8/2/2019 Rationing by Waiting Lists
1/46
(Lindsay and Feigenbaum)
8/2/2019 Rationing by Waiting Lists
2/46
Background Queues emerge in:
i) Setting 1: Quantity demanded and quantity supplied
fluctuate so optimal capacity and equilibrium price isdifficult to determine.
ii) Setting 2: If prices are below the market-clearinglevel, queues of demanders will form to ration the
available supply
8/2/2019 Rationing by Waiting Lists
3/46
This paper examines Setting 2
Case where the price is too low.
Even when demand is perfectly predictable anduniform (opposite of the first setting), lines form andgrow until the
expected wait = value of goods received
8/2/2019 Rationing by Waiting Lists
4/46
Definition of Waiting Queuing using waiting lists
Does not imply waiting in person
No cost in terms ofwasted time. One can do whatever he wants with his time except
enjoy the services of the good sought.
Question: How does the market clear?
8/2/2019 Rationing by Waiting Lists
5/46
Assumptions 1: Delay in receipt of a good can lower its value to
demanders It is this diminishing value rather than increasing cost of
obtaining such goods that produces the convergence ofquantity demanded and the quantity supplied.
2: Individual demand is unpredictable from period toperiod If demand is predictable, the demander can forecast his
desired quantity in each future period and simply order in
advance to obtain it. No diminished value for the good because there is no delay in
between the time the consumer wishes to consume the goodand the time the consumer can actually consume the good.
8/2/2019 Rationing by Waiting Lists
6/46
Section I: Theory Individual decision to join the waiting lists is described,
then aggregated by market
Sensitivity of the rate of joining to:
Expected delay in delivery
Rate at which demand diminishes with delay in delivery
8/2/2019 Rationing by Waiting Lists
7/46
Section II: Empirical results Theory is tested on the data from the waiting list for
admission to British National Health Service (NHS)hospitals.
Theory is found to be in contrast to currentexplanations of waiting lists in the NHS andelsewhere.
8/2/2019 Rationing by Waiting Lists
8/46
Current theory Current theory: Market does not clear due to backlogs Total demand Total supply because total demand includes
backlog
But rate at which services are demanded in each period= rate at which services are supplied,
There is no long-run inadequacy of resources to dealwith demand.
8/2/2019 Rationing by Waiting Lists
9/46
Implications Solution to this waiting list problem is through short-
term efforts
However, expansion of facilities typically does noteliminate waiting lists, or even substantially reducethem.
8/2/2019 Rationing by Waiting Lists
10/46
Section IA. Individual Joining: Join a waiting list when
PV of the good when delivered > Cost of joining thequeue
= C
8/2/2019 Rationing by Waiting Lists
11/46
When does the individual join the
queue?
8/2/2019 Rationing by Waiting Lists
12/46
Value functionValue of the good/rights:
i) Depends on price.i) Assume that upon arriving at the top of the waiting
list, each demander is entitled to purchase a fixedamount of the good at a price (possibly zero) below
the market-clearing price.
8/2/2019 Rationing by Waiting Lists
13/46
ii) Depends on the delay, expressed asgwhich
incorporates both: A) Discount rate effect. However, the impact ong is
insignificant as the delays in the empirical tests rarelyexceed several months in duration.
B) Diminishing demand effect.
The timing of delivery may affect a goods value due tofashion, circumstance, location, health or whim.
Major component ofg since the main thrust of the analysis isto predict the influence of differences in decay rates indifferent queues.
8/2/2019 Rationing by Waiting Lists
14/46
This gives us:
wheret = expected date in delivery
g = decay rate = vector of unknown attributes
p = delivery price
8/2/2019 Rationing by Waiting Lists
15/46
Cost function Cost of joining the list, ci =
Any costs incurred to qualify for joining other than thepurchase price
E.g. taking examinations, obtaining approvals andreferrals
Transactions costs (e.g. expenditures for transportation,legal advice, market information)
8/2/2019 Rationing by Waiting Lists
16/46
Market determinant of t
8/2/2019 Rationing by Waiting Lists
17/46
How does the market clear? Instead of clearing the market by raising the cost of
obtaining the good, waiting time clears the market bymaking it less valuable.
If a good is distributed to a population with varying vandg
Demanders with high values and low decay factors willcrowd out demanders with lower v and higherg.
8/2/2019 Rationing by Waiting Lists
18/46
Rate of joiningAssume that the purchase price and the cost of joining
as uniform across all persons, so:
Variation in t is attributable to:
Decay rateg
Vector of consumer attributes
8/2/2019 Rationing by Waiting Lists
19/46
j(0): Number of people that will join when t=0 (i.e. no
expected delay in delivery)
At t=0, the number of people is unaffected by changesing
At t>0, then an increase ing reduces the rate of joining
8/2/2019 Rationing by Waiting Lists
20/46
Elasticity
Where v= elasticity of joining wrt value placed on the good by the marginal
joiners
Substituting (3) into (2) we getFor any given expected wait, the responsiveness of those joiningwaiting lists to changes in this wait will vary positively with the demandelasticity and the decay rate.
(2)
(3)
(1)
8/2/2019 Rationing by Waiting Lists
21/46
8/2/2019 Rationing by Waiting Lists
22/46
Number in QueueAt equilibrium, the number in the queue Q is equal to
the joining ratej(t)*t.
An increase in supply reduces the equilibrium wait (seeFigure 1).
A change in expected wait should therefore have thefollowing effect on the number in the queue:
8/2/2019 Rationing by Waiting Lists
23/46
Effect of supply shifts on number in
queue So supply shifts has the following effects on the
number in the queue:
Waiting lists will not be shortened because an increasein supply results in a longer waiting lists, due to
decrease in expected delay. This holds for as long as
|elasticity of joining wrt to expected wait| >1.
8/2/2019 Rationing by Waiting Lists
24/46
Equilibrium Equilibrium is at , where
Note:
The joining function allows us to derive an expressionfor the elasticity of joining wrt expected wait in thequeue.
Increases in service rate is occasionally accompanied byincreases in joining rate (since expected wait decreases),
assuming elasticity >1.
8/2/2019 Rationing by Waiting Lists
25/46
Part II: Empirical Results National Health Service (NHS) in Great Britain.
It relates the rate at which demanders of hospitalservices join waiting lists to:
Expected delay, t
Decay rate,g of demand for these services
8/2/2019 Rationing by Waiting Lists
26/46
NHS Features of the NHS (or how to join an NHS
queue):
For non-emergency cases, Patient -> GP -> Consultant/Hospitalisation (placed in queue)
Consultants may only be visited if the patient has a referral.
8/2/2019 Rationing by Waiting Lists
27/46
Rationing by waiting lists (1) Over time, some who are in the queue would
have recovered/moved away/died while awaitingtreatment.
(2) Expected wait itself reduce the attractiveness ofjoining in the first place
8/2/2019 Rationing by Waiting Lists
28/46
Implications of Theory Rate of joining is positively related to
1) The value of the services provided, v
Rate of joining is inversely related to: 1) Expected delay, t
2) The decay rate,g
3) The cost of joining, c
Elasticity: Membership in the queue is positively related tothe rate of service where |t| > 1 and negatively related tomembership in the queue where |t| < 1
8/2/2019 Rationing by Waiting Lists
29/46
Data Base 14 administration regions
Data on mean waiting time and the number ofdischarges:
Reported annually for each ICDA disease category
Reported by region
Data of for the year 1974
22 conditions observed in 14 regions for a total of 308observations
8/2/2019 Rationing by Waiting Lists
30/46
Data shortcomings Data aggregated to regional level masks the
intraregional variability in waiting times and theother variables
However, the demanders are not restricted to a singlehospital, but may shop among alternatives in theirregion for hospitals offering the shortest wait.
Therefore the differences in hospital waiting lists withinthe region is assumed to be small relative tointerregional variation
8/2/2019 Rationing by Waiting Lists
31/46
Separate queues for admission are not explicitlymaintained in each hospital for separateconditions
Beds can be assigned for a variety of conditions.
However, since the waiting times for different conditionsvary greatly, it suggests that separate queues areimplicitly maintained.
8/2/2019 Rationing by Waiting Lists
32/46
Assigning separate decay rates The decay rate,g, for individual hospitalizable conditions is
not objectively measured.
Hence, decay rates are assigned by grouping them.
Categories are given separate decay rates
8/2/2019 Rationing by Waiting Lists
33/46
Category I conditions (high decay rate): Nonemergency cases, typically susceptible to drug therapy for
which alternatives to hospital care were available Cases that respond to treatment and are controlled within a
reasonable time in most instances, even if hospitalization is notprovided
Category II conditions (low decay rate): Nonemergency cases such as hernia or cataracts that do not grow
worse with delays in treatment but for which no alternative tohospital based therapy is available
Category III conditions (negative decay rate): Conditions that rapidly grow more serious over time Negative decay rate as demand increases rather than subsides over
time
8/2/2019 Rationing by Waiting Lists
34/46
Hypothesisj= j (t,g,v)
j/t < 0, j/g < 0, j/v>0
Estimate j = a0 +a1t +a2g.t +a3v +u
where
a1 < 0, a2 < 0, a3 > 0
8/2/2019 Rationing by Waiting Lists
35/46
Dummy variables:
Dummy variable 1 indicates inclusion in the high decayrate, Category I
Dummy variable 0 indicates inclusion in the low decayrate, Category II
Category III conditions removed from sample These emergencies are moved to the head of the queue and do not
follow the process outlined in the papers theory
No influence on the results predicted for the remaining categories
8/2/2019 Rationing by Waiting Lists
36/46
Shortcomings of tests No data on numbers actually joining each queue in
each period
Model suggests that in equilibrium the rate of joining =rate of output s
So data on service rates per period are used instead
8/2/2019 Rationing by Waiting Lists
37/46
Since rate of supply may also be influenced by
delay (not independent oft), OLS estimates ofthese coefficients may be inconsistent.
To reduce this inconsistency: Used predictors oft that are uncorrelated with the
disturbance term u
Predicted values oft are used in the joining equation to
obtain unbiased estimates of its structure Delay is structured in weeks
8/2/2019 Rationing by Waiting Lists
38/46
Different queues will have difference in the
number of potential joiners, j(0). This affects both the estimated constant term and the
slope of the joining function
Separately estimating regressions for each queue and
identifying these constants Deflate the dependent variable of one queue by the ratio of
the constant terms
This makes the two different queues comparable
8/2/2019 Rationing by Waiting Lists
39/46
Empirical results Demand
Supply
Elasticities Decay rate, g
8/2/2019 Rationing by Waiting Lists
40/46
Demand i) Demand (joining rate)
Proven that j/t < 0 and j/g < 0
Proven that there is a difference in decay rates betweenCategory I and II conditions
8/2/2019 Rationing by Waiting Lists
41/46
Supply ii) Supply (cases treated per 1000 population)
The longer the expected wait per condition, the higher isthe rate of output, s/t > 0
Beds available per capita and doctors per capita have apositive effect on supply of services
8/2/2019 Rationing by Waiting Lists
42/46
Elasticities iii) Elasticities (computed from joining and supply
equation coefficients)
Elasticity of joining wrt expected wait is lower for lowdecay rate conditions than higher decay rate conditions(both are negative)
Elasticity of supply wrt delay is positive.
8/2/2019 Rationing by Waiting Lists
43/46
Decay rate, g iv) Estimates of demand decay rates
Mean expected delay for Category I was lower than forCategory II conditions
According to American and Canadian studies, priceelasticity of demand is very inelastic at low prices
The money price of hospital care is zero under the NHS so thisconclusion is relevant
The decay rate for Category I conditions is greater than forCategory II conditions
8/2/2019 Rationing by Waiting Lists
44/46
ConclusionWaiting list queues function as rationing devices
Membership in such queue itself imposes no cost,
so the waiting lists may ration only through theinfluence of delay on the value of the servicedelivered Rates at which demand decays over time were found to be
positive for both categories Category I decay rate > Category II decay rate
8/2/2019 Rationing by Waiting Lists
45/46
Homogenous households Comparative static predictions about the response
of such queues to changing market conditions ispossible.
j/t < 0
j/g < 0, and
s/t > 0.
8/2/2019 Rationing by Waiting Lists
46/46
Non-homogenous householdsWhere markets serve households for whom
demand diminishes at different rates, rationingwill occur on the basis ofdecay rates as well asvalue
People who value the good less might obtain the goodbecause others (who value it more) are discouraged bythe waiting time and do not join the waiting lists