Some extra money from the last Federal budget went to fund new projects at universities. As part of its application for funds, each university must say how it plans to spend the money, and also do an "Economic Impact Assessment". I have been called in to help. I know little about EIA. I remember a bright, cynical, graduate student once say "Add up all the money you spend, multiply by 2, and tell a story".
So I spent a little time on the internet reading about EIA, and a bit longer thinking about EIA. I just wanted to share some of my naive thoughts about the relation between EIA and the way macroeconomists normally think about fiscal policy.
1. EIA is very different from Cost Benefit Analysis, and is much closer to traditional "Keynesian" macroeconomic analysis of fiscal policy. EIA, like traditional macro, is just looking at jobs created and income generated. It is not looking at whether the output created is useful, and whether the resources employed have an opportunity cost. To put it crudely, each dollar of extra spending is seen as a dollar of extra benefits, and there are no costs. Or, costs are benefits. Except EIA isn't really about benefits, it's about impacts, though it is easy and tempting to slide mentally from one to the other.
2. One minor difference between EIA and macroeconomic analysis of fiscal policy is that EIA divides impacts into three categories: direct, indirect, and induced. Macroeconomics divides the impacts of fiscal policy into two categories: direct, and induced. It's a minor difference, because the macroeconomist's "direct" is just the sum of the EIA's "direct plus indirect". If you spend $100 to build a new classroom, a macroeconomist would say there's a first round effect of $100, while the EIA likes to break that down into the $60 income generated on-site, and the $40 income generated off-site that the contractor spends on materials. The "induced" effect is the same marginal propensity to consume stuff in both EIA and traditional macroeconomics.
3. EIA looks more closely at where the extra income is generated. How much in the city, how much in the province, how much in the country? There are different multipliers for each. Macroeconomists normally just consider the country, and the world as a whole. Not a big deal, just a difference in what questions are seen as relevant.
4. Here's the difference. Macroeconomists normally pay attention to whether or not there are unemployed resources, and the effect of increased government spending on interest rates, exchange rates, and prices. EIA doesn't (or doesn't seem to, from my cursory reading). EIA seems stuck in the first-year classroom, where interest rates, exchange rates, and prices are fixed, and there are always unemployed resources. With full employment, so that interest rates rise, exchange rates appreciate, and prices rise, macroeconomists' fiscal multipliers tend towards zero (for permanent increases in government spending). EIA multipliers stay big (two seems to be considered conservative). EIA is pure demand-side; macroeconomists pay at least lip-service to the supply-side.
5. Now you might argue that the differences in 3 explain and justify the differences in 4. If I am a city mayor, wanting to make my city grow bigger, I don't need unemployed workers in my city in order to make income grow by creating jobs. I just need those jobs to attract workers from other cities, who may be fully employed elsewhere, but will move to my city if the jobs are better. But if that were the explanation, we would expect large local multipliers, and small national multipliers. Labour mobility is higher between cities than between provinces, and higher between provinces than between countries. We see the opposite. EIA assumes, just like traditional, simple macroeconomics, that the multipliers will be smaller for a small area than for the world as a whole, because the leakages of demand through imports (the marginal propensity to import) will be greater in a small area than in a large area. That only makes sense if there are unemployed resources, and no changes to interest rates, exchange rates, or prices.
6. Maybe it's just a "fallacy of decomposition" (I just made that up). We can imagine the Mayor chuckling "but my little project is far too small to affect interest rates and exchange rates, so let's ignore that fancy stuff!". The Mayor is wrong, of course. The feedback effects via interest rates, exchange rates, and prices, may be small, but then so is the project. Those feedback effects can be just as large, as a proportion of the project, as would the feedback effects of a larger project. In a linear model, the proportion will be identical. One big project should have the same economic impact as lots of little projects. It all ought to add up.
But I bet it doesn't add up. I bet that if the economists at the Department of Finance add up all the Economic Impact Assessments they get back they will find that the "fiscal stimulus" will far exceed their wildest estimates. (My guess is they will just "file" the EIA's.)
It's not just imaginary mayors who fall victim to the fallacy of decomposition. "Canada's little fiscal policy will make no difference to world GDP". It sounds plausible.
I have a lot of experience with Economic Impact studies (they are very popular in the consulting world). It is hard to take the estimates produced with multipliers at face value given the lack of constraints.I also don't like the lack of dynamics in EIA which is why I like to also use an error-correction model with GDP anchored to trend (plus other dynamics) and then use the AR(1) term to estimate the propagation of shocks.
Posted by: brendon | March 17, 2009 at 10:29 AM
Thanks for sharing your thoughts on this. EIA's are something that those of us in gov't who deal with infrastructure (maybe I should only speak for myself) struggle with. On one hand, as Brendon states, they are popular in the consulting world and can be quickly done. A consulting firm with a local/provincial input-output model can crank out a report in less than a week. On the other hand, the robustness of the analysis can be called into question.
Perhaps it's best that all this is kept as simple as possible. Not every government analyst who is employed as an economist has a grad degree in economics let alone an undergrad in economics. Your average analyst doing the work-up on a project for a Minister will have difficulty interpreting the complexities of a CGE model. Doing a thorough CBA takes time and expertise.
Posted by: Adam | March 17, 2009 at 10:57 AM
Brendon, Adam: thanks for the replies. You two have a lot more experience in producing and using EIAs than I have. A couple of questions:
1. The I-O models: are they looking at the structure of production (how much steel will be used in the building, and how much labout, iron ore, and fuel to produce the steel, etc.), or do they also look at the impact on consumer demand (marginal propensities to consume, import, etc)?
2. What do governments actually use these EIAs for?
In current circumstances, I would think that the most interesting question to ask would be: which particular resources will be used in the project, and are those resources currently unemployed (or underemployed) or fully employed? The standard Keynesian multiplier stuff seems sort of irrelevant. There's little reason to think that workers employed in one local project will consume a different vector of goods than workers employed in another project (except for obvious local effects in the demand for immobile services).
Put it another way: The Department of Finance presumably wants/needs to know about the details of this particular project which might differ from other projects. The I-O analysis might do that, but the Keynesian mpc/mpi/multiplier analysis won't.
Posted by: Nick Rowe | March 17, 2009 at 11:26 AM
Hey Nick,
Isn't this your idea only in a different form?
Take care,
Don
"March 17 (Bloomberg) -- Chairman Ben S. Bernanke and Federal Reserve policy makers may have to ramp up their purchases of mortgage securities and other assets after the economy and job market deteriorated further since they last met.
The Federal Open Market Committee, gathering today and tomorrow in Washington, needs to redouble its efforts after the central bank’s balance sheet shrank 17 percent from a $2.3 trillion December peak, Fed watchers said. The retreat came even as Bernanke acknowledged the chance that the unemployment rate will exceed 10 percent for the first time in a quarter century.
“It takes massive balance-sheet expansion to generate significant easing in financial conditions,” said Andrew Tilton, an economist at Goldman Sachs Group Inc. in New York who used to work at the Treasury. “More needs to be done.”
This week’s FOMC meeting could mark a shift toward more aggressive monetary expansion to fight deflation after demand waned for many of the Fed’s existing programs. One top consideration is an increase in the pace and size of a $600 billion program to buy bonds issued and backed by U.S. housing agencies such as Fannie Mae, analysts said.
Other measures could include everything from purchases of Treasuries to corporate bonds, Tilton said. The Fed has already agreed to work with the Treasury on implementing a program to revive consumer and business loans, which the Obama administration has said could reach $1 trillion. "
Posted by: Don the libertarian Democrat | March 17, 2009 at 11:48 AM
Nick, You have the spirit of Economic Impact summarized in paragraph 1. It can actually be useful to answer specific questions, if one exploits the industrial, commodity or geographic detail that some input-output models provide. Tax incidence is one example.
EIA should usually be relabelled Political Impact Assessment. They are the favourite rhetorical weapon of self-obsessed special interest groups. They appeal to people with no or just a minimal, faulty understanding of welfare economics and measurement theory.
Social wealth is an appropriate criteria; total cost accounting benefit-cost analysis is one of way of getting at that criteria.
The big problem with Economic Impact Assessment is that it provides incentives for bad management. Take a popular table fish like west coast Pacific halibut. Recreational harvesters require a relatively large and powerful boat to pursue them, sometimes several nautical miles offshore. Mismanage these fish so our pluck sport fishermen have to travel farther distances in order to find good numbers of halibut and the economic impact increases. This example illustrates why this type of analysis is sometimes referred to as the "Broken Window Theory of Economic Value".
Open version input-output models generate GDP multipliers whose values are less than one (< 1). To get around this apparent problem, special interests do several things. For one they generate gross output multiplier values that exceed one and involve considerable double-counting or n-counting. For two, they use Closed version input-out models that include induced effects from respending and generate enormous values akin to those of naive Keynesian-Hicksian macroeconomic models.
In the worst case, I saw recently in a consultant's report comparing economic impacts of commercial and recreational sectors in the Skeena River of nothern British Columbia, the consultant used gross output multipliers in a closed-version input-output model.
A number of years ago, the Input-Output division of Statistics Canada discontinued the closed version model precisely because of chronic special-interest abuse. At the time, we advised British Columbia statistics that of the short-comings of this model. That advise fell on deaf ears.
Before concluding, it behooves me to point out that the multiplier and impact magnitudes from an I-O model are highly sensitive to direct and indirect subsidies. The danger of relying on the faulty use of EIA in policy analysis is that it encourages a vicious circle of public subsidies (or reduced net taxes if you prefer).
Input output models are like stiff narcotics. For the right application, at the right time, they can be incredibly useful. Otherwise, they can wreak all kinds of social damage.
- Erik Poole, White Rock, BC
Posted by: westslope | March 17, 2009 at 11:50 AM
Don: yes, that looks like quantitative easing.
Westslope: Very informative comment. I just want to clarify the difference between open and closed version input-output models.
Open versions ignore the effects of increased consumer spending. Closed (as in "closed-loop"?) versions include the effects of increased consumer spending, in a feedback loop. Both versions assume unemployed resources (no resource constraints). Employment is determined by demand for labour (and other inputs). Is that correct?
Posted by: Nick Rowe | March 17, 2009 at 12:18 PM
My principal problem with EIA is that the size of the multipliers are not sensitive to business cycle conditions (a particular problem for EIA in BC for construction projects).
As for what they are used for by government - propaganda largely - either to secure government funding or public opinion.
I have never been asked for a specific structure of production or specific impacts on consumption - just the output - GDP, Jobs (always the most important for bureaucrats) and taxes.
Posted by: brendon | March 17, 2009 at 12:23 PM
I think the difference is indeed a "fallacy of decomposition", except that, for a single decision maker on a single project, it is not a fallacy.
If I'm a mayor deciding whether to spend $10 million, the impact on interest rates, wages, money supply etc of my project really is tiny. Of course, if we added up all the $10 million projects of all the mayors in the country, that would not be true. But I am not in the position to make a decision on all those projects - therefore I should not take their effects into account in measuring the impact of my project.
The appropriate counterfactual in my decision is where my project doesn't go ahead, but all the others do. Thus in my personal cost-benefit analysis, the inflationary impact of all those projects is not relevant, and the marginal cost (to me) of my project is just the tiny extra price impact that I cause. The rest of my costs are an externality borne by all the other cities. On the other hand, my benefit is the demand effect of just my project, most of which I capture myself.
It's the same phenomenon as the tragedy of the commons. No matter what decision I make personally, I have to bearing all the externality costs of everyone else's independent decisions. My rational choice is to graze my sheep and screw the rest of the village.
Only a larger-scale coordination mechanism can change the overall rationality of this decision, by internalising the externality and making me bear the real cost of my own expenditure.
Posted by: Leigh Caldwell | March 17, 2009 at 12:29 PM
Nick wrote above:
Westslope: Very informative comment. I just want to clarify the difference between open and closed version input-output models.
Open versions ignore the effects of increased consumer spending. Closed (as in "closed-loop"?) versions include the effects of increased consumer spending, in a feedback loop. Both versions assume unemployed resources (no resource constraints). Employment is determined by demand for labour (and other inputs). Is that correct?
--------------------------------------
Nick: Yes. You are correct on all accounts.
To elaborate further, the common closed version I-O model is actually a partially closed model. Only consumer spending of generated labour income occurs; taxation, investment revenue and other factor incomes are ignored.
Full closing an input-output model and thus making it a full macreconomic model would mean completing the relationship between supply and demand by introducing realistic resource constraints.
Nevertheless, as Michel Truchon once pointed out to me, Statistics Canada's rectangular input-output models provide an empirically accurate and useful short-term snap-shot of the overall economy. (Some refer to this as the 'mesa economy' to avoid the potentially misleading term 'macro economy'.)
Leigh: Your reference to the Tragedy of the Commons with respect to how these numbers are potentially abused in social policy decision-making is spot on. Hence my plea for total cost accouting benefit-cost analysis to borrow terminology introduced to me by west coast ecological activist Pete Broomhall.
-Erik
Posted by: westslope | March 17, 2009 at 02:04 PM
Nick: you posed these questions earlier.
1. The I-O models: are they looking at the structure of production (how much steel will be used in the building, and how much labout, iron ore, and fuel to produce the steel, etc.), or do they also look at the impact on consumer demand (marginal propensities to consume, import, etc)?
2. What do governments actually use these EIAs for?
The one's that I have reveiwed do not get into the structure of production nor consumer demand in the way you've enquired about. All I usually see are figures representing the impact on various sectors (both public & private including households) and person-year equivalents for the number of jobs created. And then those results are categorized as direct, indirect, or induced impacts and a multiplier is calculated. That's all we usually get.
Wwhat do governments do with EIA's? In a word, communications. As others have stated above, if one actually paid attention to funding announcements, you'd see language to the effect of "this funding will contribute $X million to the economy and create Y jobs."
Posted by: Adam | March 17, 2009 at 03:56 PM
Nick, Because money is fungible, wouldn't universities have an incentive to claim the money was used for a project that looked good to those doling out the funds, even if it was a project they were going to do anyway, and the money was just being diverted in higher faculty salaries?
Posted by: Scott Sumner | March 17, 2009 at 07:52 PM
Also have to remember that the person commissioning the EIA is not the only one looking at it (most likely). There are different groups within the bureaucracy, and some of them have the job of promoting projects and others have the job of killing them. Most people with the job of killing them absolutely recognise all these limitations. They will have seen thousands of these things, and recognise that most of them aren't worth the paper they're written on. In some cases, they are worth the paper they're written on but only because of the general hilarity they create around the bureaucracy.
And a second to Adam's point that they're used for communications/sales jobs.
But you still have to do it if you want the $. Question is whether you want to use your academic reputation to help get the $.
Posted by: christine | March 17, 2009 at 09:10 PM
I have no experience with EIA's, but my guess is that projects that relieve or improve bottlenecks would have a very high multiplier effect, but not one you could calculate from an I-O analysis. You'd have to do a more general kind of impact analysis.
An example would be the bridge built across the Lachine canal 10 years ago in Montreal to replace the Atwater Tunnel. The tunnel was a bottleneck because modern trucks coul;d not fit through it. The bridge should have been built 30 years earlier. In the meantime, Canada's oldest industrial area collapsed because trucks could not easily reach the factories. This sort of impact cannot be measured by I-O analysis or simple-minded IEA's.
Posted by: Anonymous | March 18, 2009 at 07:21 AM
Leigh: yes, if I am a mayor, looking at the impact on my city, then feedback effects of national variables (exchange rates, interest rates, etc.) on my city are very small. But if I am a minister of the nation, evaluating the mayor's project, looking at the impact on the nation, those feedback effects are the same order of magnitude as the direct effects of the mayor's project on the nation. Both are small, but the feedback effects are not small relative to the direct effects. I think we agree.
Brendon's approach of doing a VAR seems to be reasonable, but I am wondering if this provides any "local", or project-specific information. Presumably any expenditure of the same $ value would have the same effects, according to the VAR model.
The sort of example that Anonymous gives (the bridge over the Lachine canal) seems to be exactly the sort of project-specific "local" information that would be VERY useful to know. It's not just dollars spent, and it's not just ordinary macro analysis of fiscal policy (which the national authority could do just as well or better than the local authority). But I agree, that this sort of very relevant information, which the local authorities would have but the national one's wouldn't, is unlikely to come out of any standard EIA. It's a supply-side multiplier.
Scott: yes. But my own gut-sense of university resource-allocation decisions, seen from the inside, is that there's a fairly large flypaper effect. Not sure why.
Christine (and thanks for my morning laugh!) seems to be summing up my general sense of dissatisfaction with EIAs. They seem like a large waste of resources (as currently done). Some sort of reform seems to be required. This is how I think it should be approached:
1. Decide who the audience/reader is for an IEA.
2. What decisions will that audience take on the basis of an EIA?
3. What information for that audience would be:
a, Useful
b, non-obvious
c, where the writer of the EIA has a comparative advantage over the audience in collecting and providing that information (e.g. local, project-specific information).
Posted by: Nick Rowe | March 18, 2009 at 10:10 AM
By the way, the EIA we needed for the university project just wanted 16 numbers: a 2x2x2x2 matrix.
The 4 dimenions were:
1. dollars spent vs jobs created (divide by $100,000?). Waste of time.
2. before vs after a certain date. Useful, given current economic circumstance.
3. "Direct" vs "Indirect" (No "induced" category). This presumably means we have to divide expenditures according to whether it is for value added by the primary contractor ("direct") vs inputs purchased by the primary contractor (indirect")? I can't see the point of this division.
4. Within province vs within Canada. Probably useful, since we want to see whether fiscal stimulus may be spread across the country according to where it's needed?
Posted by: Nick Rowe | March 18, 2009 at 10:18 AM
Nick,
I don't know if you saw this:
#
By Nick Rowe on Mar 18, 2009 | Reply
That’s interesting. We get the same story told in Canada. It’s not that the banks have stopped lending, but the non-bank (or direct market) source of loans is the one that has dried up, and banks are struggling to fill the gap.
#
By Don the libertarian Democrat on Mar 18, 2009 | Reply
John or Nick,
Could you take a look at E2.6 and tell me what’s going on between the large, small, and foreign banks?
Take care,
Don
#
By Darren Larson on Mar 18, 2009 | Reply
Don, I also noticed the difference between bank sizes. The rates are dramtically higher, the fixed period of rate is longer and the time since the commitment was made is shorter the smaller the bank. One would guess that commitments made by large banks 10.7 months ago (avg) are a lot different than those being made today. We shall see if the numbers change over time. Namely, large banks could start using different indexes, margins or installing floor rates.
On the other hand, the small bank numbers show that the commitment is 6.3 months on average. Plus, anything that contains a rate with a repricing interval greater than one year has a rate of 5.94% and these were less than a month ago in commitment time. These loans are also 97% secured and less than 8% are based on prime. In essense, prime is a useless index right now.
I would assume that large banks are attempting to swap out those rates or packaging the loans out, as making a profit at 3% interest has to be incredibly difficult. For the most part, small banks portfolio their loans, thus they need a profitable interest rate.
Posted by: Don the libertarian Democrat | March 18, 2009 at 06:01 PM
Nick wrote:
3. "Direct" vs "Indirect" (No "induced" category). This presumably means we have to divide expenditures according to whether it is for value added by the primary contractor ("direct") vs inputs purchased by the primary contractor (indirect")? I can't see the point of this division.
------------------------
Somebody is likely assuming that you will use an open-version input-output model to derive these direct and upstream numbers. Presumably you feed in a gross expenditure/output number and out pops direct GDP impact and indirect GDP impact. Or you get fancy, decompose the project by commodity-level entry and feed that in. Likely not worth the bother....
To add to your repertoire of uses, these impact studies were popular with senior defence brass who had a minimal understanding of economics but were seeking additional political support for budgets. Instead of seeking to rationalize defence expenditures and more specifically procurement, senior defence officers have decided to play ball with regional benefit-oriented politicians and support Industrial Benefits and politically directed domestic procurement contracts that ultimately contribute to less effective Canadian Armed forces.
Posted by: westslope | March 19, 2009 at 02:34 PM
I have quite a bit of experience with EIAs, and as far as I can tell, they are used mostly so governments can get a rough idea of the tax revenues that flow from a project. The consultants drafting them know they are BS, as do the people reviewing them. But, certain non-economic folks love to be able to state that some amount of expenditure will create some amount of jobs, dollars, etc.
If you've ever read Confessions of an Economic Hitman, you'll be able to relate what Perkins did for his firm to the discipline of EIAs. He'd pump up the multipliers and the numbers for some hydroelectric project and exaggerate the heck out of its impact on national growth. Governments lapped that stuff up, because it made it easier to sell the people, especially when the full costs wre not considered.
I'd say companies use EIAs for projects in order to sell it to the shareholders, then to governments, who get the next biggest cut of the economic pie. With all the huge dollar figures floating around in the debates, the public is left confused, but for the most part accepts it. (A billion dollars is better than nothing).
Posted by: Aaron | April 01, 2009 at 03:56 PM