« Daylight Savings Time and the non-neutrality of money | Main | What Policies Should a Canadian Economists Party Promote? »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Why assume rising marginal cost? I know it is standard in econ, but the real world suggests that due to upfront investment costs, most of the time we are looking at firms facing falling costs as production increases or at least flat costs.

Monopolist competition doesn't answer the question of why you perceive excess supply everywhere. Why is ex post demand greater than ex ante? If they overestimate instead, you'll get excess demand and firms wanting to lower their price. You need to combine your story with the story where firms systematically underestimate demand in order to see excess supply everywhere.

Long story short: your story violates rational expectations. Which is ok, but clearly the important mechanism in fitting your model to data.

I raised the same point as Marc on the Dorman thread, but received no notice. With increasing returns, whether to scale, as in classic heavy industry, or due to other factors, such as network effects and IP, with IT or pharma, MC is decreasing with output for the most relevant portion of the curve. And that condition would apply to some, though not all, sectors, but likely many important ones. This actually works in reverse during a demand collapse/depression, since falling demand and thus lowered output means increasing unit costs, and thus a need to hold or raise prices in the face of falling demand, (which is part of why price adjustment in a depression can't occur, pace neo-classical equilibrium claims). I get the point that, macro-economically speaking, it's the limiting case for real and potential output that's sought for. But why go at that with an unrealistic stylization of a single type of case as "representative" for the whole structure of the economy. Wouldn't a mixed population of cases, in terms of sectors and firms, provide a better "foundation"?

Also, one other point about oligopolistic competition: the maintenance of excess capacity provides a signaling device vis-a-vis similar competitors, always threatening a price war, so that price wars tend not to occur.

Also, the further implication of oligopolistic competition between large firms is that pricing is largely administrative and manipulated, rather than strictly determined by competitive market pressures, (and the role of market-dominant rents needs to be brought in here). Costs will be attributed across a wide variety of operations for various accounting purposes, (rather than simply reflecting unknowable marginal products), and prices will be differentiated to appeal to different segments of potential market demand.

Offtopic, but Nick, you should be blogging this:
(Kocherlakota explains why we should abstract from monetary policy and consider fiscal policy after housing bubble has crashed).

Marc and john: You're both talking about average costs, which can be falling even while marginal costs are increasing.

I believe this is why a monetary shock that pushes us to the left of the natural rate of output seems to be more destructive than a monetary shock that pushes us to the right of the natural rate, even though it really isn't more destructive in the long run.

Good clarification, Pedro, most of the time firms in the real world are facing falling average costs. But why do economists assume rising marginal costs when flat marginal cost would be more appropriate? What is the supply constraint that is pushing up those costs -- seems like those examples are the exception rather than the rule.

I think when you try to emulate the real world the modeling does not deliver closure, so it is easier to make unrealistic assumptions and give an "answer". But how useful is that?

If MC is flat or falling, firms should be bigger. They increase in size until their MC is again increasing. Even if the MC is non-monotonic (and we have multiple equilibria), it is then locally true that MC is generally rising at every equilibrium.


Correct me if I've got the logic or math wrong, but it seems pretty simple. Average unit costs can decline only if marginal (i.e. each additional unit) costs are initially declining, though average unit costs will decline with a lag and at a slower rate than marginal unit costs. The case of rising marginal costs with falling average costs would likely occur only at the tail end of the curve, which may never be reached. Of course, this question is always a matter of firm-and-sector-specific cost structures, when upfront long-run fixed investment is the largest factor or consideration. But then there is a further implication here: with such cases, and their uncertainties, market-clearing and thus profit-maximizing assumptions don't apply "uniquely", since such notions themselves become uncertain and unclear. Rather firms in such cases tend toward investment geared toward the strategic pursuit and lock-in of rents from technical productive efficiencies and market-dominant positions, which rents or quasi-rents help defray the costs and uncertainties of high long-run fixed investments. Which is something that would seem to have large implications for actual real-world corporate investment behavior.

"Even if the MC is non-monotonic (and we have multiple equilibria), it is then locally true that MC is generally rising at every equilibrium."

Why? Some firms are creating new markets with increasing returns to scale and falling MC while others are exploiting mature markets with rising MC. There is no a priori reason why the average over the whole market should exhibit rising MC. In fact, this would require irrational behavior on the part of investors. You would imagine that there would be some level of indifference between the two, so that firms would not continue to invest if their profit margins fell below a certain value, since it would be better to deploy resources to create new markets instead.

"If MC is flat or falling, firms should be bigger."

Indeed. I believe this is called "investment", and as far as I am aware, it is still on-going.

Most firms are getting bigger, but you are right, there are some cash-flow pigs that have saturated their market and just pay dividends. I think Warren Buffet gave the famous example of See's Candy versus Geico, and how a good investor should have both types of firms in his portfolio.

Excess supply: See the interesting discussion on Fora, with John Ralston Saul discussing the world economy in his usual dry humour. [ http://bit.ly/dzpeEM ] segment 12, "Capitalism Fails In An Era of Surplus"

I see excess supply everywhere I go, because for years I have been a happy participant in Freecycle, a web-based giveaway group with some 7 million members worldwide.

The amount of stuff given away even in my small city is stunning, and it is all excess to the needs of the giver, but valuable to the taker, and too good to throw away. In an everage week our 5,000 members compete for a couple of hundred giveaways at least.

And then there are the secondhand stores. Since the 80s, they have expanded and gentrified and reached upward into the middle class for their customer base. They are flooded with stuff that people give them -- box it up, drive there, and give it to them just to get the damn stuff out from underfoot.



John and RSJ: Assumptions about rising MC and decreasing AC are generally used to model short-run equilibria. Imagine that firms have a fixed amount of buildings and equipment in use. As long as the relevant market is big enough, they'll have chosen this fixed level, as well as their price, such that they expect marginal revenue to be equal to marginal cost. If demand is higher than expected, they'll have to hire more labour than expected at the last minute. 1) They may have to get more hours from their current employees - overtime. 2) They may have to hire more workers and train them. 3) This extra labour will be using buildings and equipment that were ideal for the previous level of labour, which means the 'productivity' of this extra labour may be lower than the previous.

Investment is generally used to accomodate higher demand in the longer run. After expansion, a firm may indeed produce at lower cost.

My benchmark assumption is that the individual firm's MC curve will be roughly horizontal, in the long run. It will probably slope up in the short run, because it will be hard to double all inputs and outputs overnight. The downward sloping MR curve is what normally puts a limit on the size.

I drew the MC curve sloping up in my diagram because it *might* slope up (and probably does in the SR), and if it does slope up there will be a quantity Qmax, and I wanted to show Qmax in my diagram. Mine is a short run analysis in any case, given that the firm's price is set only in the short run. In the long run, it may change price if demand shifts (depending on the slope of MC, and on whether the elasticity of demand changes when the demand curve shifts).

Andy: I don't think my story violates rational expectations. On average the demand curve will be where my ex ante demand curve is. Half the time the actual demand curve will be higher than expected, and half the time it will be lower than expected. If demand is lower than expected, Q will fall below Q*, both in perfect and in monopolistic competition. If demand is higher than expected, the perfectly competitive firm will not increase output; it will ration sales to Q*. But the monopolistically competitive firm will increase output past Q*. It won't ration sales unless demand increases a lot, past Qmax (if Qmax exists). So I only talked about unexpected increases in demand, not unexpected decreases, because the latter are boring, since there's no difference between perfect and imperfect competition.

Scott: exactly. Increases in output past the natural rate (while they last) are seen as good times, because they *are* good times. The economy gets closer to the optimal allocation of resources. Small increases in Q have first-order effects on welfare, because Q* is less than optimal. In perfect competition, on the other hand, where Q* is optimal, small increases or decreases in Q worsen welfare, but only have a second order of smalls effect.


Sorry, but I'm not convinced. And for a couple of reasons. First, sometimes firms are happy to sell more than they planned because what they are selling was designed as a loss leader in order to get you through the door. They really don't care if you buy the sale item as long as you make some other impulse buy or decide to upgrade (think buying a PC). In other words, firms don't just sell one product. Demand for other products sold by the firm is not determined independently of other items in the firm's catalog. Second, you've implicitly assumed that the firm can in fact increase the quantity on-the-shelf in something like real time and at something like a constant marginal cost. This is clearly wrong. In order to meet demand a firm either has to buy or manufacture more product prior to a sale, in which case it incurs holding costs for supply > demand; or the firm must incur additional administrative and startup costs for any expected sale quantity (Q* in your example). This means there will be at least a step increase in MC at Q* and possibly an increase in the slope of the curve because it is likely that intermediate suppliers, sensing an increase in demand, will charge more for their intermediate goods. Now here's where it gets tricky. In the real world most firms, even those with significant monopoly powers, stock more on the shelf than the mean estimate of demand applied to a simple Wilson EOQ. The Wilson EOQ is only valid if demand is known (as opposed to estimated) and production leadtimes are known and if the customer only buys one each (i.e., no reorder point undershoot). None of those conditions apply in the real world, so firms will always have an operating level that is greater than mean demand (which is Q* in your example). In addition, firms will likely maintain some safety level targeted to protect against some level of customer ill-will. Take all these factors into account and the firm will almost always have more than Q* on the shelf at price P*, which means the firm will always welcome additional sales to eat up the stochastic component of total demand that is beyond Q*.

2slugbaits: If the short-run MC curve is locally vertical at Q*, then indeed the monopolistically competitive firm will not increase Q when demand unexpectedly increases. I just don't think it is vertical (for most firms).

"Now here's where it gets tricky. In the real world most firms, even those with significant monopoly powers, stock more on the shelf than the mean estimate of demand applied to a simple Wilson EOQ."

I would say that it is precisely because firms face downward-sloping demand curves, and so P exceeds MC at the expected equilibrium, that firms stock more inventory than they expect to sell, and so can handle an unexpected increase in demand. Since inventory is costly, you wouldn't hold excess inventory if P=MC.

Nick: I agree with what your comment and that a ratex firm can overshoot sometimes and undershoot other times.

My point was that the model where they miss only 50% of the time (and most of those misses are not very big) does not explain the phenomenon you initially described, namely "my lying eyes kept telling me that most firms, most of the time, really wanted to sell more than they were able to sell."

Plus there's the issue of how big these forecast errors would need to be to generate what your lying eyes tell you, and what form these shocks take.

Andy: I don't get it.

Assume the distribution of demand shocks is symmetric. Q* is at the 50 percentile point. But Qmax will be well above the 50 percentile point.

In my model, once the price has been set, firms will be willing to increase output right up to Qmax, if actual demand turns out to be big enough. It's only if there's a *very big* unexpected increase in demand, so that quantity demanded exceeds Qmax, that firms will not accommodate any further increase in demand. If Qmax is at (say) the 90 percentile point in the distribution of demand shocks, then 90% of the time there will be excess supply.

And only 10% of the time will there be excess demand.


Here's where we disagree. You are assuming that the firm will expand production to meet the unanticipated demand only after it's clear that actual demand will be greater than expected at price P*. I think that's wrong because the marginal cost curve after point Q* will be kinked due to transaction costs of having to place another production run. The upward shift in prices will be approximately equal to distance from the origin to the intercept of the MC curve. It's like the classic "newsboy" problem of a single period inventory decision. I think that a better explanation of how a firm will react is that the firm will start out the sale period with more than Q* on the shelf. The additional amount of Q* on the shelf will be driven by forecast error in estimating the demand curve, the transaction costs of having to place emergency production runs, the holding costs of excess inventory, and disposal values of excess inventory.

In the real world firms have inventory models that will tell them to stock more than Q* given that Q* is the expected quantity given price P*. Firms do not increase production after they realize sales will be greater than Q* at price P*; firms anticipate that sales could be greater than Q* and they increase their initial production run quantity accordingly. It's because firms commit to having more than Q* on the shelf at the beginning of the sale period that explains why they are so happy to see actual sales exceed Q*.


OK. My simple little model has no inventories. Production and sales are the same thing. I can explain excess supply without needing to put inventories in the model.

If I did put inventories in my monopolistically competitive model, (If I assumed you cannot increase production to cope with an unexpected increase in demand), firms in my model would choose to hold excess inventories. They would stock more inventories than they would choose under certainty, and with a symmettric distribution of demand shocks, would have unsold inventory more than 50% of the time).

Suppose we assume perfect competition instead. Firms face horizontal demand curves, and have upward-sloping MC curves. My conjecture is that in this case firms would *not* choose to hold excess inventories. With a symmetric distribution of demand shocks, and a linear upward-sloping MC curve, my guess is that 50% of the time firms would run out of inventories (excess demand), and the other 50% of the time firms would have unsold inventories (excess supply).

In other words, "It's because firms commit to having more than Q* on the shelf at the beginning of the sale period that explains why they are so happy to see actual sales exceed Q*." can be true, but they would only choose to have more than Q* on the shelf under monopolistic competition, not under perfect competition.


In other words, "It's because firms commit to having more than Q* on the shelf at the beginning of the sale period that explains why they are so happy to see actual sales exceed Q*." can be true, but they would only choose to have more than Q* on the shelf under monopolistic competition, not under perfect competition.

Yes. Now we're in agreement. I agree with your central premise that happy salespeople in the face of seeing unexpected customers is a symptom of monopolistic competition. I just didn't find the explanation of instantaneous expansion of production at fixed marginal cost very convincing. This leads to some interesting speculation as to how monopolistically competitive firms are likely to react to better inventory models with better forecasting techniques. We may end up with some unanticipated consequences in how it affects the business cycle.

The problem with the model is that it is fundamentally wrong, for at least any industry in which I managed supply and price profitability decisions for over 25 years. And they were all non-monopolistic.

Fixed costs or investments tend to be large but come in step functions. In capital intensive industries the step function are large. The gains in supply capacity potential tends to be oversized because of the nature of the investment, lead times.

Marginal costs and marginal supply capacity tend to be more agile, more adjustable. The marginal supply also tends to be a smaller input to the overall supply management. This obviously varies with the amount of labor input to capacity.

But my initial reaction is that this model is oversimplied and loses contact with reality because of this compression.


Two points and a technical comment.

1. What evidence do we have that Qmax is well above Q*? It certainly isn't in your model. What is the distribution of shocks and how many generate this excess supply? It is hard to map your model into the world without answering this question.

2. Even once we get that modeling worked out, my original complaint was about the motivation. In a ratex world, I just do not expect many firms to have the problem where they wish to raise prices but cannot. I certainly don't know why that would be more common than the situation where they wish to lower prices but cannot. My reading of your motivation was that everyone sees excess supply and firms that want to raise prices everywhere, which then motivated this model. The only way to reconcile your empirical claim and the model is to assume that positive demand shocks occur WAY more frequently than other shocks, a ratex violation.

(As a technical aside, I'm 99% sure that the ex ante demand curve does not mean that firms plan to see the 50th percentile of demand shocks. Such a statement would rely on both the symmetry of shocks and the symmetry of the loss function. The second of these is almost certainly not true - firms should know about this excess supply problem you're describing and adjust accordingly. Why do I bring this up? The point is that the right way, IMO, to solve a firm's decision problem in this model is to make sure you've allowed their optimization to include how the world responds to big positive demand shocks. I don't think you've done that.)

I was going to post a comment earlier about step increases in fixed costs, but I see Jess already brought it up. However, allow me to take the marketing approach to a similar problem of being a monopoly supplier of some good. And I want to maximize profit.

First off - put all fixed costs aside. And for a certain range of production - assume marginal costs are flat. So, Price, P, can be easily translated to contribution margin (P-MC) by moving the line lower.

Now, I have to assume a demand curve. Assuming I have reliable market survey data, I can guesstimate the slope of the line. And depending upon where I am on that line, I can determine total contribution (P-MC)xQ, the area of the rectangle. Three cases to consider:

A)If the slope of the demand line is 45 degrees (I'll leave out the negatives), (1:1) then it matters not where I am on the line, the size of the rectangle will be the same. And hence the total contribution.

B) If the slope of the line is steeper, say 2:1 (rise over run) - a 2% increase of price only results in 1% loss in volume, then I am gaining in area of my rectangle, so my monopolistic profits go up by charging max. price. Keep production below capacity within the production range under consideration, higher up the curve.

C)On the other hand, if the slope of the demand line is flatter, say 1:2, then for every 1% decrease in price, I gain 2% in volume, it's better to run full out within the range under consideration.

So, after all of my marketing info, I find I underestimated demand at the price I picked. What to do?

Well, depends where you are. If you had assumed B) you have overcapacity - so the demand curve is flatter than you expected, or it has the same slope, but intercepts higher up on the P scale. So, increase production.

If you had assumed C) then you are already at capacity. Nothing to do in the short term.

If you had assumed B) - then you may or may not have overcapacity, depending upon what price point you initially picked.

And if marginal costs rise in a linear fashion, then they can be removed by changing the slope of the demand line. Same analysis applies.

Once you reach capacity - the investment decision (step increases in fixed costs) comes into play.

Anyway, how I'd look at it - from a marketing perspective, on a very simplified level.

2slugbaits: yes. We are basically in agreement. I was just now trying to work out a simple model of optimal inventories of fresh fish/bread (you have to throw it away if you don't sell it) and showing that the expected unsold inventory would increase as the gap between P and MR got bigger, but I can't quite get my head around a simple exposition. I see your model as a very short run version of my model, when unexpected extra sales have to come from inventory, not production.

Jesse: all models are oversimplified, from the perspective of someone living and working in the model. But if you want to make my MC curve look like a flight of stairs, that's OK. It won't affect the model much.

Andy: and I'm 99% sure that your technical aside is true too. With linear MC and MR the loss function should be symmetric though. And with symmetric shocks, and a risk-neutral firm, I think Q* would be on the 50 percentile (as long as we never hit Qmax?)

"What evidence do we have that Qmax is well above Q*? It certainly isn't in your model."

I needed to get both Q* and Qmax on the same page!

My hunch is that we rarely hit Qmax. Why? Because I rarely observe firms turning customers away. Though there was a line-up at Tim Horton's this morning, and I saw one customer leave rather than join the line. But maybe Tim Horton's has a fairly elastic demand curve, so P is close to expected MC at Tim's. If I were modelling Tim Horton's properly, I would treat excess staff rather like 2slugbait wants to treat inventory.

"Even once we get that modeling worked out, my original complaint was about the motivation. In a ratex world, I just do not expect many firms to have the problem where they wish to raise prices but cannot."

Tim Horton's does not raise the price of coffee and doughnuts when there's a rush. I don't explain why, but it doesn't.

"I certainly don't know why that would be more common than the situation where they wish to lower prices but cannot."

I'm saying it *isn't* more common. I'm saying they are equally common (at least, with a symmetric loss function, symmetric distribution of shocks, and risk-neutral firm).

JVFM: It's the *elasticity* of demand that matters, not slope.

If the demand curve is unit elastic (a 1% change in P leads to a 1% change in Qd), then Total Revenue (TR=PxQ) is the same at all points on the demand curve. It's a rectangular hyperbola, not a negative slope of -1. If MC is zero, such a firm would be indifferent about what price it set. If MC were positive, it would set an infinite price to maximise profits.

If MC is positive, the elasticity of demand must be greater than one at the profit-maximising equilibrium.


Full capacity is the quantity that can be produced *before* you need to raise prices. We have data on this, and the "average" capacity utilization is about 75%. In terms of U shaped cost curves, the upward sloping part of the U is only when you hit 100% capacity utilization.

For the vast majority of firms, short run marginal costs are declining up until the right hand part of the U, which is rarely reached. When it does reach full capacity, marginal costs go up very steeply -- overtime is time and half, weekend shifts are double time, etc. That's a steep jump and firms want to avoid making that jump.

Imagine a production chain in which every firm needs to buy inputs for other firms, and every firm is a price taker for its inputs.

Now, assume that at any time, capacity utilization is distributed somewhat randomly among these firms.

Then long before you hit 100% capacity for the economy as a whole, some firms will be at full capacity already, and those will then raise input prices for other firms, and that shows up as an increase in the short run *average* cost for everyone else using that input. It doesn't hit the marginal cost for each firm because it has nothing to do with how much that firm produces -- i.e. the micro firm cannot scale back production and have its costs per unit produced lowered.

So given that there is a production chain, and that around 3/4 of all sales are business to business, the average firm primarily operates in a situation of declining short run marginal costs, and is limited by how much it can sell, rather than by rising marginal costs. This must be the case, unless you assume that each firm has identical capacity utilization rates. You need another mechanism to determine price and quantity sold.

JVFM: It's the *elasticity* of demand that matters, not slope.

If the demand curve is unit elastic (a 1% change in P leads to a 1% change in Qd), then Total Revenue (TR=PxQ) is the same at all points on the demand curve. It's a rectangular hyperbola, not a negative slope of -1.

I knew I'd get into trouble not using the economists "elasticity". Let me try to restate my argument.

Case A): Total Contribution = (Price - margin) x Q = constant. It is the same at all *assumed* Q and P. You call this unit elastic. I did not say the Total Contribution (if I was to plot it) had a negative slope of -1.

In your example, you are assuming a particular P vs Q demand line. Neglecting an increasing marginal cost line, if the slope is -1, then unit elastic. A steeper line, then less elastic. If flatter slope, more elastic.

The point that I was trying to make is that, depending upon whether the company had pre-guessed that the demand curve was less elastic, unit elastic, or more elastic would determine where they were currently operating. Less elastic - less than capacity. More elastic - going full out. Unit elastic - it depends.

And so how they respond to overdemand depends upon where they figured was the optimal place to be to begin with.

The way you randomly drew your chart, you are assuming more elastic (the blue line)- and so in this case, I'd argue that if I was running the ops, I'd already be operating at max throughput. No excess capacity.

NR, I was thinking further over your comments about a "box hyperbola". OK. I see your point. But, for all intents and purposes, within a "certain range of production" it can be assumed to be constant slope. Point taken.

I've posted a response over at EconoSpeak, for those who are interested. The only thing I'd add here is that the shape of the MC curve is not relevant to this debate. MC can be horizontal (as it often is in the neighborhood of realized output), and Nick's model would still apply. True, Qmax would be indeterminate, but the point about wanting more buyers at P would stand.

"Tim Horton's does not raise the price of coffee and doughnuts when there's a rush. I don't explain why, but it doesn't."

How would that work, exactly? If there is a morning rush, then donut prices are doubled, and if no one is in the store, then donuts are free? Would customers need to call ahead to determine what the prices are? And how could the store possibly give them an answer, as a rush may occur between the time of the call and when the customer arrives?

There is no such thing as rigid price or a flexible price, there is only a price-setting mechanism that may or may not respond to short run fluctuations in demand. And each price-setting mechanism contains costs. Not just menu costs, but also bargaining costs, information processing costs, as well as an enforcement mechanism ensuring that all agents are putting in honest bids and have the means to pay. All of that requires real resources.

Perhaps the donut-pricing structure is already optimal, so that prices are as sticky as the donuts :)

On the other hand, in the financial markets, you do have real-time auctions, but even there you need market-makers that are necessary for the market to function, and the infrastructure costs of running a stock-exchange are high. Only because the number of transactions is so large in comparison to the value of the underlying is it critical to have a highly efficient price-setting mechanism. Also, they are not selling consumption goods -- it would be difficult to put in a naked bid or offer on consumption goods, but if you can't have naked transactions, then you can't have market makers and you lose a lot of liquidity.


I do not understand that:

but if you can't have naked transactions, then you can't have market makers and you lose a lot of liquidity

Naked shorting is illegal in the US today.

Some research does indicate it may be beneficial though.

But NS is not needed for market making.

Am I missing something ?


From //www.sec.gov/spotlight/keyregshoissues.htm

"Naked short selling is not necessarily a violation of the federal securities laws or the Commission's rules. Indeed, in certain circumstances, naked short selling contributes to market liquidity. For example, broker-dealers that make a market in a security generally stand ready to buy and sell the security on a regular and continuous basis at a publicly quoted price, even when there are no other buyers or sellers. Thus, market makers must sell a security to a buyer even when there are temporary shortages of that security available in the market. This may occur, for example, if there is a sudden surge in buying interest in that security, or if few investors are selling the security at that time. Because it may take a market maker considerable time to purchase or arrange to borrow the security, a market maker engaged in bona fide market making, particularly in a fast-moving market, may need to sell the security short without having arranged to borrow shares. This is especially true for market makers in thinly traded, illiquid stocks such as securities quoted on the OTC Bulletin Board, as there may be few shares available to purchase or borrow at a given time."

Take two potential monopolists:

One assumes the demand curve is less elastic (call them Potash Corp / Canpotex cartel). So, they restrict production.

The other assumes the demand curve is more elastic (call them BHP Billiton). So, they plan to produce at full capacity.

From BHP Billiton's perspective, total profit of Potash is suboptimal - "their demand curve is wrong" - hence a hostile takeover attempt.

Now, throw in the Sask gov't. It has a different profit dynamic again depending upon royalty schedule (rising pct as a function of price?). At a simple level: Status quo = certainty. Change = uncertainty.

Can this be modeled? Haven't seen it yet.

david: "If MC is flat or falling, firms should be bigger. They increase in size until their MC is again increasing. Even if the MC is non-monotonic (and we have multiple equilibria), it is then locally true that MC is generally rising at every equilibrium."

I assume the first sentence means that firms should increase output, to the point where MC rises because of falling prices. due to the demand curve, whatever its projected or actual shape may be. It's the demand curve that's the constraint. But there is no *necessary* generalizable reason why MC can't continue to decrease as demand is saturated, or why dropping price would increase both demand and profits. (Slope and elasticity of the demand curve, etc.) To the contrary, liquidating excess fixed capital capacity might be required perversely to raise MC. And when firms in a sector experience falling MC and thus expand output, that doesn't occur by some gravitational force/action-at-a-distance, but by a competitive process, in which likely stronger, more technically efficient and more capital intensive firms bankrupt or acquire weaker firms, (which is generally how capital stocks get liquidated). And those newer, stronger firms would have reason to maintain their market-dominant position precisely by maintaining excess capacity, (due to the discontinuous scale effects and lowered capital costs involved). Not to mention the long lead times and lags in realizing capital intensive investment, and the high uncertainties involved.

Of course, you could draw MC and demand curves any which way, zig-zag, step-wise, discontinuous. etc. and there might be corresponding empirical cases without any single generalizable case, which is an empirical matter among a mixed population. But if fixed capital is $.70 of costs and labor and materials $.15 each, when, say projected demand is 1000 units, and if, labor and materials rise to $.20, while fixed capital costs decline to $.583, when actual demand is 1200 units, then MC is still decreasing for the extra 200 units, by $.017. And I would maintain empirically that such a curve of decreasing MC throughout the empirically relevant potion of the curve would tend to be the empirically predominant case for that minority of large-scale, capital-intensive oligopolistic firms, which are responsible for the largest share of output in an advanced industrial economy. Thus a large part of MC curves in the economy would be open-ended.

My problem with Nick Rowe's account is that he assumes oligopolistic competition, rather than providing any account of how and why it comes about and what its likely dynamics are. It's simply not epistemically adequate to assume equilibrium and then provide a set of artificial assumptions to buttress the case. That amounts to assuming that firms are constantly optimally adapting to their market environments, that they always evolve in relation to their market environments to the edge (or, better, zone) of "criticality", whereby they are both sufficiently robust in their organizational structure and sufficiently open to the stochastic "chaos" of their environment, (which, of course, includes other competing and complementing firms), such that an overall optimum prevails. But that's an empirically independent matter, in which all kinds of mal-adaptions, dysfunctions, and disruptions, and thus a cumulative potential for crisis, can occur. (There is also a problem of assuming that firms are reducible to or derivable from the terms of market transactions, rather than, qua production systems, subject to different and separate kinds of constraints than just those imposed by markets).Of course, the basic question that was to be addressed by this thread wasn't the micro-economics of the firm, but the macro-economic question of why excess supply = production capacity tends to predominate over excess demand in modern capitalist economies. And the other half of that question is how demand is constituted from recycled production incomes.

It reminds me of a restaurant my wife and I like. The owner, sadly, may have to close up due to lack of business. He would love ten times the customers, but would lose money if he lowered his prices enough for this to happen. Partly this is also due to information problems. I'm sure if everyone knew exactly how good his food was he would get a lot more business. Lowering the price doesn't do that much to solve this information problem; directly it does nothing.

Partly we also see here price stickiness. Businesses can't just adjust their prices that quickly or easily as quantity increases.

Richard: restaurants always seem to me to be one of the very best examples of monopolistically competitive firms. Each restaurant sells a different product, even if only in the eyes of its customers, or in its location, and so faces a downward-sloping demand curve. Each restaurant is in competition with lots of other restaurants, not to mention home cooking, so there is little oligopolistic interdependence. And entry and exit is relatively easy.

John: You wanted the macro version? Here's the macro version, in an old post I did: http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/01/macroeconomics-with-monopolistic-competition-in-pictures.html

And the dynamics? Think about restaurants, and new restaurants opening up in what they see as profitable gaps in the multi-dimensional product-space, and closing or moving if their location in product-space becomes too crowded and so unprofitable. The location choices of ice cream seller on a circular beach is one simple model. But I don't need to do all that to make my simple point here. We need the simplest possible model that can make the point. Not to add in lots of unneccessary bells and whistles.

"But if fixed capital is $.70 of costs and labor and materials $.15 each, when, say projected demand is 1000 units, and if, labor and materials rise to $.20, while fixed capital costs decline to $.583, when actual demand is 1200 units, then MC is still decreasing for the extra 200 units, by $.017."

No it is *not*. It is very clear from your example that you are talking about AC, not MC. In your example, the MC is $.15 at 1000 units and rises to $.20 at 1200 units. Fixed costs ($700 in your example) are included in Average Costs, but are not included in Marginal Costs. They are not part of the *extra* costs of producing one *extra* unit, which is the definition of MC.

Yes, a monopolistically competitive firm in long run zero-profit equilibrium will have a (locally) declining AC curve. But that says nothing about the slope of the MC curve.

First understand the basic stuff, like the distinction between AC and MC. Then we can maybe talk about fancy stuff like macro, and dynamics.

JVFM: In practice, no firm knows where its demand curve is. It knows the current point on the demand curve, but does not know the elasticity at that point. And if it guesses wrong about elasticity, it is not maximising profits. So yes, if a potential purchaser thinks it can guess that elasticity of demand better, and can make higher profits, it should be able to succeed in a buyout.


"Naked short selling is not necessarily a violation of the federal securities laws or the Commission's rules"

That was true before Sep 2008. After Sep 2008:
Hard T+3 Close-Out Requirement; Penalties for Violation Include Prohibition of Further Short Sales, Mandatory Pre-Borrow.
The Commission approved a final rule to eliminate the options market maker exception from the close-out requirement of Rule 203(b)(3) in Regulation SHO. This rule change also becomes effective at 12:01 a.m. ET on Thursday, Sept. 18, 2008.

As a result, options market makers will be treated in the same way as all other market participants, and required to abide by the hard T+3 closeout requirements that effectively ban naked short selling.

Re. consumption goods:

One would argue that naked selling is *more* widespread among online consumer goods "market makers" than on the stock exchange. I personally experienced failure to deliver from Amazon, on several occasions, although the item was marked as being in stock. There is less supervision too: try to complain to BBB.

No, unless there is some technical obscurity that I'm missing, marginal cost means the cost of each additional or incremental unit, (since by any reasonable accounting a reduced loss or opportunity cost should count as an increased unit margin). And since in the example, the capacity of capital stocks exceeds most or all of any possible demand curve, by hypothesis, the unit capital cost declines to $.583, as distributed over all units, even as combined labor and materials costs rise to $.40. Marginal costs could be rising even as average costs were still falling, but that would be at the tail end of the curve. And there is nothing that says that tail would be reached and, in real world cases it often isn't.

However, it does seem that I've missed another bit of jargon, and interpreted "monopolistic competition" to mean "oligopolistic competition", which in cases of large-scale capital-intensive production firms, (whether through high upfront physical investment costs or high upfront R &D costs), provides the focus of my point. But I see that "monopolistic competition" merely means general competition, which is "imperfect" only because of indirect competition over product differentiation, (and thus supposedly a tendency to MC=MR rather than being forced to MC=MP). But then I fail to see all the more the relevance of your point to the issue, which I took to be providing any account as to why the economy as a whole tends to maintain a high degree of excess capacity and thus potential aggregate supply above any likely level of AD. (Which is something that comes home to roost in crises, where AD declines or collapses).

Alan Blinder of all people, by the way, wrote a book about a decade ago based on empirically surveying 200 hundred or so corporate CEOs and cases of falling MC curves were found to be quite prevalent there.

NR responds: "So yes, if a potential purchaser thinks it can guess that elasticity of demand better, and can make higher profits, it should be able to succeed in a buyout."

And that's the conundrum regarding the failed Potash takeover. Why was the Sask gov't so against it? The G&M on Sat had a blow by blow account of how BHP failed, but this comment caught my interest:

Tony Robson, the co-head of mining research at BMO Nesbitt Burns in Toronto, said BHP made a fatal error by threatening to tamper with Canpotex. “The biggest mistake was politics,” he said. “BHP and its Canadian advisers underestimated the impact [to the province’s revenues] of leaving Canpotex. It’s very clear the province gets its potash income from price, not volume.”


Interesting, in an economics, rather than a political way. I wonder if that is true for most production scenarios (impact on Sask's total royalty take).

Nick & Andy,

I had difficulty interpreting your discussion because it seemed to me that you are working with incompatible assumptions. In particular, when Andy says that his "reading of [Nick's] motivation was that everyone sees excess supply and firms that want to raise prices everywhere", I thought only the first part of this was true: Nick sees the appearance of excess supply everywhere but half the time (the interesting half) this is actually due to excess demand. The other half, firms might be happy to lower their prices because there really is excess supply. But if that is so, why did Nick not dispute Andy's interpretation?

Anyway, I was struck by this observation: "I certainly don't know why that would be more common than the situation where they wish to lower prices but cannot."

Again, there are two parts to this: 1) firms wish to lower prices, and 2) they cannot do so. In my reading of Nick's model, the second proposition is not explicitly required; it might be that it is easier to discount prices than to charge a premium in the short run. But in that case, rational expectations of demand are incompatible with rationally expected prices; prices should be systematically biased to be higher than P* (by which I mean the price associated ex ante with Q*.)

That would constitute a competing hypothesis of why excess supply seems to be "too frequent." In that sense, symmetrically sticky prices are implicitly assumed by Nick's conclusion. But that doesn't seem to be what Andy is driving at, so far as I can see. Clarifications?

John: It is not an obscure technicality. But you are very definitely missing it. You got the definition of MC right, but you didn't apply it correctly.

Let me try with this example.

Suppose Total Cost = $100 + ($5 x Q) , where Q is output

MC = (change in TC/change in Q) = $5

Average Cost = TC/Q = $100/Q + $5

In this example, the MC curve is horizontal at $5.

The AC curve is downward-sloping, and asymptotes towards $5.

Yes, you said "Oligopolistic competition", which is different, because of the game-theoretic problem of strategic interaction between a small number of firms, rather than the large number of firms in "monopolistic competition" (misleading name, I know). But you really must get MC vs AC straight before we can talk about different types of competition. Because if you don't understand that distinction, you won't understand the difference between Marginal Revenue and Average Revenue. And you need to understand that before you can understand why the downward-sloping demand curve of monopolistic competition makes a very big difference from the horizontal demand curve of perfect competition.

Phil: your first paragraph is spot on. I should have disputed Andy's interpretation. Half the time firms will wish they had set a higher price, and half the time they will wish they had set a lower price. But nearly all the time, given the price they have set, they will want to sell more than they are currently selling. ("nearly all the time" means "as long as quantity demanded is less than Qmax").

Phil's second bit would be correct as a competing hypothesis to explain the facts. But either I am misunderstanding Phil, or Phil goofed by getting it the wrong way around. (Or maybe I am goofing right now!). If firms can raise prices immediately when there's excess demand, we never see excess demand. But if firms can't cut prices when there's excess supply, we see excess supply.

Or maybe I am misunderstanding Phil. Perhaps Phil is saying that firms always set prices high, because they know they can always cut them later of they need to, but can't raise prices if they set them too low.

Nick: Yes, I was just proposing the second interpretation: that if firms know that prices are more sticky upward than downward, and their demand expectations are rational, then they will systematically set prices too high, knowing they can always lower them if necessary. That was one way of explaining why Andy thought that your model assumed prices must be downward sticky: if they are not, then your explanation would not necessarily hold - the tendency to excess supply might be real, due to a tendency to have prices that are too high. Re-reading what I wrote, I think I have it right-way around.


I wasn't talking about the options loophole, but about the 3 day settlement window. The naked position needs to be closed within 3 days or there will be a settlement failure, but 3 days is generally enough, as there hasn't been a lot of failures.

The buyer shows up on Monday, and the seller on Tuesday, and the market maker can make it seem as if both parties are buying and selling whenever they want. You do not need to wait until a seller shows up. Note that I am not talking about naked shorting as an arbitrage technique, nor about naked positions in derivatives -- I'm describing it purely in the cited SEC language, where market makers "stand ready to buy and sell" but there may not be immediately offsetting transactions, so they will be need to be naked for a brief period of time. This obviously adds liquidity to the market and is the purpose of having a settlement period.

That is funny about failure to deliver via postal mail.


I think the reason why there is confusion of AC and MC is because your model is timeless whereas most people think in terms of time. So you are probably thinking there is a $100 oven, which does not depend on Q produced, and the per unit costs would be the labor and inputs. So the $100 has no Q dependency and goes away when you differentiate. Also the production process is timeless -- there are no units produced per week, just units produced. The time axis consists of two points "short run" and "long run", and in the short run, capital is assumed to be fixed, and quantities are produced instantly.

But a large firm is always replacing the ovens simultaneous to producing; there are many ovens, and some will be at the point of wearing out. So even over the short run, if the expected Q decreases, the firm is able to change the capital stock, for example by not replacing an oven, delaying software upgrades, closing plants, or by cutting down on marketing expenses, etc. The $100 has a time dependency, and Q has a time dependency, so the $100 has an implicit Q dependency. Alternately, if Q is higher than expected, then when an oven needs to be replaced, the firm can get a more efficient one, decreasing the marginal cost curve in real time as Q goes up.

But if firms are allowed to change their short run cost curves in real time by adjusting their capital stock, then perhaps excess capacity is just a byproduct of on-going AC minimization. Only when the Q demanded is a large shock that exceeds the ability of the firm to adjust their cost curve would you reach excess capacity. So in this case, rational expectations would suggest excess capacity, given that the capital stock for the firm as a whole was about as flexible as the magnitude of the shocks.

Oops, in the last paragraph, I meant to say "Only when the Q demanded is a large shock that exceeds the ability of the firm to adjust their cost curve would you not have excess capacity so that firms would need to turn the marginal customer away."


If one wanted to be obnoxiously logical, one would remark that any transaction whatsoever that has a temporal settlement leeway is potentially "naked".

By the same logic, any retailer can be regarded as a "market maker" thanks to the above mentioned temporal leeway that makes possible either "naked" sell, "naked" buy or both if the transacting parties decide to do so (lying about having the goods in stock, or cash to pay-- the situation is really symmetrical).

That's perhaps why SEC established the T+3 rule having arbitrarily decided that anything that is settled(or failed to settle) beyond the three day period is "naked", but within such a period is not "naked". It is similar in its arbitrariness to the legal drinking age:

In a "naked" short sale, the seller does not borrow or arrange to borrow the securities in time to make delivery to the buyer within the standard three-day settlement period.

In other words, "nakedness" happens not because the seller does not have the security on hand, but rather because the seller was unable to procure it in allotted time.

Seriously, in my opinion, in the established security markets, the "market makers"/"specialists" are largely rent-seeking parasites that can easily be eliminated. Their role in providing market liquidity is highly dubious and utility of such virtual liquidity along with purported market stability provision is suspect. It ain't no rocket science and no technological miracle for market participants to trade directly and settle their mutual obligation immediately, in real time, with finality and without "nakedness". Everyone in the world wishing to trade could do so with his PC and a simple piece of software so that bid/ask spreads are created naturally. A modern PC with an e-bay like piece of software could easily and efficiently replace the market maker.

The middlemen can migrate to the world of customized transactions of various kinds where their intermediation may in fact be useful.

The online retailer "naked" transactions are, arguably, more excusable, due to various logistics, than their analogues in the world of the commercial or government paper handler/"market maker".


"If one wanted to be obnoxiously logical, one would remark that any transaction whatsoever that has a temporal settlement leeway is potentially "naked"."

Heh. Several times I tried to short with a "modern PC" and retail account, and was told "cannot locate shares to short".

Only the broker dealers get to be naked during the settlement, they may pass this privilege onto certain customers, but not to retail customers in general. Retail customers need a locate, whereas middlemen do not.

I can understand that if you have buy side experience, then perhaps you would not see the actual constraint in your trading experience. But again, that doesn't have anything to do with the argument that market makers have a legitimate short term need to be naked.

In terms of market makers being rent-seeking parasites, no argument there. I also think that you might be able to devise a distributed system to replace them, but doing so would be harder than you think, as someone would need to bear the risk (and concomitant ownership and financial responsibilities) of going naked when the number of buyers is not the same as the number of sellers. I don't think you could offload that responsibility to "the network". At the very least, there would need to be an entity with funds backing "the network", to which traders would have recourse if the network failed. This entity would demand some premium, and you are back to the current situation, more or less. Perhaps nationalizing the BDs is the only option -- e.g. creating something along the lines of a Fedwire for stocks, in which the government guarantees settlement. But even that would be problematic, if you gave retail direct access to the network. You can imagine distributed attacks if they use the wrong price-setting algorithm when there is a discrepancy between bids and asks. Ultimately, you need someone with judgement and assets at risk to decide what they are willing to pay when they take the other side of the trade.

Really running continuos auctions is extremely difficult, even for high volume markets such as the stock market. Doing this in the goods markets is insane. But this instantaneous auction assumption is at the heart of a lot of other economic paradigms, at least in my unprofessional opinion. Perhaps others will disabuse me of this notion.

It's no wonder that people are puzzled why prices are sticky if the models assume that continuous auctions are occurring, or that continuous auctions are a good approximation to the price setting mechanism for donuts. If that was true, then prices would not be sticky. In the stock market, prices are not sticky. In the commodities market, prices are not sticky. I don't think it has anything to do with psychology, but with basic cost-benefit trade-offs when the underlying is not fungible and the ratio of net sales to transactions is not many orders of magnitude greater than 1. I guess this was Nick's "kink" post.

I think you are right, but I am pretty sure this is exactly what Steve Keen has been pointing out for some time. But it is interesting that you came to this conclusion with standard micro-economics. But there may be other factors involved as well. Maybe there is a consistant tendency for suppliers to overestimate their marginal revenue. After all, entrepeneurs are entrepeneurs because they are optimistic folks, and a significant percentage of them go broke, even in good times.

If demand is stochastic, won't firms hold excess inventories for sort of 'buffer stock' reasons? Because of time to build and other considerations, they want to have some spare capacity on hand to take advantage of extra demand. Firms don't take their demand curves as given but have sales forces etc. exerting effort to push out those demand curves. There's no point in having a sales force taking orders you are unable to fulfill.

Aren't there also implications for the relationship between demand and measured productivity? If you always have some slack in the firm, then an increase in demand will look like an increase in productivity too, won't it?

RSJ: my hunch is that you are right, but I can't really prove it. The markets that most closely approximate the textbook models of continuous flow auctions of perfectly divisible identical goods with zero transactions costs etc., do have very flexible prices. It's the markets that are very different that seem to have sticky prices.

reason: I'm not familiar with Steve Keen's proposed solution to the problem. But the monopolistic comp solution is fairly standard. Some New Keynesian macroeconomists saw it about 20 years back.

Your sample selection bias solution is neat. I could believe it for new firms. But for older firms, it's hard they could be overoptimistic and make consistent mistakes day after day for years.

Luis Enrique: Yes, they will hold inventories, if they can. I'm trying to do a short post, like this one, only building in inventories.

The marginal revenue product of a sales force is the extra sales per extra salesperson, times the gap between P and MC. So it makes sense to pay people to try to push your demand curve right if P is geater than MC, as it is under monop comp. Yes. Agreed. (Not in my simple model, but can be added).

Yep. The "labour hoarding" hypothesis says that measured productivity should fall in a recession for very similar reasons.

Yes, they will hold inventories, if they can. I'm trying to do a short post, like this one, only building in inventories.

I'd suggest picking a specific industry. The economic rationale for holding inventory differs depending upon the nature of the business. Just in time (JIT) production limits the need for inventory (say automobile assembly). Some high tech consumer electronics (with high margins) have high opportunity costs - and wide dis'n channels - so more inventory in the pipeline. Walmart has driven down costs (and gained market share) through their sophisticated materials management systems, integrating right back to the manufacturer.

One other thing about inventory - it can be used as a short term economic buffer - because it shows up on a firm's balance sheet - not income statement. The general accounting rule for recording inventory is - the lower of cost or market. And in the "cost" it would include some portion of overheads/semi fixed costs. Of course if the demand continues to fall or remains flat and you are still building inventory, the day of reckoning when you have to cut back production/take more drastic action eventually arrives.

thanks for your answer. But with regard to this:
"Your sample selection bias solution is neat. I could believe it for new firms. But for older firms, it's hard they could be overoptimistic and make consistent mistakes day after day for years."

I'm not so sure. Older firms are also not static. They have new managers all the time, and new product releases. Maybe internal forces in the firms tend to select for the optimistic ones (for instance people get promoted for being involved in successful projects - some projects are successful - so people promoted for running successful projects tend to think all projects will be successful).

is this the Steve Keen in question?


The guy doesn't even understand the most basic expected utility theory. Pretty scary.

Oh God. Adam is totally right. And he really does mean *the most basic*. Like, utility might not be (and almost certainly isn't) linear in wealth.

What I can't understand is how Adam managed to be so restrained in his above comment.

It's a lack of knowledge of the first year textbook on "why people buy insurance". Because of diminishing marginal utility. So maximising expected utility is not the same as maximising expected wealth.

reason: maybe. In addition to being really neat, your theory may be a part of the truth. My gut just tells me it can't be a big part of the truth.

Your theory is like the "Winner's Curse", where it is the most overoptimistic person who places the highest value on a used car who buys it at auction. But the winner's curse does not survive: self-reflection ("hang on, why is everyone else dropping out of the auction? Maybe there's a winner's curse?"; and it does not survive learning from repeated experiments.

Nick: "What I can't understand is how Adam managed to be so restrained in his above comment."

Well, with effort. Seriously though, Keen shouldn't be teaching economics. If he was taking first year econ he'd be failing!

One of my abandoned half-assed ideas concerned 'rational' over optimism by firms. It's based on my observations when I used to be a business journalist that too many firms - big old firms - claimed to be planning on growing their market share for that to make sense in aggregate. They can't all grow market share. I think you can tell a story about it being rational for individual managers all to plan to grow market share. Somebody's going to lose share, but who wants to tell their boss they're planning on losing market share that year? You lose your job right there. The rewards are asymmetric, so managers don't plan like a rational expectations planner. My idea was that during good times when the pie was growing firms accumulate excess (just sub-optimal, not loss making) labour, investment, overheads etc. and get more and more exposed to a negative shock, so when the downturn comes they don't think 'no worries, we planned for this' but 'oh shit we need to cut costs'. Hard to make work, because really presupposes existence of habitual excess profits without explaining their existence.

[as ever, if somebody realizes this idea isn't so half-assed and could work, get in touch!]

co-incidentally was discussing Keen last night - my supervisor likes to say that heterodox economists often ask the right questions, even if they don't always have the right answers and often don't understand what they purport to refute. You can't fault Keen for his long-standing focus on debt levels, but Lord some of his 'take downs' of mainstream economics are embarrassing.

Shareholders require growth. So senior management is increasingly rewarded for short term performance/stock price appreciation. And this translates throughout the company in terms of KPIs (key performance indicators). On average, I think CEOs last two years, something in that neighbourhood. I'd look at that aspect - the internal reward system. It may lead to overcapacity due to overly optimistic forecasts. Not sure if it would necessarily lead to higher inventory levels. Maybe.

Luis Enrique and JVFM: that reminds me of this old post of mine:

interesting! It doesn't have a built in boom and bust business cycle dynamic, which is what I was angling for (I know, yours is a growth model, not a model of business cycles). What I also had in mind (but didn't say above) is that the over investing, over hiring, might have a sort of Keynesian multiplier demand effect, so some self-fulfilling potential for some period of time ... and having firms over expanded during the expansion means they have to over correct in the correction, which creates a larger negative Keynesian demand multiplier in that phase. So I'm thinking of an endogenous cycle story with an amplification effect caused by some moderate collective irrationality. But why would the partially self-fulfilling expansion run out of steam? Perhaps the gap between plans and reality increases every period until people can no longer ignore / deny it, and managers have to correct. Easier to say than to model.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad