Navigation
Papers by Melberg
Elster Page
Ph.D work

About this web
Why?
Who am I?
Recommended
Statistics
Mail me
Subscribe
Search papers
List of titles only
Categorised titles

General Themes
Ph.D. in progress
Economics
Russia
Political Theory
Statistics/Econometrics
Various papers

The Questions
Ph.D Work
Introduction
Cost-Benefit
Statistical Problems
Social Interaction
Centralization vs. Decentralization

Economics
Define economics!
Models, Formalism
Fluctuations, Crisis
Psychology

Statistics
Econometrics

Review of textbooks

Belief formation
Inifinite regress
Rationality

Russia
Collapse of Communism
Political Culture
Reviews

Political Science
State Intervention
Justice/Rights/Paternalism
Nationalism/Ethnic Violence

Various
Yearly reviews

Philosophy
Explanation=?
Methodology

 

[Note for bibliographic reference: Melberg, Hans O. (1997), Justifying the welfare state : Biased but useful (Review of Barr), www.oocities.org/hmelberg/papers/971127.htm]

 

[Note: I am not fully satisfied with the quality of this review; it still needs quite a lot of polishing, and I probably overstate my case a bit. There are also additional arguments, such as the nature of liberty, that I would have liked to discuss. Nevertheless, since I have promised to make something available every week, since I am short on time, and since I believe there are at least some sound and valuable arguments in this review, I have decided include it among my papers.]

Justifying the welfare state
Biased but useful (Review of Barr)

by Hans O. Melberg


The Economics of the welfare state
Nicholas Barr
Weidenfeld and Nicolson, London, 1993
ISBN: 0 297 82158 - X (Paperback), 495 pages

Introduction
The central argument of Nicholas Barr's book The Economics of the Welfare State is simple and clearly stated:

"The welfare state [i.e. income support, health care, education and housing] is much more than a safety net; it is justified not simply by any redistributive aims one may (or may not) have, but because it does things which private markets for technical reasons either would not do at all, or would do inefficiently. We need a welfare state of some sort for efficiency reasons, and would continue to do so even if all the distributional problems had been solved." (433)

The term "technical reasons" is shorthand for information problems and externalities. For instance, since people themselves can affect the probability of becoming unemployed, it is very difficult to create a well-functioning private market for unemployment insurance (the problem is technically known as moral hazard). Another problem with unemployment insurance, is that the failure to take buy such insurance (if it were possible), has external effects. It is not only you who suffer the consequences if you become unemployed, but your dependants may starve, I may be hurt by rising crime committed by the unemployed, and there might be a cost in terms of reduced "social solidarity". Altogether, the problems of moral hazard and externalities - in Barr's opinion - make a strong theoretical case for compulsory and publicly provided unemployment insurance. Publicly provided because no private firm will offer insurance when there are great moral hazard problems; Compulsory because of the externalities involved.

In a manner similar to that above (appealing to externalities and information problems) Barr argues in favour of different parts of the welfare state (health programs, education, income support, social insurance). The question is then: Is this really a good justification for the welfare state?

My first response is that the argument is certainly not ideologically neutral, as he claims. Second, I think the argument is incomplete and somewhat biased. Incomplete because there are many links between the welfare state and efficiency which are not discussed, and biased because some of the negative links that are discussed are either discussed very briefly, or often dismissed as inconclusive.

Neutrality and John Wayne
Consider the claim that welfare economics is "an area in which economic theory is capable of strong results which can justify the general idea of the welfare state to a surprising extent without resort to ideology" (3).

The problem with this argument, is that aggregate externalities do not automatically justify compulsion at the level of the individual (see p. 194). For the sake of argument, imagine a single stubborn individual - call him John Wayne - who decides that he does not want unemployment or poverty insurance. Being single, we cannot use the externality argument of starving children (see p. 194 for his list of externalities justifying compulsion). Imagine too, that he will not become a criminal if he becomes unemployed (not all unemployed people become criminals!). Still trying to justify compulsion, we might then argue that him being poor causes me displeasure, but I would question whether we should count these kinds of preferences. This is an issue Barr does not discuss, but I think that accepting a person's preference for other people's welfare in this way implies that we also have to count changes in utility caused by envy (which we do not want). Lastly, there is the argument that there are costs associated with disposing of the body of John Wayne (if he becomes unemployed and starves to death). Apparently Barr thinks this is an argument worth mentioning, but these costs are - in my opinion - far from significant enough to outweigh the utility John Wayne receives from "being independent", and hardly significant enough to justify infringing his rights. In sum, for this individual the externality argument does not apply (or, more weakly, it is not strong enough), and compulsion is not justified. To argue that we nevertheless, for the sake of the well being of most people who are not like John Wayne, should have a compulsory system, is not a neutral political position. (I am not one of them in this case, but libertarian like this exist).

The example of John Wayne who does not want unemployment or sick-insurance, may sound far fetched, but consider the following example with the same structure (from the newspaper Vaart Land, 25. November, 1997). Three researchers have estimated that the total cost of alcohol in Norway was about $10 billion (included lost work hours, alcohol related health-problems, crime and more). However, the state only receives 1.4$ billion from the sale of alcohol [I do not trust these figures, but that is not the point here]. In other words, there is an externality: The people buying alcohol do not pay all the costs. Does this imply that we should raise the price of alcohol?

The problem, of course, is that not all the people who buy alcohol commit crimes, stay away from work and so on. If we raise the price for all we also force those who do not abuse alcohol to pay the price for those who does. This is not efficient (economically speaking), it does not appear fair, and it is certainly not a "neutral" position.

Neutrality and means
Another claim that Barr repeatedly makes, is that his arguments are amount the best means to realize a policy aim. The specification of the aim is ideological, but the specification of means is a positive issue: It is simply a question of finding the means that most effectively realize you aim. These questions, he thinks, can be settled without value judgements.

My immediate reaction is that this is an example of "ends justifying the means" - a position which in turn may imply using quite unethical means to achieve your aims. However, Barr can counter this argument by saying that the fact that we object to some means really imply that we have an aim which we did not specify initially. In this sense the statement is still true, although close to meaningless.

A more problematic difficulty is the fact that means and ends are not as independent as one may think. For instance, Barr writes that "... the treatment of private property is not an end in itself but a means towards the achievement of stated aims" (428). The problem with this is that an instrumental justification for something may not have the desired effects. As an illustration consider the following quotation:

"But there is a deeper question: can the planners implement constitutional reform without a normative commitment to individual rights? If they introduce a constitutional system and abdicate from some of their powers simply in order to get the economy going, the economic agents will always fear that the rights will be abolished as soon as the economy gets into trouble ... " In short, "the instrumental benefits of freedom are essentially by-products, and will be forthcoming only if freedom is valued for non-instrumental reasons." [Elster, Jon (1988), Jon Elster Goes to China, London Review of Books, vol. 10, no, 19 (27. October), pp. 7-9]

Hence, to choose property regime according to its instrumental efficiency in fulfilling some aim, may be self-defeating: only non-instrumental justification will bring about the full benefits of private property (and, you cannot instrumentally choose to be non-instrumental and expect to be credible). As an illustration consider a potential private farmer in Russia. He will not start to work buy land unless he can be sure than the government will not change its mind and collectivise the farm again. A government that justifies private land by efficiency reasons leaves open the possibility that land can be collectivised, and hence the farmer may decide that the risk of buying and working on his own plot is too high. As a result, the instrumental justification for private land is self-defeating.

[Note: This argument is not simply that you should consider how the choice of means affect your the fulfilment of your aim before you choose the best means. The problem is that the means-end framework cannot be used to produce what Elster defines as "essentially by-products" - states that by their very nature cannot be brought about if you consciously try to do so, e.g. if you try to hard to fall asleep you will most likely fail. I do not know to what extent the economic growth resulting from private property really is an essential by-product; this is an empirical question which I know little about.]

Sub-conclusion on neutrality
The case for and against the welfare state can never be decided in a neutral fashion simply by appealing to positive economics (see also Atkinson in Barr and Whynes, 1993, p. 43). Moreover, the argument that positive economics is simply about finding the best possible means to achieve whatever aim the policy-makers specify (a neutral and technical exercise), sometimes run into the problem that instrumental reasoning in itself defeats the aim.

Economic arguments: Biased and incomplete
The headline of this sub-chapter may appear a bit negative, so I shall immediately admit that Barr has produced a very useful, clear and systematic overview of the argument for and against the welfare state. The same cannot be said for some right wing authors, such as Robert Skidelsky's book After Communism who proceed in a less rigorous and organized manner (see my review of Skidelsky).

What is the right frame for deciding questions of state intervention? Barr's starting point is the Invisible Hand Theorem which says that under certain assumptions free markets produce an outcome in which nobody can be made better off unless some are made worse off (i.e. the outcome is Pareto efficient; the first theorem of welfare economics). Moreover, it can be shown that all possible distributions can be achieved simply by making changes to the initial assignment of property rights (the second theorem of welfare economic). The intuition behind this was given by Adam Smith more than 200 years ago, but it was first formalised by Arrow and Debreu in the 1950s (this kind of work is known as general equilibrium analysis). In sum, the free market seems to be the best mechanism to maximize welfare given our preferences, resources and technology.

To show that state intervention can improve on the free market, we have to show that the market somehow fails. One way of doing this is to examine whether the assumptions behind the Invisible Hand theorem hold in the real world. If they do not, intervention may (but it need not) produce a better result than the free market left to itself. The three major assumptions, as presented by Barr, are: Perfect competition, perfect information, no market failures.

Perfect competition implies that nobody in the economy are strong enough to influence the price (i.e. have monopoly power). In the real world, we sometimes find that this is not the case; Microsoft can certainly influence the price of computer operating systems and unions can influence the wage rate.

The condition of no market failure, as discussed by Barr, includes the conditions that it must not be possible to enjoy the full benefits of a good which another has bought (e.g. if I pay for the defence of our country it is impossible to prevent you from free-riding on my contribution); there must be no external effects (effects which are not taken fully into consideration by the decision-maker e.g. pollution); and no increasing returns to scale (when a doubling of the inputs more than double the outputs).

Lastly, perfect information implies that you must know the real price of the product, the quality of the product, and probabilities of future events (e.g.. when you are going to die, how likely it is that you will suffer an industrial injury etc.).

These are, as Barr points out, very strong conditions. But, as he is less keen to point out, the failure of these conditions do not automatically imply that intervention is desirable. The first problem is that when one of the conditions do not hold, the breakdown of the other conditions need not be a problem (the general theory behind this is known as the theory of the second best and was developed by Lipsey and Lancaster, 1956). For instance, there might be a small externality in the production of good X, but X may also be complementary to another product (Y) which has very positive external effects. Thus, intervention to raise the price of X will reduce the negative externality associated with X, but it will also reduce the consumption of Y and thus the good externalities associated with Y. In sum, the net effect of intervention is unknown: it can be positive as well as negative.

The problem is simply that often we do not know enough to be sure that intervention will produce a positive result. The consumption and production of goods are linked in so many ways, the conditions for a second order optimum so complex, and our knowledge of the causal mechanisms in society so limited, that intervention is not automatically justified as soon as we discover an externality (see, Ng Welfare Economics for a good discussion of the complexity and the information required to fulfil the conditions to achieve the second best solution).

From the discussion above it should be clear that one need not be an extreme libertarian to be weary of arguments for intervention based on externalities. On the contrary, traditional conservatives always argued that we simply do not know enough about society to justify large scale intervention (Burke, Oakeshott). This is a pragmatic argument, not one based on rights. In Barr's overview of the three major political theories (Libertarian, Liberal and Collectivist), the "conservative" position is called "empirical libertarianism", in contrast to moral libertarians (Nozick). The other two main directions are also divided; Liberals can either be utilitarians or "Rawlsian" and Collectivists are either Socialists or Marxists.

It is also a bit unfair of Barr to accuse Hayek of taking "little account of information problems" (103). First, it is unfair to accuse somebody who wrote before the problems of moral hazard, adverse selection and the rest of modern information economics were known of "failure to acknowledge" these problems (433). Second, Hayek's main focus was on those information failures which led intervention to fail, for instance that the government do not know the preferences of people as well as people themselves; that the government cannot gather the information that the price system so efficiently communicates; and that the government do not have the necessary knowledge of how society works to be certain of achieving the desired result. This investigation may be biased, but understandably so knowing the time in which he was writing; and Barr's relative lack of focus on these issues is no less biased.

Although Barr does not give us a systematic defence of the government's ability to intervene to produce a better general equilibrium, he does give some evidence that intervention at least have predictable and positive partial effects. For instance, to show that poverty is reduced by income transfers and the pension system "removed from poverty over 80 per cent of the pre-transfer poor" (p. 146). Thus, intervention does work in increasing the income of the poor today. However, it the question is phrased slightly differently - not whether the system implies that the poor gets more money at a given point in time, but whether poverty is reduced over time - the answer is less satisfactory (see David Whynes in Atkinson and Whynes, 1993, p. 64). The introduction of social insurance and other measures were thought to gradually eliminate the need for public assistance, yet the opposite happened: "As a percentage of total expenditure, national assistance increased, rather than fell" (Atkinson in Bar and Whynes, 1993, p. 27). Hence, while the intervention was and is successful in relieving poverty at one point in time (which itself is generally accepted as a good thing), it has failed in reducing the problem of poverty over time.

Leaving aside the question of whether the government has the ability to intervene to produce a better total outcome, we might question whether it has the desire to do so. The government consists of people who, like most of us, have personal and selfish interests. Hence, instead of intervening to produce a more efficient and/or fair result, they may end up doing what is in their own personal interests. For instance, giving in to lobbyists in order to get money to be re-elected.

Altogether, intervention is not easily justified. We must show that there is a market failure (not only locally since one failure may cancel another), that intervention has the potential to improve the situation (again, not only locally and statically, but a new better general equilibrium and preferable reduce the problem over time), and that intervention actually will produce a better result. Finally, we must accept a political theory which allows us to use compulsion as a means to achieve a better result.

Left out
Having discussed the general theory behind state intervention, I will no go on to discuss three possibly relevant mechanisms which are not mentioned in Barr (the risk of too much intervention, private but non-market provision and the non-linear cost of taxation). The failure to mention these is one reason why I think the account is somewhat biased, and this time the accusation is more fair that Barr's claim against Hayek since these issues are well known.

A slippery slope: The risk of too much intervention
At one point Barr notes that government spending as percentage of GDP has grown from 21% in 1920 to 41% in 1992. Moreover, welfare spending as a percentage of government spending has increased from 28% in 1920 to 60% in 1992 (p. 174, all numbers refer to the UK). The question is then whether this increase represent a problem?

Conservative like to argue that the increase indicates the growth of state power at the expense of civil society. Left wing academics often reply that the conservative argument fails to distinguish between government controlled spending and transfer payments. Transfer payments, like pensions, increase the state share of GDP, but the state does not actually decide how to spend the money the pensioners get. The left is right in that we should distinguish between transfer payments and other types of government activities, but this does not necessarily solve the problem. Even if they are transfer payments, they still represent an increase in state power. It is the government which decides the size of your pension and the degree of redistribution in society.

Regardless of the debate about the difference between transfer payments and government consumption, there are several reasons why we might be worried about the growth of the state controlled sector. One of these is that the scope for unproductive rent-seeking activities increase. Different groups invest time and resources in persuading the government to listen to their demands. Another problem is simple that a large government increases the risk of abuse of government power.

As an illustration of the last point, consider the system of state transfer of money to support voluntary organisations in Norway. Lately, and predictable one might add, there were several major scandals associated with these transfers when it transpired that some organisations had claimed higher membership than they in fact had had. The government reacted by enacting new rules; from now on they had to report more details about each member (name, address, date of birth and so on). In itself this may seem innocent and reasonable, but consider the use a "bad" government could make of this information. How much easier would it not be for the Nazis if the Norwegians already had had a register of the names and addresses of the members of an organization working for Jewish interests in Norway. How much easier would it not have been for the Communists to purge the old elite when they occupied the Baltic countries if they could find all the names of the leaders of most voluntary and political organisations on a computer file?

To some these arguments sound far fetched. There is little chance, they say, that the information will be abused in this way. I am no so sure. History has made many strange turns, and we cannot exclude the possibility of future "bad" governments. Note also that even if most people are not evil and power hungry, the state must still be organized to consider the possibility of its "worst" members coming to power. The reason is that politics may under certain circumstances be dominated by the mechanism of adverse selection, only the worst - the most ruthless and power hungry - survive. Hence, as an analogy to Rawls' maxmin rule (the best society is that which maximizes the life prospects of the worst off), we might offer the following maxmin rule: The relationship between the state and civil society ought to be organised so that the worst possible government can do the least possible harm. [Both rules, incidentally, seem too risk-averse]. A large state does not fulfil this condition.

In what sense is this discussion relevant to Barr? First, Barr argues in favour of interventions that would both increase the size of the state and its information about citizens. Second, Barr's approach is to decide each issue on its own, based on considerations of costs and benefits. Looking at each market separately, one might conclude (as he does) that the benefits of government intervention is larger than the costs. However, each measure may also have an externality: An increase in government power. Isolated this externality may not be enough to affect the cost benefit consideration in one market, but in aggregate the result may be a state that is more powerful than we would like.

Private, non-market provision
In his discussion Barr focuses on two major alternatives for organising social insurance and assistance: Private vs. Public. He then tries to demonstrate that some forms of insurance is impossible to provide in the market, and that public intervention can remedy the situation. For instance, moral hazard and common shocks makes it difficult to organise private insurance against inflation. Similarly, the impossibility of insuring fully against inflation makes public intervention in the pension system desirable. So, given the failure of private markets to provide insurance, public intervention is necessary.

The option that is ignored in this argument, is private non-market provision of social insurance and assistance. For instance, in traditional societies the young are often obliged by norms and traditions to take care of their own parents. One could argue, as some conservatives do today, that as the state takes more and more responsibility for tasks that used to be provided on a non-market private basis, it creates a self-fulfilling process whereby private non-market provision breaks down and the need for public assistance only increases (until it becomes too large to handle ...).

Personally I do not know what to make of this argument. It is certainly coloured by a rather romantic view of how things were done in the past; yet it is also true that the state can never afford to pay for all kinds of care, or that "paid" care can ever fulfil the same function as care within the family.

Although Barr does not discuss private non-market provision of social insurance, he does consider private charity as an alternative to social assistance. In that connection he makes a rather interesting argument about how the income distribution is really a public good: I may be willing to give in order to decrease poverty, but the effect of my contribution is so insignificant that it is not worth the costs. To solve this problem, many people may agree to each give some money only as long as the others also give ("I give as long as you give"). Although I did not want to give money alone (since it would not decrease poverty significantly), I may be willing to support this scheme (since if many give the same amount it will reduce poverty significantly). The argument is both true and interesting, but one may doubt whether this is really how people who fail to give think. Is it not more likely that donations are motivated by bad-conscience, pure altruism and individual concern for a concrete family that suffers - than an abstract concern for the aggregate income distribution. If donations are motivated by the first three mechanisms, the free-rider problem does not exist.

The non-linear cost of taxation
A third problem not mentioned as far as I could see by Barr, is that the marginal cost of increasing the tax rate to raise money for social assistance is larger than the average cost [This is also an argument for considering the system in aggregate, not only the isolated costs/benefits]. True, it may be possible to raise money is less costly ways (negative income tax), but in practise revenue is usually raised in ways that cause dead-weight loses.

A simple model illustrating the non-linear cost of increasing the tax rate, is presented by H.S. Rosen in chapter 14 (taxation and efficiency) of his book Public Finance (see, p. 312 for the mathematics). In short, we are trying to estimate the size of the so called "dead-weight triangle" resulting from taxation. The area of a triangle is given by:: (1) A = 1/2 base * height
in our case:
(2) A = 1/2 (di) * (fd)
fd is simple the change in price, so (using D as a symbol for delta)
(3) fd = D p
(4) D p is the same as (1 + t) p - p = t p
Hence, fd = t p
Now, di is the change in quantity ( D q).
(5) D q = D q (Dp/Dp) = (Dq/Dp) Dp
Recall that the price elstaticity (n) is defined as:
(6) n = (Dq/Dp) * (p/q)
which implies that
(7) Dq/Dp = n (q/p)
Using this result we have:
(8) di = n (q/p) Dp
which can be transformed into
(9) di = n (q/p) t p
We now have expressions for di and fi so A =1/2 di * fd becomes:
(10) A = 1/2 n b t t p
(11) A = 1/2 n p b t2

The important point to note about equation 11, is that the deadweight cost of taxation increases exponentially; increasing the tax from 20% to 21% is more costly than increasing the tax rate from 10 to 11%. If we are willing to draw policy implications from this (which is dangerous given the already mentioned problem of first best vs. second best), it is that we should think twice before making the state sector so large that it requires high rates of taxation.

Bias II
One source of bias was arguments simply not discussed. Another source, is arguments discussed very briefly. For instance, the topic "government failure is discussed on two pages, that is less that 0.5% of the whole book. A balanced book trying to discuss the pros and cons of state intervention should give more space to this topic.

Another source of bias, is that the author quite consistently concludes that his own arguments are strong, while the conservative counter-arguments are complex and uncertain. To justify my argument, I offer the following sample of quotations:

* About his own arguments for the welfare state:
- "In all these cases there is an overwhelming case for compulsion ..." (199); "These arguments are compelling." (201)
"... there is a cast-iron efficiency argument for at least some public involvement with pensions." (218)
- "The efficiency case for substantial public, PAYG involvement is therefore strong." (237)
- "Thus there are solid arguments of both efficiency and social justice for public provision of subsistence benefits on a non-insurance basis." (244)

* When discussion possible arguments against his proposals:
- "These insights, however, should not be overstated" (94, about government failure and the public choice arguments)
- the literature is "large, complex and controversial" (203); "The simplicity of the original argument disappears the deeper one digs ..." (204, when discussing whether unemployment benefits cause unemployment)
- "considerable controversy" whether funded is more than effective that PAYG pension system in increasing output (223)
- "... the issue remains unsolved" (230, about the effects of savings, pension system and economic growth)
- Welfare state and economic growth : "terra incognita" (436) and no "obvious correlation" (436)

Of course, it might really be true that his arguments are strong, and all the others weak and uncertain. However, the choice of words to express this, and the detailed dissertation of counter-arguments in contrast to the less critical attitude to his "strong" conclusions, makes me suspect that his own "moderately egalitarian aims" (437) have made the presentation somewhat biased.

To mention one possible example of this, consider the effects on innovation of public vs. private pensions which Barr argues is "theoretically indeterminate" (229) and empirically "unresolved" (230). Two pages later he suggests that the introduction of robots could be one answer to the demographic problem (i.e. the growing number of pensioners compared to people in the work-force). At this point one might make the connection that in a system with privately (employer) provided pensions, the firm has a great incentive to introduce robots (save pension expenses). In a public system, however, this motive for introducing robots is weaker since you do not receive the benefits in terms of reduced pension-payments when you introduce robots. Barr does not make this incentive argument. Moreover, his list of supply side measures (232) to avoid the demographic problem leaves me wondering whether he has discovered a simple way to increase economic growth which is not exploited today. Clearly his measures also have costs, and the question is whether the costs are larger than the benefits; increasing education always sound good but at one point the net return from education also turns negative.

Conclusion: Good (and Bad)
I want to re-emphasise that this book is by far the most rigorous, systematic and clear overview of the pros and cons of the welfare state that I have ever read. In a review there is a tendency to focus on points of disagreement, and this gives a negative impression of the book. This is not my intention, and not my opinion. Overall I thought this was a very useful book. The style was clear, although the author goes a bit over the top when he constantly organises his discussion using the "first, second, third" framework.



Appendix: A model of the the insurance market, including adverse selection and a critique
(From Barr (1993) p. 111 - 131)
Note: I think there is an error in the third equation on p. 123, above equation 5.17. This is corrected in the following presentation.

* Demand for insurance
y1 = income in bad year
y2 =income in good year
p1 = probability of bad year
p2 = probability of good year

Expected income : y = p1 y1 + p2 y2
Expected utility : U = p1 U(y1) + p2 U(y2)

There are two ways of achieving expected utility: By receiving a certain income (y*) every year, or by recieving y1 with probability and y2 with probability p2. The difference between the average income and y* represents the amount you are willing to pay in order to avoid the risk of income fluctuations (v = Value of certainty):
v = y - y*

The difference is larger the more risk-averse an individual is.
So, the individual will demand insurance as long as the value of certainty is larger than the net cost of insurance (c):
Demand condition: c > v
The net cost of insurance is the gross price (g) minus the expected benefit: pL (where p is the probability that the event will happen and L is the size of the benefits if it happens). In short,
c = g - pL

* The supply side
The gross price (premium) is defined as G, the expected outlay to the claimants is pL, and the costs of operation (administration etc) is defined as T. Hence,
G = pL + T

In short, the insurance company will not sully insurance unless they price they receive from selling insurance is hogher than their outlays and their administrative costs.

* Altogether
We must have positive demand, technically feasible supply, and demand and supply must intersect (it is possible to have damnd and supply curves which do not intersect):
Positive demand: v > y - y* (i.e. there must be risk aversion)
Supply conditions: p must be
1. independent across individuals
2. less than one
3. estimable
Intersection:
Recall, net premium c = G - pL
An intersection exist only if

v > G - pL

(Its value must be greater than the price) which implies:

y - y* > T

In other words, individual risk aversion must be strong enough to cover the administrative costs.

Adverse selection
Assume that there are two types of individuals: Low risk and High risk. Since there is a difference in risk, they will have to pay differnt price for insurance; the higher risk, the more you pay:

Gl = pl L + T
Gh = ph L + T

If the company cannot distinguish between high and low risk, the might try to charge a premium based on average risk:

G = (a ph + (1-a) pl) L + T
However, at this price the low risk individuals will find it profitable to buy less insurance, and high risk individuals to buy more insurance. Thus, the pooling eqilibrium is inefficient: Some buy too much and some do not buy enough. Moreover, the pooling equilibrium is not be stable since a company that offered a contract with a lower price and a lower premium will attract the low-risk individuals. Neither is the separating equilibrium always possible (stable). In short, either the result is inefficienty, or the market fails to exist at all.

A short critique
Recall that according to Barr necessary conditions for supply is that p must be: 1. independent across individuals; 2. less than one; 3. estimable
. I do not think 1 and 2 are logically necessary conditions:
On Independence
Inflation is a "common" shock (if there is inflation, we all experience inflation, not only some individuals), and as thus it cannot be insured against, Barr argues.However, if people have differnet probability estimats of the likelihood of future inflation, I still think it is possible to insure against inflation even if it is a common shock. A rich person believing that inflation will be lower than 5% is willing (and able) to bet against some poor pensioners who fear (and strongly believe) that inflation will be 10%. Also, as Barr notes, it is possible to achieve some insurane against inflation using the futures market; which in turn is inconsistenw with the statement that common shocks like inflation cannot be insured against. Barr might be more correct if, as he sometimes does, argues that inflation cannot be insured against completely, but this argument implicitly concedes that independent p's is not a logical and necessary precondition for the existence of insurance.

Estimable probabilities
It is not necessary that p is estimable, only that people think it is estimable. For instance, the probability of different rates of inflation 20 years ahead may be inherently impossible to estimate. Still, people might have different beliefs, and this in turn makes insurance possible. Whether this kind of insurance is efficient is another matter.

[Note for bibliographic reference: Melberg, Hans O. (1997), Justifying the welfare state : Biased but useful (Review of Barr), www.oocities.org/hmelberg/papers/971127.htm]