Stimulus more or less? A failure not being acknowledged. PART 4

The economic impact of fiscal stimulus isn’t a partisan issue. But, it is easy to be misinterpreted, intentionally or otherwise.

It’s time to turn to the actual article entitled “How the Great Recession Was Brought to an End” by Blinder and Zandi. While the previous posting focused on what The Hedged Economist thought was most useful (Appendix A of the article), this posting focuses on what is of interest to most people commenting on the article. Clearly, since it is also the focus of the executive summary, it is what the authors thought was most important. Unfortunately, it is also the source of a lot of misunderstanding and spin.

One has to read it closely and not extend the findings beyond the actual results. It is also almost essential to deconstruct how the results were derived. But, if one does, one finds it is fairly free of partisan nonsense. Further, the effort is rewarded by a sharper appreciation of the limitations of the analysis. It is to the credit of the authors that they frequently point out the limitations, but it seems legitimate to fault them for not always emphasizing the most important limitations.

There are aspects of the assessment that some might call methodological biases with partisan implications, but they could also be characterized as methodological assumptions. They will be discussed after pointing out some other reasons for citing just this one analysis.

First, The Hedged Economist has had the opportunity to walk through the details of the Moody’s Economy.com Macroeconomic Model equation-by-equation and linkage-by-linkage. That’s either masochism or curiosity. It’s time-consuming in any case. The review was probably sufficiently recent to still be relevant. (Models evolve if you didn’t know). Equivalently detailed reviews of others models aren’t as current, and the investment that would be required to review the models used for every other assessment would challenge any realistic time constraint (certainly mine).

Further and probably more importantly, Moody’s Economy.com Macroeconomic Model is well-documented and available for review. That doesn’t seem to be true of the models used for every assessment. Thus, it isn’t The Hedged Economist review that matters; it is that anyone could do the same thing.

The article contains a description of modifications to the model that were made for the assessment. Given an understanding of the model, they all make sense. If one was evaluating the model itself or wanted to use the Blinder / Zandi simulation as a starting point for their own assessment, one would need the actual equations for the modifications as well as all adjustments, but there’s enough information that is easily available to justify recommending the article. Authors make forecasts and do assessments, but the model is a tool that tells something about how the forecast was made.

Now, on to the article, let’s see what is says and what it doesn’t say. The executive summary of the Blinder and Zandi article provides their expert opinion. Those who only want the expert opinion of the authors should read it, but read it carefully. They never say what the impact of the policy was. Rather, they say what the impact of the policy would be relative to some other policy. Not stressing the unrealistic nature of the other policy is a major deficiency of both the executive summary and the article. With that in mind, one would do better reading the summary than reading most reports on the article. It’s only one page.

Also, a minor point, but the reader should note that in the executive summary some results are presented in terms of “payroll employment.” That is understandable given the accuracy of payroll data. However, it may overlook self-employment dependent on whether the self-employed person takes income as wages or proprietors’ income. It’s a minor point, but in previous postings The Hedged Economist has pointed out that over this cycle self-employment has behaved differently from previous cycles (or at least with a different lag).

The first section of the article is the assessment itself. It contains enough of an explanation of how the conclusions were reached to stand alone. It can stand alone for anyone who has a general understanding of how macroeconomic models work and is prepared to accept Blinder and Zandi’s modeling efforts without questioning. Appendix A was discussed previously, and Appendix B is a summary of the model.

The presentation of findings is only five pages of text and five tables, yet one suspects that many people who comment on the article didn’t even bother to read this section, didn’t realize its implications, or chose to ignore the implications. Maybe they read the Executive Summary and not very carefully. Perhaps they felt they could cite the article to support a conclusion they had already reached.

Most of the five pages of text in the first section of the article are used to explain how the analysis was done. That in itself is a reason to read it. The explanation of the approach provides the basis for an informed judgment about how to incorporate the results into one’s own assessment.

The basic technique is to run multiple scenarios assuming various policy responses. The definitions of the scenarios shed light on the uncertainty associated with the results of the analysis. The scenarios’ definitions also are essential to understanding what results could possibly be achieved and thus how relevant the results could be.

First, let’s look at what the definition of the scenarios says about the level of uncertainty associated with the results. It is worth noting the scenarios are not for a point in time or time from implementation to today. The scenarios are through the cycle. In other words, every scenario includes and element of forecast.

The primary scenario, common to each comment on impact, is the baseline scenario, the Moody’s Economy.com forecast. The authors are quite open about the fact that “most private forecasters…misjudged how serious the downturn would be.” Although said in a different connection, it indicates the difficulty of forecasting. If the baseline is wrong, every impact assessment is wrong. But, they could all be wrong without their rank order changing. The authors go to some lengths to point out that they are making estimates.

The other scenarios are “counter-factual.” They are based on different policy responses. So, not only is the forecast an estimate, as the authors point out, all the scenarios are estimates. Estimates are what any assessment is about; so, that there are estimates shouldn’t be a great concern. If they are, one should read Appendix A and skip the article. Here we have two experts making estimates for us for free; life’s great. One shouldn’t begrudge them the effort required to understand what is being estimated.

Not defining what is estimated is the source of most of the misinterpretation and misrepresentation. So, let’s dive in. First let’s look at scenarios common to every estimate.

The forecast is one. That’s inherent in any estimate since we are not through the cycle. At best, we are in the recovery phase. Any assessment that clips the estimate at the present would, by definition, be short-changing any positive impact.

A second scenario common to the entire assessment is the “no policy” scenario. But, here’s the problem: “no policy” is NOT a realistic counter-factual. Using a “no policy” alternative comes off as setting up a straw-man in order to make the policy look good.

That impression is only reinforced by not responding to statements citing the article like Treasury Secretary Tim Geithner’s statement: “The combined actions since the fall of 2007 of the Federal Reserve, the White House and Congress helped save 8.5 million jobs and increased gross domestic product by 6.5 percent relative to what would have happened had we done nothing.” (emphasis added). The deceptiveness of Geithner’s use of “we” deserves comment. As the article points out, the policies pursued to stabilize the financial system and stimulate the economy involved two administrations, two congresses, and the Federal Reserve.

However, the flaw isn’t in Geithner’s statement; it is in the “done nothing” counter-factual which he and the authors are both using. Government exists and its very existence is doing something. Not stressing how ridiculously unrealistic the “no policy” alternative is, more than anything else, was a major oversight by the authors. One is inclined to wonder why the authors stress that the assessment involves estimates without emphasizing they are estimates from an unrealistic counter-factual. As an academic exercise, their “no policy” scenario makes sense, but not stressing this specific limitation of the analysis is truly unfortunate.

What is particularly unfortunate is that the authors make passing comments that could have been excellent points of departure for a discussion of a more realistic counter-factual. For example, they note: “Some form of fiscal stimulus has been part of the government’s response to nearly every recession since the 1930s….” They then describe a few without future comment.

Now, one need not be a genius to understand the logical reason for a no policy alternative. It’s easier, more dramatic, and appears less speculative. Why easier? It is easier to compare actual experience with one counter-factual than generate and justify two counter-factual scenarios. Why more dramatic? The difference between another policy response and the actual policy response would be less than between the actual policy and nothing (assuming a positive response to both policies). Why does it appear less speculative? Even if “no policy” is a ridiculous counter-factual that hasn’t occurred in close to a century, and to some extent because it is so ridiculous, one can’t mistake it for an advocacy position. More importantly, it doesn’t require speculation about what might have been the policy. “No policy” was never on the table.

Despite the clean alternative that “no policy” provides, it ducks the important question. The important question is what would have been the performance of the economy under a different policy response? The authors point out: “It is … not difficult to find fault with isolated aspects of the policy response.” They then go on to question some. So, they acknowledge the issue.

One debate that deserves comment is the discussion of whether the stimulus in particular was too much or two little. When looked at in detail, much of the argument for a better response from a larger stimulus is based on the absurd notion that the response to stimulus is dollar-for-dollar the same regardless of the size of the stimulus. It ignores decreasing marginal utility, or, at a minimum, subscribes to the purely ideological notion that decreasing marginal utility never applies to government actions.

The authors aren’t guilty of such absurd reasoning. They point out that as the economy approaches higher levels of resource utilization additional stimulus has a smaller beneficial impact. Similiarly, at the other extreme, in an interview about the analysis (to be discussed in a subsequent posting) Dr. Zandi mentions that the beneficial impact is greatest when there is slack in the economy. However, there is still one of those pesky methodological biases or assumptions lurking in the analysis. Namely, the non-linear response is achieved without addressing whether the actual responses are linear. The difference in the response is due to the state of the economy. It is equally likely the response is NOT independent of the size of the stimulus. But, let’s depart from squabbling over methodological trivia. Linear specifications in simultaneous models are a simplifying assumption that yields tremendous benefits.

There are more basic questions about potential alternative policies. Basically, the underlying philosophy of the entire policy response should be questioned. It is not the need for a policy response, but the assumption inherent in parts of TARP and most of the fiscal stimulus that is questionable. Both are predicated on the assumption a trickle-down approach is best. In essence: give the money to a government, an investor, an automaker, etc., and just count on it to flow to the general benefit of the population.

In the popular media, the trickle-down approach is most likely to be challenged in the area where it actually doesn’t apply: TARP loans. As an example, read postings like Ritholtz’s “2008 Bailout Counter-factual.” Alternatively, one only needs to turn on TV or talk radio (of either conservative or liberal leaning) to witness the total failure of commentators to differentiate between a loan and an out-and-out income transfer. One has to be awfully glad that people who consider a loan a bailout aren’t managing your personal finances. One also suspects their personal finances and certainly their investment advice is something to avoid like the plague.

The actual policies focused on where the problem built up or collected (financial industry balance sheets) and the resulting cash flows in the general economy (consumers forced to cut expenditures). Imagine as a counter-factual that instead of relying on a trickle-down strategy, the government had recognized that the problem had its origin in balance sheets. As a counter-factual let’s have the government focus on the ultimate source. Ultimately, consumers and investors hold the assets and have the liabilities. So, let’s focus on households rather than banks and governments.

As mentioned in the previous posting, a careful analysis of the content of TARP and the stimulus bills can yield a slightly higher estimate of the fiscal stimulus. It works out to just about $10,000 per household using a 2008 household count. So, instead of targeting governments, banks, auto manufactures and their unions, insurers, etc. -- as a counter-factual assume a one time $10,000 rebate to every household.

Since we’re aiming to reduce both TARP and replace the fiscal stimulus, earmark it as only useable to pay off debt. Since elected officials consider themselves such financial wizards, one has to allow them to prioritize which debts get paid first. But, clearly, one priority should be to get non-performing loans out of the system. Therefore, targeting debt in arrears makes sense. Also, since housing finance was a recognized problem, debt secured by owner-occupied residential real estate seems a likely target. Since elected officials are partial to the UAW and automakers, put auto loans as of the enactment date in there, too. Further, the government would have to take steps to ensure the rebates were actually used to shrink balance sheets rather than as a credit substitute. It’s doable.

Now, what about households with no debt? Well, to control the deficit, their rebate could take the form of non-transferable government bonds. Since the government is executing a massive transfer of liabilities from consumers’ balance sheets to their own, the bond idea appeals. However, just plain old money has a certain appeal as a stimulus. Each has its own appeal. Keynesians may prefer money that has to be spent while fiscal conservatives might like the bonds.

This is a weird counter-factual, admittedly, but no weirder than “no policy.” Further, it focuses the discussion where it belongs. Is a trickle-down approach like ARRA and TARP subsidies the best we can do?