This essay talks with a gas tax holiday that was proposed by Senator John McCain in April 2008 that never happened. Mostly it deals with the response to McCain's proposal by Princeton economist/New York Times columnist Paul Krugman. What it attempts to do is look at the real world evidence for the effect of taxes on pump prices of gasoline. I concluded that while McCain's proposal was naive and the idea was probably not a good one, he was more right than wrong about a gas tax holiday reducing pump prices a bit. I conclude that Krugman's response was more wrong than right. And that the discussions of the issue on the Internet, while extensive and heated, were not very helpful.
For what it's worth, I think that the result of a four month tax holiday would have been that the government would have foregone about 8.6 billion dollars in highway construction/maintenance funding; that consumers would have seen about a 12 cent (3%-4%) decrease in pump price and that gasoline retaillers, distributors, and producers would have divied up a 3 billion dollar (6 cents a gallon) windfall.
A more interesting and worthy subject I think is why economists failed to carefully examine professor Krugman's model and to evaluate what data is available. Should they not -- given time -- have found that the data, such as it is, suggests that Krugman's model is wrong and its answer incorrect? If they can't get this rather simple situation right -- and I'm pretty sure that they did not -- can we trust them to get anything right?
On April 15 2008, as gasoline prices soared, Republican presidential candidate John McCain proposed a Summer holiday from the 18.4 cent Federal gasoline taxes. New York Times columnist and Princeton Economics Professor Paul Krugman quickly countered that a tax holiday would not result in lower prices at the pump. The discussion quickly deteriorated into lay persons maintaining that Krugman was out of his mind, and economists maintaining politely that laypersons were ignorant and none too bright.
I've been meditating on this for quite some time and have come to the conclusion that there is something to be said for and against all sides of the discussion. In general, I don't think Senator McCain or the lay people understood the economist's arguments. On the other hand, I think that the economists did a poor job of explaining their arguments. And they were -- in any case -- most likely more wrong than right.
I should preface this by saying that the gas tax holiday probably was not an especially good idea. The amount involved was only 18.4 cents per gallon with gasoline prices over 15 times that amount. Moreover it seems likely that consumers wouldn't even have seen all of the 18.4 cents. And future funding for highway maintenance and improvement -- already diminished by reduced consumption of gasoline -- is directly tied to that 18.4 cents.
The short version is that this is an argument about who actually pays taxes and pays fewer of them if the tax is reduced. Most people would assume that a gasoline tax is paid by whoever buys the gasoline. But there is another possibility which is that the tax is actually paid by the producers and is built into the price of gasoline. That is to say that producers underprice their product by some amount to allow for the tax that consumers will pay. The difference is that economists are going to claim that the price of gasoline is not set by cost of materials plus cost of refining, plus costs of distribution, plus profits, plus taxes as many might assume. No, the price is set by inexorable market forces of supply and demand. Since the price is set by the market not by totaling costs, removing a cost (tax) puts the money saved into the pockets of the producer, not the consumer who will pay the same price no matter what. (Anyone who pointed out that gasoline supply and demand hadn't changed all that much in the past decade while pump prices in the US had trebled would presumably have been declared a to be a Philistine and likely would not be invited to any more economic discussions).
It's more complicated than that. It only works that way if the supply of gasoline is not sensitive to purchase price and the demand for the product is. There is a rather elaborate body of theory wrapped around all this dealing with subjects called "tax incidence" and "price elasticity". It is important to understand that economists actually believe this stuff. They don't pass Econ 101 if they don't at least act as if they believe.
The problem here is that there is really quite a lot of evidence that something is wrong with the Economists' formulation in this case. Perhaps gasoline supplies are more accommodating to price than is assumed. Perhaps demand is even less price sensitive than supply. Perhaps Econ 101 tax incidence theory is at least in part incorrect even though the derivation is plausible. Not mine to judge. The following factors can be identified.
Anyway, the economists answer looks to be partially or possibly completely wrong.
In essence, the argument revolves around a simple chart. According to the Wikipedia the Supply Demand chart was first published by Fleeming Jenkin in 1870. But it draws upon ideas that go back at least a century before that. It is said to be one of the pillars of microeconomics. The chart shows price on the vertical (y) axis and quantity on the horizontal (x) axis. Two lines (curves) are drawn. One represents price vs supply. The other is price vs demand. Since supplies of anything are expected to increase with increasing price, the Supply line usually slopes up to the right. Demand on the other hand usually decreases with increasing price so the Demand line slopes down to the right. Somewhere on the chart the supply and demand lines (probably) intersect. Pathological cases (e.g. lines don't intersect or intersect multiple times) exist but probably aren't relevant. The intersection point is called the equilibrium point. It is believed -- supported by some plausible logic and some experimental evidence -- that market forces will (or at least can) force market prices toward a price that may well be the equilibrium price.
Professor Krugman and, it would appear, economists in general, argued as follows: The supply of gasoline for the Summer season is fixed because US refining capacity is limited and much of the gasoline for the Summer driving season is refined ahead of time and stockpiled in inventory. The supply being fixed, the price must be determined entirely by demand. Since price is controlled only by demand, removing or adding a tax, can have no affect on the consumer price. Price will remain unchanged. (If that were true, why would prices be climbing to quite high levels in response to changes in raw materials cost? Supply is, we are told, fixed. Somehow demand is driven not by price but by raw material cost?) In the case of a gas tax holiday, the 18.4 cents a gallon would be split in some fashion between refiners, distributors and retailers. Consumers would see none of it. (Parenthetically, this is a very large logical jump especially if tax changes are large.)
In his initial article Krugman tells us that this is "Econ 101 tax incidence theory". And I think that if Krugman and the Economists (herein after dubbed KATE) got themselves into trouble this is where they did it. They turned the debate into a debate about tax incidence theory which they possibly understand and lay people don't. In many cases, that might not be a problem. Unfortunately, petroleum pricing and taxation look not to be an example you would pick to illustrate tax incidence if you were teaching Econ 101.
Tax incidence is the concept that the person, place, or thing that is held accountable for paying a tax is not necessarily the entity that comes up with the money. The nominal taxpayer will, if he she or it can, lay off some or all of the actual cost of the tax onto whoever they can stick with it. Ultimately, the burden of the tax will fall on those least able to pass the cost on to someone else.
The conventional example is that in the US, employers and employees nominally each pay half of the cost of Social Security program contributions. Many economists believe that the incidence of the Social Security taxes falls mostly on employees because the employer taxes are money that would otherwise go to the employees. However, a few souls believe that the employees would see none of that money if the tax were eliminated and therefore believe that the incidence in this case falls equally on employers and employees.
That illustrates a problem with tax incidence. It is not necessarily a rigorous concept. People often do not agree what the incidence of a given tax is. http://web.archive.org/web/20090206080146/http://are.berkeley.edu/~cavazos/teaching/eep1/slopeelasticity.pdf
All economists are taught, and many seem to believe, that there is a rigorous way to compute tax incidence in some cases. It depends on yet another Econ 101 concept -- Price Elasticity. Price Elasticities are the ratios of product price changes to corresponding changes in supply or demand. Those who took and remember any Analytic Geometry, will observe that the price elasticity looks a lot like the slope of the supply or demand line (or its tangent if the "line" is a curve). Strictly speaking, it's not because elasticity is normalized to remove the dimension of price/quantity, but in practice the distinction is often unimportant and sometimes ignored. If things are inelastic, changes in price have little affect on supply or demand. If, on the contrary, they are elastic, then changes in price strongly affect the quantity in question. Near horizontal lines indicate high elasticity. Near vertical lines indicate low elasticity.
Economists hypothesize that tax incidence will reflect the relative elasticities of supply and demand (at the equilibrium point). If the elasticities are similar, the tax will be split fairly evenly between producers and consumers. If the elasticities differ, tax incidence will fall more heavily on the less elastic of supply or demand. We can determine the tax incidence by examining the ratio of supply and demand elasticities. This hypothesis is neither crazy nor stupid. The basic idea is that taxation will affect supply and demand. To the extent that the producer/consumer can't alter production or consumption to avoid the tax, they will simply have to pay up. There is a detailed description of how all this purportedly works at http://rri.wvu.edu/WebBook/Garrett/chapterthree.htm. If a 10% tax is imposed on widgets, then widget manufacturers will desire to increase their prices by 10%. But that would (probably) decrease widget sales. So the market will then seek out a new equilibrium with manufacturers charging more than they used to and consumers paying more. But the increase in price will only be the full 10% if consumer demand is not affected at all by the increased price -- i.e. is totally inelastic.
1.0 I find the logic of the assumed relationship between tax incidence and price elasticity to be somewhat uncompelling. Further, it appears to me that economists are prone to assume those parts of the supply/demand model that support their current thesis to be fixed then to push the remaining elements around the chart to wherever they need to be in order to "prove" their argument. That's possibly unfair to them. I may just be too dumb to understand.
2.0 While most things that are true are also logical, the converse is not necessarily true. That's why scientists are much happier when experiments confirm theory. I can not find any evidence that any of this body of economic theory has ever been validated by experiments with controlled variables or broad, credible analyses of real data that indicates that tax incidence and/or elasticity actually work as modeled. I did find some experimental support for convergence of prices to a price that might well be the equilibrium price. That's about it.
3.0 I have not come across an explanation of how supply-demand curve pricing is tied to the common assumption that prices are at least influenced by the cost of raw materials, processing, distribution, overhead, and some profit. Neither is it obvious where to look on a supply-demand chart for the effects of, for example, competition.
At this point, pragmatics enter the picture. First of all, Krugman's assumption that gasoline supply is totally inflexible, is not entirely true. In fact, there is a large inventory buffer of refined gasoline that can be and is drawn down somewhat during the Summer depending on demand. Within limits, additional gasoline can be and is imported during the Summer. Since storing inventory over the Winter costs money, refiners juggle refining, imports and inventory in order to enter the Winter with a minimum but adequate amount of inventoried gasoline. Finally, a significant amount of Ethanol production was coming on line in 2008. In point of fact, the EIA estimated a week before Krugman's editorial that -- based on slightly reduced demand, high inventories, and increased Ethanol production -- US refining would probably be down around 150000 barrels a day (about 2%) in the Summer of 2008 from levels in the Summer of 2007.
US Gasoline supply is an extremely complex subject and apparently one where those of us who are unfamiliar with the industry should tread lightly. Krugman apparently didn't tread at all. He simply made an assumption. My best guess is that his assumption was basically wrong. There is a constrained supply of gasoline. But not exactly a fixed supply. Probably not fixed enough for the logic that was built upon it at any rate. Neither the economists nor the non-economists spotted this at the time.
Second, gasoline demand is notoriously inelastic with regard to price. Between May 2007 and May 2008, it appears to have taken a 25% increase in the pump price of gasoline to achieve less than 2% reduction in demand. This suggests a very steep demand curve. At least that is what it suggests to me. On the other hand, starting in the Spring of 2008, US consumers actually did start using significantly fewer refined petroleum products. That would seem to indicate that at high enough prices, gasoline demand does become elastic. What that suggests is a curve, not a straight line. And a curve with poorly defined characteristics at that.
Even if we believe that a meaningful tax incidence can be accurately computed from supply and demand lines (and I, at least, am not completely convinced), we will end up trying to allocate tax incidence from two very steep and not very well defined lines. Pragmatically, that's unlikely to yield useful results. So, the correct answer would seem to be that we have no firm idea on theoretical grounds how much of a gasoline tax holiday would actually show up at the pump. It is perhaps not surprising that lay people failed to realize this. One thinks economists could, and should, have done better.
The proposed gas tax holiday involves quite a small change percentage-wise in gasoline price. Large tax changes almost certainly do not behave as KATE predicts. If a seven dollar a gallon gas tax were imposed in the US, I think that everyone including KATE would expect the price of gasoline to jump by seven dollars a gallon the day the tax went into effect. Those, including many economists, who advocate higher gasoline taxes to discourage consumption are counting on that. In econospeak, the incidence of large tax changes for gasoline falls (almost?) entirely on the consumer. If it did not, the supplier would be expected to manufacture and distribute gasoline at a (huge) loss.
Here we have some actual data. Countries with high gasoline taxes tend to have high pump prices for gasoline. In March 2008 the pump price for gasoline in Norway was $8.73 a gallon. Great Britain -- $8.38. Canada $5.35. Taxes were 170%(2007), 117%(2007) and about 33%(2007) respectively. All three countries are oil exporters so the higher prices probably aren't somehow caused by the process of importing petroleum. They are surely caused in large part by taxes. US gasoline taxes at a 2007 price of $3.00 per gallon would be about 17%. US and Canadian gasoline tax rates vary substantially from place to place. Those used above are typical not universal.
There is an assumption that tax incidence works the same in both directions (i.e. is reversible). Imposing then removing an 18 cent tax will bring things directly back to their starting point. Maybe. A lot of physical systems do not behave like that. Irreversible changes like baking a cake are quite common. Unbaking a cake isn't necessarily forbidden by the underlying state equations. But it doesn't work. Hysteresis -- taking a different path down than was taken up -- is also common. Quite a few devices in common use -- household heating systems or mechanical timers for example -- depend on it. I'm not going anywhere with this. Just mentioning it for the perusal of anyone who feels that tax incidence is too simple or well understood to yield wrong answers. For an example of a simple sounding problem in reversal that is virtually impossible to analyze, see the Wikipedia on the Feynman sprinkler
Oh yes, and looking over the Econ 101 level information on tax incidence on the Internet, I don't find much indication that Econ 101 students are told that tax incidence can vary with the size of the tax. If it can't (in Econ 101 land) the problem is solved. The large tax increase case demonstrates that tax incidence for gasoline falls (almost?) entirely on the consumer. McCain is right. Krugman is wrong. I'm not going anywhere with this either, because redrawing supply as a curve rather than a straight line can resolve the large tax change issue while leaving the possibility of different incidence for small taxes. And that seems intuitively to be plausible.
I have prepared charts which show some variations on gasoline supply-demand charts:
(Parenthetically, I think there is probably more to be learned from these charts than I have learned and that they could be better and show more if only I were smarter).
What Chart D makes clear is that even small errors in the shape of the curves or their placement in the quantity-price plane are likely to make quite significant changes in tax incidence. Most likely, we have only the haziest idea what the actual impact of a tax holiday would be.
No that's not all. There are several relevant cases where some information is available on how tax incidence works for small changes in gasoline taxes. One of these actually has been formally studied. One person (ONE) mentioned that during the gas tax holiday debate. [https://web.archive.org/web/20150630024937/http://www.williampolley.com/blog/archives/2008/05/gas_tax_holiday.html] Probably others posted similar or related information in venues I didn't read and didn't encounter via Google. But the number seems to have been small.
State gasoline taxes vary substantially from state to state. If KATE's logic prevailed, we would expect that pump prices would reflect demand, but not taxation. Two similar counties in adjacent states with different taxes would have virtually identical pump prices. In fact, maps of pump price by county can be found on the Internet. It is relatively easy to identify high gasoline tax states on those maps which are color coded by price. Prices tend to be higher in high gasoline tax states. If you plan to check this, be warned that a number of states have pump price based sales taxes on gasoline as well as or instead of per gallon taxes. You want to use Spring 2008 data with taxes that represent Spring 2008 pump prices not those of, for example, 2002.
Would it be possible to make a guess at gas tax incidence based on pump prices in jurisdictions with differing tax rates? Probably. Unfortunately, pump prices vary quite a bit from pump to pump in the same jurisdiction, so deriving an average pump price for each jurisdiction would be a rather daunting task. Add in other factors that require correction such as prices often being higher at stations located at Interstate Highway off ramps and you have the material for a serious study. In the one situation where I can actually observe the difference, it appears to me that the difference in pump prices between Vermont and New York is about ten cents a gallon -- which happens to be the difference in state tax rates between the two states. But that may be coincidence or observation error.
In 2006-8, a number of counties in upstate New York experimented with "capping" (reducing) gas taxes in order to lower prices.
www. co.oswego.ny.us/info/column/2008%20columns/070308. html It is asserted that the reduced tax had no affect on gas prices at the pump. I can find no numeric data to prove or disprove that statement although there are claims that data exists. I am inclined to discount these claims somewhat because pump price maps show that some of the counties (e.g. Oswego) where gas caps are claimed not to have reduced prices seem to have slightly lower prices than nearby counties.
The New York gas tax caps are not, as it turns out, the only state gas tax holidays, reductions, caps ever. The problem is that the tax tinkering generally involves small amounts -- a few cents a gallon -- at times when prices are rather volatile. That makes interpreting results kind of difficult. There have been a couple of academic studies of past tax tinkering that conclude that 50%-60% of the tax cut in the cases studied is passed through to consumers. See [https://web.archive.org/web/20150630024937/http://www.williampolley.com/blog/archives/2008/05/gas_tax_holiday.html].
In many states, gas stations on tribal land are exempt from state taxes although there may be tribal taxes in some cases. I'm not entirely clear on how this works. I believe that Indian communities are sovereign governments that are exempt from state taxes(?), but in some cases they may have traded voluntary compliance with state taxation for other concessions(?). Anyway, we potentially have taxed and untaxed gas being sold in neighboring stations at times. Prices of gasoline on reservations are said to be lower than those in adjacent non-reservation areas -- by up to 25 cents a gallon. I found plenty of anecdotal evidence that is the case. And I believe it to be the case. But I do not have a tidy chart showing comparable reservation and non-reservation gas prices. Neither have I been able to find pairs on the Internet of current reservation, non-reservation pricing in likely places like the St Regis Mohawk Reservation or the Navajo Reservation. I did look, honest.
Demand curves are different on Indian land?
There were actual gasoline shortages in the US in 1973 due to the Arab Oil Embargo and again in the early 1980s due to the Iranian Revolution. That certainly would seem to meet Professor Krugman's assumption of fixed supply. Reasoning from Supply-Demand charts, one would expect that gasoline prices would rise to whatever level is necessary to exactly match the available supply. Is that what happened? No. Prices did rise substantially, but actual demand management was provided by odd-even fill up days and by stations simply running out of gas to sell. As I write this, the newspapers assure me that there are severe gasoline shortages in parts of the South due to refinery shutdowns and damage from Hurricane Ike. These are reflected to some extent on gasoline price maps, but even though the worst shortages are said to be in Tennessee, prices are highest in Michigan, Indiana, and Georgia
This argument may be somewhat tainted during the oil shortages due to the imposition of odd-even day purchase rationing. It nonetheless appears to me that "price is affected by supply" is a more accurate statement than "price is dictated by supply"
The right answer would appear to be that a gas tax holiday would result in a reduction of gas prices at the pump of no more than 18.4 cents a gallon and possibly much less. The difference between the magnitude of the tax and the amount returned to the consumer would go to the (huge, malicious, evil,etc ...) oil industry. We probably lack sufficient information to predict exactly how the benefits from the tax holiday would be split. Academic studies of past situations say 50-60% to consumers. But that was at somewhat lower real gasoline prices and there is some reason to think that the incidence will fall more heavily on consumers if gasoline prices are high. If I had to guess, I'd say about 35% to the producers, 65% to the consumers.
Let me repeat that the idea of a gas tax holiday was probably a poor one. Revenue to fund highway work is foregone and the benefit probably goes in part to companies that probably don't need the money. However, John McCain probably didn't know that. He's an airplane driver who knows a whole lot about campaign finance reform. He should have asked someone about it. If he didn't -- his fault. If he did, it's by no means clear he would have gotten a straight, accurate answer -- not his fault.
Let me also let Paul Krugman off the hook to some extent. He was functioning as a commentator, not an economist. Commentators do broad analyses of current issues. They need answers now, not four months from now. Krugman consulted a leading economist -- himself -- who gave him a quick, plausible, superficial answer. Not Krugman, the economist's, best work I think. But it is not clear that Krugman, the columnist, would have gotten a better answer from some other economist.
As for economists. With the notable exception of [https://web.archive.org/web/20150630024937/http://www.williampolley.com/blog/archives/2008/05/gas_tax_holiday.html] the economists I read on this subject devoted all their time to supporting Krugman's probably faulty analysis with little or no apparent thought and analysis. That's not very helpful to us non-economists. It is all very well to explain tax incidence to us unwashed bozos. But it'd be better to get it right, eh?
Gasoline pump pricing in the presence of small tax changes in a supply constrained marketplace appears to be fairly complex. It is fairly difficult to confidently estimate the pump price affects of temporarily suspending the 18.4 cent federal gasoline tax. The most probable result would seem to be that federal revenues would be reduced by 18.4 cents a gallon; that consumer prices at the pump would drop by something less than 18.4 cents a gallon; and that gasoline producers/distributors/retailers would receive a modest windfall. This is probably not the result expected by John McCain nor is it the different result predicted by Paul Krugman. The lay people and economists who offered opinions don't seem to have guessed very well either for the most part. The extensive discussion on the internet was for the most part curiously deficient in insight and failed to converge on a better answer although a better answer probably exists and should not, one thinks, have been enormously difficult to get to.
Experimental verification of equilibrium pricing.
Analysis of tax incidence
Copyright 2006-2012 Donald Kenney (Donald.Kenney@GMail.com). Unless otherwise stated, permission is hereby granted to use any materials on these pages under the Creative Commons License V2.5.
This page will have been validated as Valid HTML 4.01 Transitional prior to posting on the web site. W3C Logo Image omitted.