, Robert Noyce, and Steve Jobs that do create episodes of increasing returns.

Innovation and entrepreneurship are externalities not built into the 60-year old Cobb-Douglass function even as recent evidence in the neoclassical tradition, itself, like endogenous technological change (P Romer 1990) hints at the possibility of increasing returns tempered by mechanism like new types of events channel congestion rather than price-based competition.

On this, my conclusion is that the DSGE model yields only the results expected from a neoclassical model.

Strictly speaking, it cannot compute solutions that involve change in institutions, innovation and entrepreneurship.

As such, it may quantity numerical gaps, like investments that may be filled up, say, by infrastructure, but recommendations in external areas need to be justified with further studies.

Second, the Cobb-Douglass production function implies diminishing returns and also implies convergence when the real world indicates that convergence happens only with take-charge leadership (modern industrial policy!) and not the free-markets.

SYNTHESiST has all the case studies including emerging America and the strong hand of Alexander Hamilton building the federal infrastructure in the early 1800s (also the subject of Thomas Sargent’s 2011 Nobel lecture).

The Philippines is proof, herself, that divergence is more the reality if only gaps like savings and investment are filled by investments or foreign aid in neoclassical analysis.

Finally, our successful neighbors like Japan, South Korea, Taiwan, Singapore, and China show that co-evolution and institutions through modern industrial policy are necessary for success.

Summary and conclusions on DSGE

Much work needs to be done, in addition to the substantial effort already expended, to have a new policy analysis tool that works for the BSP and the country.

It is good to see that our technocrats are just behind the leading edge of innovation. Apparently, we have the skills but do not have the venturesome mind to improve on what we emulate if only for relevance.

Within its limited goal of enriching the policy discussion on inflation targeting, adapting the realistic changes will make for a relevant tool.

Selected references:

Galison, P. (1997). Image and Logic. University of Chicago Press.
Kirman, A. (1992). Whom or What does the Representative Individual represent? Journal of Economic Perspectives , 6 (2), 117-136.
Kuhn, T. S. (1962). The Structure of Scientific Resolutions. University of Chicago Press.
McNelis, P., Dakila Jr, F., Glindro, E., & Co, F. (2010). Dynamic Stochastic General Equilibrium Model for the Philippine Economy. Center for Monetary and Financial Policy. Bangko Sentral ng Pilipinas.
McNelis, P., Glindro, E., Co, F., & Dakila Jr, F. (December 2009). Macroeconomic Model for Policy Analysis and Insight (a Dynamic Stochastic General Equilibrium Model for the Bangko Sentral ng Pilipinas). Bangko Sentral ng Pilipinas. Center for Monetary and Financial Policy.
Nelson, R., & Winter, S. (1982). An Evolutionary Theory of Economic Change. Belknap Press of Harvard University Press.
North, D. C. (2005). Uncertainty in a non-ergodic world. In D. C. North, Understanding the Process of Economic Change (pp. 13-22). Princeton University Press.
Rizvi, S. (1994). The microfoundations project in general equilibrium theory. Cambridge Journal of Economics , 18, 357-377.
Rotemberg, J., & Woodford, M. (1997). An Optimization-Based Econometric Framework for the Evaluation of Monetary Policy. In NBER Macroeconomics Annual (Vol. 12, pp. 297-346). The MIT Press.
Schorfheide, F. (2011). Estimation and Evaluation of DSGE Models: progress and Challenges. National Bureau of Economic Research.
Schumpeter, J. (1934). The Theory of Economic Development. Harvard University.
Smets, F., & Wouters, R. (2003). An estimated Stochastic Dynamic General Equilibrium model for the Euro area. European Central Bank.
Taylor, J. B. (1999). A Historical Analysis of Monetary Policy Rules. In J. B. Taylor (Ed.), Monetary Policy Rules (Vol. 31, pp. 319-347). Chicago: National Bureau of Economic Research.
Tovar, C. (2008). DSGE Models and Central Banks. Monetary and Economic Department. Bank for International Settlements.

References (SYNTHESiST Posts):

Beduya, M. (2011, December 14). Dutch disease raised real interest rates and caused deindustrialization. Retrieved December 20, 2011, from SYNTHESiST: http://synthesistblog.com
Beduya, M. (2011, February 7). Economics and Crises – an epic history of self healing. Retrieved December 20, 2011, from SYNTHESiST: http://synthesistblog.com
Beduya, M. (2009, March 5). Kuhn and Galison defines virtuous cycle from new knowledge and new tools. Retrieved December 21, 2011, from SYNTHESiST: http://synthesistblog.com