University planning, decentralization and resource allocation

UNIVERSITY

PLANNING, RESOURCE

DECENTRALIZATION ALLOCATION

AND

STEPHEN A. HOENACK. PATRICK D. MEAGHER. WILLIA~I C. WULER and RONALD A. ZILLCXTT Management Information Dl\lslon.

Umverslt)

of Mmncsota.

Mmneapolls,

Mmnesota

_ _. 534.~3

This paper proposes a system of resource allocation m unlversltles which addresses both the prohlcm of useful slmulatlon of altcrnatlvrs and efficlent substltutlon of resources It IS the authors’ blew that the rcasonb \I hq various university plannmg models proposed in the last few years have not been found useful In pracncc 1s that they take inadequate account of the decentralzcd nature of deaslonmakmg wlthm unlversltles. In this regard the three main pomts of the paper are’ (I) universltl plannmg models should have flcxlble aggregation categorlea which can bc determlned and re-adjusted by the deaslonmaker as he uses It. (2) resource constralnts m umversity planmng models should bc IleKlhlr and negotiable. and ncpotlatlon over resources should he part of the analyst’s domam of myulry; (31 planmng m unlvcrsltles must take mto account the problem of mtroducmg mccntlves uhlch cause the behavior of Indlvlduals wlthm the umverslty to be dlrected to the needs of the m\tltutlon‘s cllentele~

I. INTRODCICTIOR Limited growth and financral constramt,

it is apparent. are the most pressmg problems of higher education m the decade of the 1970s. Recent experience has shown that sound planning principles are of necessity usurped by hasty ‘retrenchment’ programs when umversity managers face contracting budgets armed only with resource allocation procedures developed in the expansive environment of the 1YSJs and 1960s. At that time. it was possible to honor nearly all proposals for additional funds to some extent. now. it may no longer be possible to do so and still fund any one of them adequately. Hard choices must be made and relative values placed on alterrrative courses of action. In any financial environment. needless to say. good resource management could be useful. However. linancial constramt aggravates three problems of higher education which make good management a requisite to survival. Fn-st. and most immediate. 1s the problem of relative growth: which university programs should grow. which should not, and which should contract? Second is the problem of Rexibllity: how should funds be allocated to develop new programs so that proper responses can bc made to changing demands of umverslty constituencies? Third 1s the problem of self-interest of umvcrsrty departmental administrators: how can they bc encouraged to use resources efficiently and consistently

with

articulated

university-wide

goals?

* A good descnptlon of most of these models IS probIded hy Lamson and Powel [la]. The models adcanccd m the many publlcatlons of the NatIonal Center fol Higher Educatlon Management Systems m Boulder. Colorxdo. ha\e rcccl\rd con,ldrrahlc uttcntlon In recsnt jeara 751

Today’s utmcrslty manager is concerned less with the budgetary consequences of educational decisions than with the educational consequences of budgetary constraint. He must operate on tbo levels, one mformatlonal. onejudgmental He must have access to II$W~trfro~ which shows quantitative consequences of alternative actions that are under his control. and he must spcclfq crltcrrc~ relevant to his own situation. which arc both objective (dollar inputs and program outputs) and sub]ectlve (values) m order to choose among these actions. Usually the relevant aggregation categories vary from program to program and decision to decision, and the criteria vary. often substantially. among different declsionmakers. It is obvious, therefore. that resource allocation plannmg tools based on standardized aggregation categories and criteria cannot be useful to university declsionmakers. Then only use is likely to be assisting the Federal government m monitoring data on federally funded uniberslty activltlcs within aggregation categories which do not neccssarlly place umversltles m a reasonable perspcctlve For this reason alone a number ofresource rcqulrements models which have been devclopcd and advanced m the past few ycarb arc unlikely ever to be helpful *ithin unl\crsities * A further and more slgnlticant problem is that even If these models were adequate m a mechanistic sense. that IS. even lf they simulated helpful alternatives. they would not provide incentives for desirable resource substitutions. The large majority of significant resource substitution decisions in umversitles are made at hlehly decentralized levels. In order to lmprobc declslonmaking at these levels a combination of both the simulation of helpful alternatives and the mcentlves to choose alter-

‘58

S. A. HOENACL P. D. MEAGHTR.W C. WAILERand R. A. ZILLGI~I

natives supportive of articulateduniverslty-wide goals 1s needed. This paper discusses a system of resource allocation which addresses both the problem of useful simulation and efficient substitution of resources. Because various models proposed m the last few years have not been useful m practice, we have developed a model which hopefully avoids many of its forerunners’ inadequacies. We are attempting, therefore. to meet two goals. First, a model’s cost-of-use should be commensurate with results. Often, for example, when analyzing a small operation, the answers could be obtained very inexpenslvcly on the back of an envelope.* When used on a very large operation the model must be directed specifically to the subtleties and trade-offs of concern to the declslontnaker. Second. a model should simulate alternatives relevant to decentralized authority; alternatives based on realistic aggregation and on flexibility in defining categories of resources and ObJectives. In order to meet these goals we have taken the position that planning m the level of detail exemplified by existing models can and should be done only at the departmeti level Accordingly. this paper proposes a ‘departmental planning tool’. Since planning at the collegiate and allumvcrsity level must deal with a range of issues far beyond the scope of any informational model, includmg our own, we have suggested a process and certain crlterla by which departmentally generated informatlon can be evaluated at these levels. However. for resource substitution, no information generating system is sufficient in itself; efficient substitution can be guaranteed only by the provision of economic incentives consistent with the maximization of articulated university-wide goals. Accordingly. the concluding section discusses the design of an incentive based budgetary system which depends primarily on the elimination of recurring budgets. Only within the context created by such a system can the university manager use resource allocation models to improve decisionmaking.

2. DEPARTMEKTAL

PLANNING

This section describes a proposed method for departmental planning, in three parts. The first part describes the basic concepts underlying a ‘cost&output schedule’ which shows what 1s Judged to be the maximum departmental output available for several levels * A practical ~x~lys~s of costs of computation for the basic resource requlrcment predlctlon model advanced bq the National Center for Higher Education Management Systems has been prowded by Hopkms [ 133. t Ordmarlly departmental outputs are numerous and the practical departmental plannmg tool described m the second part of this sectlpn takes multiple outputs mto xcount Howcvrr. III order to keep this dIscussIon of basic concepts as simple as possible. single output cases ~111 be consldered here

of funding. The second part presents a simple numerical example of a cost-output schedule. The third part presents a departmental departments to construct

plannmg

tool

that

enables

such a schedule by exploring quickly the results of a wide range of altcrnativc uses of faculty resources. the corresponding costs, and the meeting of objectives.

Once a department’s budget is tixcd for a particular fiscal year a host of decisions need to be made concerning hiring, lay-offs. work assignments. and so on. Associated with the decisions finally made arc a series of instructional, research, and service outputs. some of which are hard to measure. We regard outputs as those which are planned by the department head. In the case of instruction. these outputs can ordinarily be readily measured in units such as student credit hours taught. However. m the case of research. the only rcadlly available measure is the input of faculty research time. Probably this measure is also the only useful one. Department heads must always make subjective Judgments concerning the likely result, and the value of the result. of an indlvidual’s expenditure of a given amount of research time. This visually cannot be quantiiicd. and even if it could bequantified. a department chairman often would not want his views on mdlvldual faculty research productivity to be known. It is our view that department chairmen can plan just as well with the research time proxy for research output as with any measures of the output. Needless to say, however, anyone ambitious enough to do so can attempt to meusurc the quantity and value of research output Theoretically. for any glvcn level of funding there is a set of outputs corresponding to each set of actions that the decisionmaker could take. These actions would obviously reflect the values of the declsionmaker. One could visualize a table showing outputs that one could expect under a vdrlety of funding levels.

Fundmg level I$) 0 50.000 100,000 200.000

Instructional output 0

Research output 0

Other outputs 0

We will call such an idealized table a ‘costGoutput schedule’. Throughout the remainder of this discussion we will use the terms ‘cost-output schedule’ and ‘costoutput curve’ mterchangcnb1y.t 7% .&LzI)~~ of tlz~ CUI’LXJ. We bould expect the shape of the cost-output curve to look somethmg hkc the

Unlverslty plannmg. decentralization and resource allocation

259

output

output

0

cost

FIN 1. A typlcal cost-output

curve

one shown In Fig. 1. The segment S, corresponds to what many call building up to ‘crltical miss’. That is. it takes a certdm minimum faculty to offer a program at all, and for a time, demand may not grow very quickly. Later, however. output may grow quite rapidly without requiring many additional resources. The segment S2 illustrates this growth. Finally. the curve (segment S,) takes off on a line flatter than .S2. but not so flat as S,. Actually. we would prefer to have the line travel along the path of S, where output mcreases faster than costs, and we experience economies of scale.* EJficicncy It may not be unusual for a department to be operating less efficiently than It can. As an illustration consider Fig. 1. The curve OS shobs the relationship between output and cost under the assumption that departmental resources arc utlhzed as efficlently as possible. By efficient use of resources we mean that the department has carefully assessed the relative effectiveness of resources in achievmg departmental outputs, and has deployed these resources accordmgly. (In the second part of this section we ~111 illustrate a practical planning tool which can be used by departments for the purpose of determmmg efficient uses of resources,) Points below the curve OS represent opportunities for increasing output with no corresponding increases in costs, or for maintaining output at lower cost. The pomt P m Fig. 2 illustrates that situation. Efficiency here means that more output could be obtamed for the same money or the same output for less money. Growth. The amount of money a department needs to grow depends on the relatlonship of its current operation to its cost output curve. Figure 3 illustrates this pomt. A department which 15 usmg its resources efficiently to provide level one (~5,) output is represented by P,. This department needs an amount E m order to grow enough to provide level two (L2) output efficle‘ntly A department providmg level one output inefficiently 1s represented by P,.

* At some pomt the CLI~W maq Rattcn out agaIn u hen dlsrconomlcs of scale begm to occur

Fig 2. Eficlency and the cost&output curve

This department needs no additional resources to provide level two output because the increased output can be achieved through better use of currently available resources. An over-worked department providing level one output. represented by PO. needs the most resources to provide level two output efficiently. In this case additional resources are required both for the increased output and for alleviatmg excessive workloads. Ddir7c. Trlmmmg back a department’s operation also depends on the relationship of operation to costoutput curve. but equally important is the phasing of the cutback. In the illustration shown in Fig. 4, an operation could be cut back from the level indicated by the point P to another level Q in any number of ways. The diagram shows two paths. one followmg the costoutput curve and the second running somewhat below it at planned inefficiency. The point R, on the cost curve corresponds to a big bite, B, on the cost axis. The point S1. which is at the same output as R 1, corresponds to the little bite h, on the cost axis. Planned reductions may often of necessity fall below the cost-output curve. Adjt~~tr~~ts to &sirrd output /erels. It is worth notmg that while it is desirable for departments to operate on the cost-output curve OS. representing efficient resource use at a desired level of output. it may be necessary to operate off this curve while reaching the desu-ed output level. For example. it may be impossible A

output

FIN. 3. Growth and the costPoutput curve

‘60

S

A

HOI

NA(

h.

P.

MI AGHI K, I&’ C. WI

D

Table

Output

1’

/*

3 3

curw

to hire faculty of the desired quality or quantity over the short-run. thus necessttating overworking other faculty. Or it may be tmpossible over the short-run to fill sections with students, thus necessttating operating below the costtoutput curve.

This section illustrates, in a simplified way, how a department chairman might develop his cost--output schedule. Suppose that the mttlal operating level of a department is that shown in Table 2. The department chairman is dissatisfied with the amount of research time available to junior faculty; however, he is unable to hire additional faculty due to budgetary restraints, and therefore decides to consider reallocating resources in order to free time for junior faculty (faculty level one). He first estimates the amount of time. m his judgment, that a faculty member needs to teach a course at a given level. Table 3 contains his estimates. expressed as the percentage of full time required to teach a section. The reason for the lack of research time for junior faculty 1s apparent from the following calculation: (50 per cent time at course level 1) x (15 sections at course level I) * Full time equrvalent.

Table 2. Initial operatmg

1

Average class stze (students)

I

2

3

‘i

50 15 25

50 33 25

75 33 33

75 50 33

ttmes sections)

at

per cent time at course level 2) x (5 sections at cour5e level 2) = 7.50 + 2.50 = 10.00 FTE* faculty. Thus. there is a workload equivalent to ten FTE faculty at level one and only ten people to perform it, leaving no time for research. Similar calculatrons indtcate 34 per cent time at faculty level two and 33 per cent at level three is free for research, which this department chairman considers about right. Clearly, one way to provtde research time at level one is to cut back on the number of sections taught. The chairman makes arrangements shown in Table 4. This arrangement provides 35 per cent time available for research at level one, 34 per cent at level two. and 33 per cent at level three.

Secttons

I

2

15 5

5 5

20

Table 4. Allocatton Faculty level

Secttons

I

I ? ; Total. Average class stze output

of faculty

at course level 3 4

7 8 5

5 5

13

10

4 6 10

50 650

40 400

30 300

2 Y IO

teachmg

loads

Total secttons

No. of faculty

13 16 14 43

10 8 7 25

10 100

1450

To the chairman. research time is now adequate. but he 1s troubled by the fall in instructional output. He is unwtlling to make any changes in class size at upper division and graduate levels (course levels 2 3 and 4). but he decides to allow average class sizes at course level of tllustrattve

at course lewl 3

department

4

10

4 6 10

3 x 10

50

40

30

10

1000

JO0

300

100

Total secttons 20 16 14 50

output (students

courses

Percentage of full ttme reqmred to teach a course at level

1

Sl

to teach

+(50

Ftg. 3. Declme and the cost-output

2 3 Total

of effort requtred vartous levels

Faculty level

/I:

Faculty level

ZILLGITT

and R A.

3. Esttmates

,,$ Rl ’ // /’ /* ,/’ /’ Q ,/./*

ILLR

1800

Number of faculty IO 8 7 25

Unwers~ty

plannmg.

decentrallz~tlon

cost Fig 5 Derwatlon

of pomts on the cost-output

curve.

level I to increase to 60. This will give him an additlonal 130 units of output for a total of 1580 at all levels. Schematically. the choices of the department head look like the cost-output schedule in Fig. 5. His initial position (1) produced the most mstructional output, but at the expense of research time for junior faculty. His next position (2) solved that problem. but reduced output too drastically. Posltion (3) is the point on his cost-output schedule representing all the concessions he is willing to make in each of the decision areas he IS considering. It should be noted that cost was held fixed throughout this process. but the same kind of decisionmakmg would allow the chairman to find the curve at a lower level of cost. Thus, the chairman might move from point (1) to point (4) in response to a budget cut. We have shown only two dimensions of the chairman’s cost-output curve. but there are others. This example considered only the trade-off between classroom instruction and research. Obviously. the planning procedure could be generalized to consider the trade-off between research, undergraduate mstructlon, public service. service load teaching, degree credit teaching. and so forth. Departmental planning then becomes a more complex multidimensional problem in which a computer can be very useful.

More generally, departments within a university can make use of the cost-output schedule concepts to lmprove their own eticlency by determining the maxlmum mstructional output they can produce at a given * An Important attrIbute of the DPT 1s that the user can select the time horizon relevant to his decisionmaklng needs. the lmmedlate quarter or acadennc year. or a time period defined further in the future. Longer time periods can be useful for plannmg relanve to changes m the number of faculty posItIons avaIlable due to retxements and tenure decIsIona and for plannmg relative to prechcted changes in student demands. t Note that the ‘costs’ are of research mput tlmr Research output IS not valued m any objectwe fashion.

and resource

allocatIon

361

cost. conststent with their own objectives. The systematic method by which they construct this schedule, which we call the department planning tool (DPT). enables them to schematize current instructional resource deployment and to explore quickly a wide range of alternative allocations of resources. The tool has two Important characteristics: 1. It assumes no fixed relatlonshlps between resources and mqrnctlonal output: this IS a matter of departmental Judgment. It allows the user of the tool to impose his own choices, such as: (a) Assigning specific staff to specific instructional. research, or service functions. (b) Emphaslzmg certain forms of Instruction (e.g. graduate mstructlon). (c) Choosing the time context of his analysis * 2. The user IS able to make a range of choices and to see the results of these choices in terms of: (a) Satisfaction of student demand. (b) Faculty workload. (c) Section size. (d) Total and average (per unit) costs. (e) Availability of research and service time. The first characteristic of the DPT insures that the department head himself explicitly makes the assumptions and defines the conditions on which the cost calculations are to be made. The departmental planning tool provides a convenient framework for recording each assumption and condition and for making the calculations once the assumptions and conditions are determined The data which are needed for determining the cost for a level of instruction are as follows: ( 1) The fraction of each faculty member’s time spent at teaching (by level). at research and at public service or other activities. (2) Each faculty member’s salary (the departmental planning tool permits aggregating faculty by rank If preferable). (3) The number of sections offered at each level. (4) The desired average section size. The most Important assumption IS the first: the allocation of faculty time. This 1s a workload assignment. and the fraction of a professor’s time assigned to a course level multiplied by his salary is allocated to the cost of that course level. Thus. the department head, using the departmental planning tool. can determine the implicit value of faculty time spent doing research and public serv1ce.t If a department head wishes to attach no value to these activities. he does so by showing 100 per cent assignment of all faculty time to instruction. Whatever the allocation. the instruction and research functions arc given dollar cost figures, and any interested observer can learn the precise assumptions under which the costs were determined. Clearly, then. the existing allocation of faculty time for a department produces a level of output at a particular cost. However. there IS no L( pr~orr reason to assume that this 1s the maximum output consistent with departmental objectives which can be produced

262

S. A. HOENACK,P D. M~AGH~K.W. C WEILERand R. A ZILLGITT Table 5. Departmental

Course level Satisfied demand Unsausfied demand* 1 Student 2 Instructor 3 Asst. Prof. 4 Assoc. Prof. 5 Professor Total sections Section size

I

2

3

4

700

300

250

96

300

100

50

15 5

‘0 35

’ 4 2 2 10 30

8 2 10 25

plannmg

Numbers of faculty (FTE)

3 5 8 12

6 3 3 9 9 30

worksheet

Sect1onsfdculty

‘5 ‘3 13 1.4 10 16

Faculty research/service t1mr (FTEI

I.0 3.2 6.7 II 4

*The estimation of unsausfied student demand 1s Important for many. perhaps most. uses of the DPT Student registrauon systems can be des1gned to provide accurate student demand data. although the expense is not necessatxly justified. At the University of Minnesota. students arc rat1oned among courses according to random selection of alphabetical segments within which student last names fall We are thus able to estimate total course demand based on a frequency distribution of last names among the alphabetical segments. Ad IIOCmethods ofestImating student demand can often be developed w1thout the addmona expense of a pre-registration system or survey.

for that cost, or equivalently, that the same output could not be produced at less cost. The departmental planning tool allows quick analysis of alternative faculty assignments and shows the associated cost. thus enabling the department head to examine a large number of ways of providing a given output and select that one which satisfies his objectives at the lowest cost.* The essential elements of the systematic analysis of alternative choices provided by the DPT are described in the following example. Use of a computer permits both a more detailed description of one’s choices as well as exploration of a wider range of alternatives. However, a computer is not necessary. and the following example can be worked out with paper, pencil, and a desk calculator. Table 5 shows the status of an academic department resulting from one deployment of its current faculty. The column headed Faculty shows that the department has 30 facultiesat five levels: six at the first (lowest) level, three at the second. and so on. The row labeled ‘Total sections’ shows that the department teaches 20 sections at level 1. 10 at level 2. 10 at level 3 and 8 at level 4. The array bounded by the ‘Faculty’ column * The most rfic1ent situation will be stated both m terms of cost versus outputs and 1n terms of meeting departmental objectives. It may well turn out, for example. that surface efficlencles 1n terms of lower costs are more than offset by increases m cost at other course levels, substantial reducuon in research time. unmet demand. or overly large section sizes. t In practice. all adjustments are converted to integer values when dIsplayed to the user. $ Column moves and array moves are represented m a similar fashion. However. changes in array values are far reachmg, involving changes m total sections. satisfied demand. and sections per faculty member.

and ‘Total sections’ row shows specific assignments of faculty at the five levels to courses at the four levels. At the head of each column correspondmg to the four course levels are figures showing the total demand for courses broken down into satisfied and unsatisfied demand given the faculty deployment. At the foot of each column is the average section size, which IS assumed to be a parameter determined by departmental policy. In order to generate possible alternatives. the department chairman can make ‘moves’ which allo& him to change all values m a row or column at one time. or change selected values m the assignments array. To properly evaluate his alternatives, he should make only one move at a time. an array move, however, allows him to change any or all values in the assignment array. A row move is defined as changing the sections per faculty member for a given faculty level. This obvrously changes the number of sections that can be taught utilizing the existing faculty. The new values m the array are computed in proportion to then previous value. Thus. if sections per faculty member at level 3 increase by IO per cent. all values m the array corresponding to level 3 increase by 10 per cent, and with section sizes held fixed at a policy-determined level, total sections taught will also change.t Note. however, that a row move which changes section sizes alone will change only the satisfied demand figures: and not total sections taught. Nor~-111strt1~ti0l1(~/ ~CS~II~CL’S. There are certain areas of non-instructional resources which the DPT 1s not intended to calculate. The departmental planning tool is intended only to develop and evaluate alternative deployments of faculty resources. It should be emphasized that while the procedure shows instructronal outputs, there IS no attempt to quantrfy research or service

Unlverslty

plannmg.

decentrallzatlon

outputs. Thcrc are no standard units with which to measure these outputs. What the procedure does show is the change in time available for research and service as the result of a given set qf choices. It would not be practiclll to habe all departments use a similar procedure to handle supplies and equipment or support personnel Some types of departments, chemistry and anatomy. for example. have relatively heavy supply and cqulpment expenses. for obvlouslq good reasons In other departments. for example veterinary medlcme. support personnel Includes not on11 secrrtarlal help. but cm be substituted to some extent for faculty in the research and teaching laboratories. Smce the manner m which supplies. equipment, and nonacademlc personnel rxpcnsc IS incurred varies considerably among departments, each department can provide an analysis establishing the relationship between these Items and Its instructional function. Then. usmg the plannmg procedure. quantities and costs Incurred for these Items can be shown for each level of departmental output. The DPT IS not intended for use in resource allocation within non-academic admmlstrative and support units. Generally. these umts vary so much m then activities and orgamzatlon that probably no single evaluatory procedure exists which can handle them all. Many of these units produce services which are available commercially, and In such cases an evaluation can be based on alternative market prices. 3.

COLLEGI4TE PLArLNING AUD DECISIO\~lAKING

CENTRAL

The three parts of this section discuss the integration of the departmental plannmg tool into collegiate and further centralized le\cls of university governance. In

* A thoughtful appllcatlon of formal programmmg models to departmental and colleglatc decwonmaking at Io\va State Llnlcersltj 1sgvrn by McCamleq [l6]. Excellent general discussions of the economic context of programmmp models. and the programmmg context of economic analyses are provldcd rcspcctlvcl> by Dorfman, Samuelson and Solou [5] and Bnumol [2]. A general descrlptlon of decompwtlon prmclpler \\hlch guldc the appropriate relatlonshlps hctuecn collcglatc and departmental plannmg IS probIded m Chapter 23 of DantrIg [3]. $ Soctlon 3 dwxsaeh this fact In mow detail m the context of how the monopoly power often cnnfcrred upon departments m stdtc nnlvcrsltw can Icad them to be able to &ate costs hecausc they arc the only subsldlzed source of serwce to state rwdent5 1 Our approach defiers from the model described by Geotirion. Dyer and Ftxnberg [S] 111two ways (1)The dew sionmaker usmg our departmental plannmg tool selects alternatl\r rcsonrcc .~llocatlons wthout evpllcltly consldermg his prefercnccs He dots. howrbsr. observe the consequences m terms of values of Indlvldual arguments m his objectlvc filnctlon He can make lrnpllclt wclghtmgs If he wshes. In the Gtx~ffr~~n ct ul model these wclghtmgs are formulated for choices actually tlndcr conalderatlon (2) An Important aspect of our approach is that resource constramts arc Hc\~blc ,~nd negotlablc

and resource

allocation

363

the first part, we discuss our general approach to a LIIUversity-wide plannmg process and compare our approach to that of other formal planning models. The second part describes m detail a process for planning at the collegiate level which complements use of the departmental planning tool. The third part considers the role of the central decisionmaker m an administrative planning process based on these departmental and collegiate tools.

Pws~~~~~~~Io~I of thr plnrzrhy process. The departmental planning tool IS designed to be an integral part of a planning and declsionmaking process that extends to the collegiate and centralized levels. This process can best be visuahzed as a pyramid, with the DPT as its base, in which information about alternative resource allocations is increasingly condensed. As the process evolves. only the mformatlon necessary to decisions that should be made at each successlvc level IS filtered out and presented to the appropriate decisionmaker. In other words. use of the departmental planning tool condenses the large amount of mformation necessary for departmental decisions into sets of feasible alternative levels of cost and output for presentation to collegiate deans. At the college level a similar process occurs; this greatly reduces the amount of information central administrators need to consider m making decisions. If proper presentation of the planning process is made. it will be conslstentlq and reliably possible for decisionmakers at both the collegiate and central levels to see which alternatlve allocations are judged best and which are rejected and on what criteria these Judgments are based. Compurrsor~,s with ,forrnul propwnrtm~~ rtdrfs. The procedures for departmental and collegiate planning which we outline are very similar in substance to procedures implied by formal optimization models.* The differences between our approach and that often taken m the literature are based on several assumptions. discussed below, concerning issues Involved in actually implementing a planning procedure in the university environment. (I) The concept of a production posslblhty frontier IS m part. at least. determined by subjective and negotiable factors, not by a fixed set of quantities and relationships. This fact is glaringly apparent in universities, as any dean or budget officer will readily attest.? (2) We believe It 1s neither fruitful nor reallstlc. m the beginning stages of a planning process, for deans or department heads to try to formulate explicitly their objective functions in advance of solving a problem.1 One reason is because the relative weights on indlvldual research outputs are extremely difficult for any one decisionmaker to determine for himself and arc almost never agreed upon by different decisionmakers. Similar dlfficultles exist. of course. for the rcscarch and instructional trade-offs. As an alternative. as we suggested m the beginning of Sectlon 3, department heads should be able to observe the consequences of alterna-

161

S. A. HOEMCK. P. D. MEAGHEK. W C. WEIL~K and R. A. ZILLGITT

tlve courses of action m terms of the values of the arguments in their objective functions. They should make their own subjective welghtings and keep them private If they wish. (3) While It IS technically correct that any optlmlzing model can be defined to encompass any aggregation categories its user desires, m fact. most of the large scale higher education models advanced m the past few years have standardized aggregation categories. This obviously limits their usefulness m making deczntrahzed decisions. The difference m our approach and that of the relevant programming literature IS more subtle. It concerns the allocation of the decislonmaker’s time in his involvement with the model In our view. the decisionmaker’s process of thmhing through and defining aggregation categories forces him at the same tnne to think through and specify his declslon problems in reahstlc. operationally meaningful terms This task reprcsents a large proportlon of his analytical effort m using our proposed planning tools.

The plannmg process described below assumes that collegiate and departmental planning are. and should bc, mteractlve. They should occur at the same time. AdJustments at cither level should result m adjustments at the other until a satisfactory equilibrmm pomt is reached. The procedure described IS simple, a little crude. and fairly obvious. We behevc that these qualities are vlrtucs. Complex and subtle planning models at the collegiate planmng level. especially in larger colleges, may be so difficult to use that they would simply not be employed, at least by the right people. It seems prcferable to have a procedure that is smiple to use. reasonably accurate for the estlmatlon of major resources. and easy to understand. For this reason the procedure described here highhghts for the collegiate decision the major elements that characterize the tnission and workload of each department and leaves the subtleties of planning to the department. TIz~JDPT 111co//cl/itrt<, /~ltrr~mt(l As discussed m the previous section. the departmental planning tool can be useful to a department chairman in several ways. It can be employed at an) static level of funding to assess alternative uses of resources to attain certain outputs. It can also bc useful m the budgeting process itself to develop sets of feasible alternative levels of cost and output. with priority ordermgs attached. for presentatlon of a budget request to the college dean. Later, if need be. it can be used to work out the detailed plans necessary to work withm the limits of imposed budgetary constraint. The r,c~/coj t/w cd/cqc~. The primary role of the college in the budgetmg procedure is to establish limits. Specifically. the college may establish the followmg norms and IimitatIons: (1) It may estabhsh dcpartmcntal budget ceilings. (3) It ma? speclrj hkely or dcsn-ed levels of student

demand (e g. by restricting or mcreasmg enrollments). (3) It may assent to or reject the development of new programs. (3) It may mandate the elimination of existing programs. (5) It may specify a desirable level of departmental research and service. (6) It may control the level of service instruction provided to other colleges. (7) It may control the amount of instruction that its own students seek in other colleges. Stage5 ufcollegiatr phr~irzy. These norms and limitations are established in three stages of collegiate planning: estimation of demand for instruction within the colleges; tentative deployment of resources; and negotiation of ditrerences between the college and its departments concerning the deployment I. Est~~tr~rorl c!frlcrflclrzrl--The first task of collegiate admimstratlon is to speclfq likely or desired instructlonal load. This specification must consider two clements; service mstruction, or instruction provided to students m other colleges. and internally generated mstructlon. or instruction provided to its own students. Depending upon collegiate circumstance and ob~ectlves. collegiate leadership may assume either of two styles at this stage of planning: the college may be passive with respect to student demand, and plan to supply whatever students are likely to ask for; or it ma> be interventionist--encouraging new activities. trimmmg others. or funneling students into instructional areas that conform to a preconceived plan. The college will. of course, have to accept enrollments in other units as given. but it may deslrc to institute policies that encourage or restrict the amount of service instruction. The outcome of this stage of plannmg IS an array of numbers showing demand by course level for each department in the college. The form that this array would take 1s that of the familiar induced course load or student preference matrix. Several of these arrays, corrrspondmg to different policy options and operating levels. may be produced for both the short and long terms. 2. Rrsowce deploJrnrnt-The second stage of collegiate planning. tentative resource deployment. IS very similar to the departmental planning procedure described in the previous section. It differs m that the concern of the dean is to establish hmlts. not to make detalled plans; he does not attempt to take mto account as many subtle factors as his department chalrmcn collectl\ely can. HIS first plan is a negotlablc one which objectively reflects his attitudes. goals. and constraints. A display that can aid the dean in estimating departmental requirements is shown in Table 6. The demand figures m the display are derived from the previous stage of planning, estimation of demand. Class sizes. initlallg, should be based on actual experience. but the dean may alter them at will. Similarly, the Initial

.s,- 5: hJ

Notation: L,-Cource level J D, -Department I d,,-student demand at level j m department I c,,-ytveragc class sue. level 1. department (. s,,-number of s&Ions. level 1, department 1( = d,,lc,,)

Table 6. Collegiate

resource

estlmatlon Instructional racu1tj

Ke\carch and XI-\‘,ce Total facultj

366

S A HOI N4C‘K. P D. Mr ,\rizzz R, W C WI ILI K and R A ZILLGITT

number of faculty, average faculty salary. and sections per faculty should be based on experience, but subject

to manipulation. Research and service may be handled in one of two ways: the dean may specify the actual number of faculty involved m this functzon. or he may specify a ‘research and service allowance’ factor. This factor would be expressed as a percentage of a faculty member’s time that one could expect would be devoted to scholarly and service activities. This array shown m Table 6 can serve several purposes. First, it permits quick comparison of all the college’s departments. Second, by altering any element, it permzts the dean to observe the sensmvity of a department’s budget with respect to constraints he might wish to impose. 3. Nryotiatzorz-When the collegiate dean has completed hzs estimatzon of resources, he may provide the department with the demand alternatives he has explored and the constraints he has imposed in estzmating resources. Using these data. the department may use the departmental planning procedure to see what adjustments are necessary to operate within the dean’s framework. It may be that the department and dean differ with respect to the values of the parameters and variables within the scope of the departmental and collegiate planning procedures described in this paper: differences involving factors outside the scope of these planning procedures may also develop. In thus case it will be at least clear that a dzfferent type of negotiation. based on different terms is requzred. If such negotiations lead to resource changes, zt may subsequently be necessary to return to the departmental planning procedure to change its quantifiable variables. (c, CfPztzXzIdrcl.siorlrzzukiry If the planning the departmental information with contend in order

procedure 1s carried out correctly at and collegiate levels, the amount of which central administrators must to make deczszons should be substan-

* The exact numbers would. of course. be dependent on the fiscal envzronment at the tzmc. t Although there are no doubt many criterza whzch could be used to deczde what is the best allocatzon of departmental or collegiate resources. .dlocatzons whzch are labeled clearly infcrzor ought to be readily apparent to any reasonable observer. For example, an mferzor alternatzve should be much more ewpenszve or involve a level of output that. for the cost mvolved. IS unreasonable. In the unhkely case that there IS mdzfference between two alternatzves. there should be a clear explanatzon of why thus zs the case .$ However, zt zs Izkely that the publzc wzll become mcreasmgly reluctant to subszdzre such a program and that zt WIII tend toward the vzew that such students should pay the full cost of then- trainzng out of then own pockets or from student loan programs. The case where student demand for traznmg zn some field substantially exceeds market demand for graduates zs a partzcularly agonizmg problem for a pubhc mstztution. Some students wzll demand to know why they

tzally reduced. Each dean. after appropriate consultatzon with hzs departments. assembles a set of costoutput schedules from the allocations of departmental resources mutually judged acceptable under ahernative budgeting constramts. The schedules show what the collegiate umt would elect to do in terms of output under alternative levels of funding. Therefore, it is possible for each dean quickly to show central adminzstrators what his situation would be in advance of any likely budgeting decision. since the costtoutput schedules would describe exactly. zn quantitative fashion, a unit’s output at. say. 90, 100 and 110 per cent of zts current budget.* Ancillary mputrelated information provided to central administrators might include the mzx of tenured and nontenured faculty to bc used at each level of performance, student/‘faculty ratios. and any long run zmplzcations for items such as physical facilities. Qzznzztzt~ztz~c co17.seqww~.s of ~~hwutiw uctions. Use of the informatzon generated by these decentralized decision procedures enables admmistrators to see quickly the quantitative consequences of alternatzve actions. Some ahernatzves wzll be judged to be better than others for any number of reasons. such as cost per degree, cost per credit hour. cost per FYE student, use of tenured faculty time for tcachmg as opposed to research, or student demand satisfied.? In addition. there are certain other objective measures of departmental programs which are appropriately the concern of central decisionmakers and which would azd them in deciding between alternative levels of funding. These mclude: 1. Drqrces prur/uceu’~Use of degrees granted combined w’ith the use of attrztion data and time-to-completion data are important measures of program quality. Use of this data also makes zt possible to separate clearly the impact of service load teaching from workload generated by a department’s own majors. Finally, the effect of altermg a program’s level of funding can be measured zn terms of degrees produced as well as m terms of mstruction offered. This is essential information zf, for example, it 1s deemed desirable to tune the rate of output to market conditions for trained manpower in the field 2. Job rnurlirts---The existence of job markets is a second objectzve measure zmportant to central deczszonmakers. Students in nearly any public university program receive a conszderable subsidy from the state, usually justified by the belief that the public derives a benefit from the actzvzties of an educated person in excess of those accruing to the person himself. But if the graduate cannot obtain the usual sort of employment for persons wzth his training, the reasons for state subszdies of education stand in danger of being subverted. Absence of a job market for the graduates of a program, however, zs not necessarily sufficient reason for cutting back zf there is stzll considerable student demand and these students have full knowledge of then bleak employment pr0spects.f Student demand

Umverslty plannmg. decentrallzatlon and resource allocatlon under these circumstances, if informed,* indicates either that the program’s training is a consumer good or that the students are willing to take a considerable risk of not gcttmg a job.? Non-c/utrrtrrtut;t~~, tlrc~.\ion.s. If central administrators are probIded with properly condensed quantitative information. they can direct their energies to grappling with important qualitative Issues which directly affect resource allocatlon. In general these Issues require adminl\trators to act on changing values on the part of the univcrslty’s constituencies. Judgments concerning the weaknesses and strengths of individual parts of the institution. estlmatlon of the value of various research and scrvlce actlvltles. assessment of faculty and student preferences. and so forth. These intang~blrs arc‘. ofcourse. not explicitly handled by the quantltatlvc procedures already dlscussed, although it should be clear that some are impliclt in the judgments made by deans and departments in assembling the quantitative Information. Many umversitlcs have become institutions fulfilling what the name impllcs: the offering of a diversity of educatlonal experience covering nearly the whole of the universe of knowledge. It is unhkely that unikersitlcs will soon abandon the idea of universality even though It IS often the case that only a few institutlons can asplre to excellence in all programs. However, to argue that a certain program should exist is to aay nothmg at all about Its size. It IS true that student demand. state subsidies. and Job markets are important external Influences on the size of a program, but Internally. pubhc universities have an obligation to consider other determmants as well. One of these determinants 1s efficiency. There may be such cconomlcs of scale Involved in certain programs that only large umversities would be able to attract enough students or researchers or both to attam an cfficlent size. This concept IS important for

cannot bc admltted and others. once they have graduated. ~111 demand to kno\v \vhq the) were admltted If there were no job opportumtle5. It IS unlikely, though. that m a real sltuatlon the fob market will present a clear and simple pitturc What IS lIkeI!, 15 that thcrc ~111 be Job openmgs for a hmltcd number of graduates. In this wse a combmatlon of subtle pol~c~cs M hlch habe the effect of controlling enrollment a$ a result of Indl\ldual decisions wll probably be more palatahlc to the puhllc than an enrollment ceilmg or prohlhltlcc tultlon fees. which give the appearance of coer-

cion. A comhmatlon of pohc~cs mvolvmg tultlon fees. mformation for proqxctlvc atudents. and admissions standards could be used to ‘tune’ the program to the needs of all w-ved. * Freeman [S] prowdes conslderablc ewdence that student career choice behawor relatlvc to several occupations IS senaltive to labor market condItIona. t In this dwusslon

to analyze the propobltron that programs at some mstltutlons are of such high qualIt> that tho Job prospect?, of their graduates are relatlvcly unmflucnced bq labor market condltlons

267

the quality of instruction and research. but it is also important to society because it has implications for the efficiency of the educational system. Violation on the side of smallness results m needless duplication of some educational facilities. In the opposite case. if costs increase with size. the implication is that not enough duplication is present Complete information on the cost-output schedule of a program is probably the only way to be sure that the program is being operated near optlmal size. But it should be remembered that the current costs of a program should also be revlewed to see if the current level is Justified. This activity should not he confused with the one described above. The former is to determme if a program is at optimum size, which is one sort of efiiclency: the object here is to see if the program is hemg operated without waste. whatever its current size, which is another sort of efficiency. In particular the reviewer might question the current student,‘faculty ratio, class sizes. or the amount of time faculty spend doing research as opposed to teachmg. The development of a realistic cost-output schedule is an essential element in the decisionmaking process. hut it ~111 not make decisions automatically. It can offer considerable help in evaluating the efficiency of an operation, but there will always be an element of judgment involved in deciding if either or both of the two sorts of efficiency discussed here are absent. If a unit, m the best judgment of all considered. is not operating efficiently in the latter sense. then it seems reasonable that in an environment of growth it should not receive additlonal funds until it is efficient; m a stable environment it should he cut back. The cost-output schedule described here can help with the decision by showing what alternatives are available for a given level of cost. But what about efficiency in the ,former sense? It is here that the complex interaction of student demand, job market. and state subsidles must he weighed by the central decisionmaker. Careful consideration of such elements will help determine whether a program should grow, remain stable, or become smaller; but the complexities are such that the magnitude of the desired change may be unclear. A cost-output schedule can help with this problem by showing what a given level of growth or dimmution would cost. One program may be able to grow by 5 per cent with only a 2 per cent increase m funding; another may require 10 per cent or more to increase its output 5 per cent. By showmg such relationships, decisionmakmg is put in proper terms: the declsionmaker chooses a given level of output instead of simply allocating more money and letting output fall where it will. .l. INCENTIVES IN A UNIVERSITY PLANNING AND BUDGETING

SYSTEM

ae do not attempt

The preceding sections have shown how department chairmen and deans organize and simplify information to be used in university resource allocation decisions.

26X

S.

A. HOFUACE.

P. D. MEAGHER, W. C WEILER and R A ZILLGITT

However, no resource simulation system is sufflclent to gua:antee efficient resource substitutability. In order to accomplish this. it must be in the self-interest of both dccentrahzed and centralized decisionmakers to use resources in a manner consistent with articulated university-wide goals. This section discusses some economic incentives which. if introduced mto a umversit> m conjunction with the information generating tools described above, could strongly encourage efficient use of educational resources. It illustrates the type of analysis which would accompany the design of a budgeting system based on incentives. As will be seen. careful analysis IS needed because incentives can have perverse as well as beneficial effects.*

Though some university-wide goals may be unique to ;I particular institution. we suggest four of universal applicability which form the basis for the present disCLISSIOII.These are: (I) Meeting student demand. University course and degree oRerings should be flexible to adjust to student demands. (2) Supplymg venture capital. Collegiate deans and central admmlstrators should have a regular sup* Any unlreraltj budgeting system. Q hether mtentlonally or not. has economic mcentlve effects See Brencman [3] for an an,~lqs~s of the mcentlvcs lmphclt m the budgeting system at the I Inl~crst~ of Ccll~forma. For a provocatIve dlscusslon ofeconomlc rclatlonahlps among departments within man) un~wrs~txs under the rwtmg tqpcs of budgeting systems. xc Boldcrston [I], The problems of economx eficlency in many types of burcaucratlc orgamratlons, m addltlon to umversitw. ha\c wmL~r~tlcs. An excellent and valuable dlscus~on of the potcntlal hrnefits of mtroducmg market mccntlva in hureaucr,ltlc orgamzatlon~ 1s prowded by NEkxncn [ 171. In the opmlon of the authors. because of the dnect contact hethccn chents (students and purchasers of rcac:lrch) and aupplws (mdlvldual faculty). unwrsitlcs are umqucly able to benetit from the mtroduction of market incentIves t A ‘regular. recurrmg budget’ may be consldered to be an) hudgetlng sqstrm under \bhlch the operating units are almost certam to rccewr for the next year at least the amount of then current fundmg ‘base’ Barring extraordlnary financlnl dlficultles at the central level, oprratmg units are fax11 sure under such a sqstrm that they ~111 share more or Icns equally, or m proportlon to their current fundmg Icvel. an) cutback m total funds the umverslty has for mstructlon Research umts may fare differently If then fundmg 1s from an outsldc agrncq such as the Federal government. Even then, m :i rccurrlng system. the mstltutlon may be ulllmg to dcvotc It5 own resources to softening the Impact of fund wlthdrltwl $ A completcl) operational system should take the un*que cnxumstanccs of r~ particular mstltutlon into account WC are attemptmg to develop one for the Univerat\ of Minnesota 4 Thl 1s not the only mechdnlsm hy hhlch economic nccntive, could be brought to bear on decentrallzcd decwonmaking m a umvetwt? For alternatwes. see Frledman [7] and Hoenack and Norman [ 1 I].

ply of uncommitted funds that they can devote to new ventures m teaching and research: venture capital as it were. (3) and (4) Efficiency University operations should bc efficient. as defined in prebious sections. That is: (3) Dollar efficiency. Whatever the level of output of an operating unit, it should be. by reasonable quantitative standards or explicitly stated value judgments. the maximum level of output attainable for a given dollar input. This is one kmd of efliclency. (4) Optimum size. No alternative level of output for any operating unit should result in a lower unit (average) cost. This is a second kind of efliciency. We should note that this type of efficiency implies the former type. but not conversely.

At present, most universities operate on the basis of regular. recurring budgets.? Our four goals arc not well-served by recurring budgets. I. St~rcier~rdernu&. Under a recurring system. decisions at departmental. collegiate, and central levels need not be based directly on student demand. but on administrative utility functions, which may or may not serve the welfare of students. 2. I,L~~tzrrrwpitd Rigidities inherent in recurring budgets cause programs to live beyond their usefulness. The recurring system makes it difficult to cut back outmoded programs, so that there are few If any uncommitted monies available for new ventures and needed expansion elsewhere 3. DO//LII.&icicw~~. Because they tend to perpetuate the status quo allocation of resources. recurring budgeting systems inhibit the consideration of alternatives that would promote dollar efficiency. 4. Optimm sm. During the late 1950s and in the 1960s. nearly every department could count on receiving its current budget plus .Y per cent in each successive year. In many cases. universities grew far beyond optimum size. With increasing problems of coordmation. control, and congcstlon there are now many umversities whose outputs and associated unit costs arc too large. even though@ t/~~t /rc~~/c?f’out/~ut.no more can be achieved with the current dollar input. Under a recurring system there is no practical way to sculc back to optimum size Only by actively considering large numbers of alternatives and adopting the best one can proper scale eventually be attained. E&ents

of‘n hettcr. ,VL/J’to .SL’Vl’t’ thefew ycrlcYcll$/ads

Although no operational preferred budgeting system will be developed here.++ It is clear that some system involving non-recurring budgets IS needed. For purposes of discussion it will be assumed that such a system would have three principal characteristics: (a) Student fees and governmental and other subsidies would be paid to the university’s central administration 4

(b) Flexlbihty

and venture capital would be obtained m two ways: administrative (or committee) discretlon and financial incentives to action on operatmg units. (c) Operating units at the collegiate and departmental levels would not know from one year to the next Just what theu- budgets would be. except to a degree determined by a set of administratively enforced financial incentives. What follows is an analysis of some incentives likely to serve our four goals assuming a non-recurring budgeting system. For any particular institution, a similar analysis should be extended to Its own unique goals.*

Gl\en current institutional reward structures, there arc few incentives for faculty to consider student demand for mstructlon. Student demand is more likely to be met if faculty are rewarded for showmg the same spirit of mnovatlon and entrepreneurship in their teaching that they have exhibltcd for so many years in obtdming research grants and producing scholarly pubhcations.

* Probably the central admlmstratlon of an mstltutlon hopmg to shift to an mcentlve system should coordmate. compllc. and evaluate such analyhes t The aggressive pursuit of rescarch contracts by acadcmlc personnel durmg the post-war period suggests thdt economic mcentlves strongly affect academx behavior In this section we arc proposmg balancmg these mcentlves with economic mcontlvcs uhlch reward mstructlon The reader should note the rclatlvely large amount 01 freedom for academics under such a system Those rclatlvcly capable of performmg InstructIonal services UIII be rewarded. Those prefcrrmg research and having a comparntlve advantage at It will not bc coerced to teach Instead. 1 Just ask an) tenured academic whose recent salary crease\ have been percentage-wlae smaller than mcreasx the Consumer Price Index $ Under the recurrmg budget often subsldlxd with mstructlonal hides the Important fact that our fundmg ‘pure’ research for which Federal government. can capture

nm

syhtcm ‘pure’ research IS revenues. This practice society needs B polq of no agency. mcludmg the the economic benefits

11Departments would also have an added mcentlve employ auxlllary faculty on B temporary hasps to adjust tluctuatlona m student course demand.

to to

7 It IS clear that the necessary cidJustmrnts would take borne time. so the reward system would have to he mtroduced over a dcfinltr period to enable faculty to adjust to the changed rnvxonment. ** These Issues are dIscussed m more drtall m Hoenxk and Norman, (111
analysis

1s provided relevant

and Weishrod

[9],

bq Hoenack

and

to this point. see Brene-

If departments and./or colleges were rewarded for Instruction, they would have an incentive to make It rrlcvant. interesting. and of high quality. Research has occupied a pre-emment posltlon m the faculty reward structure because faculty hnow their economic welfare depends on it.? Faculty economic well-being. It should be pointed out, involves more than Just tenure.: A department whose faculty fail to obtain a fair number of research grants may find Itself m serious financial trouble Perhaps if high quahty mstructlon were rewarded. a department or college which failed to offer It would also sutrer financially if it did not have sufficlrnt ability to generate offsetting research rcvenuc.$ Departments would then have a strong mcentlvc to sort out workloads among faculty according to rclatlvc abllties to attract research and instructional revenues 11They would reward mdlvldual faculty for instructlonal as well as research performance.‘1 Some may object to tallormg programs to student demands on the grounds that students do not have sufficient information to make mformcd cholccs. HOMever, mdlvidual faculty and departmental chairmen as suppliers of instructlon. and mdivldual students as demanders of instructlon. are more likely to be Informed about their preference\. needs. and optlons than commlttees and admimstrators can ever be. no matter how good their Information systems. Furthermore. m many umversltles. commlttces and admimstrators are sub]ect to. and often participants m, politecal processes which can be pcrverT,e to ctliciency m adaptmg to student demand.** The operational way of creating Incentives for instruction in the context of this section IS to base departmental budgets partly on cnrollmentsirt However. this must be done carefully becuusc reward structurcs of mcentlves. evpliclt or Imphcit. ma> have perverse effects as well as beneficial ones. Burgeomng enrollments may not always be a sure sign of qualIt instruction. For example. enrollment based budgets also provide an incentive to recruit students and to uflatc degree requirements artificially in order to retain students as long as possible. In order to encourage units to graduate students. it might be necessary to modify an enrollment based reward structure by tymg budgets to graduation rates as well as enrollmsnts.$: Ideally. this would encourage faculty to motivate \tudents to complete the program. or to recruit better qualified students. Conversely. It might also rncouragr some units to ‘water down’ degree requirements. Therefore, this incentlbe might best be applied 111programs uhere employers’ requirements or licensurc serve as a standard for degree quality. Incentives are also applicable to students as well as to faculty Even if students are capable of maklng informed choices of courses and programs. two factors tend to prebent them from accounting for cost dIfferent&s in choosing their programs: (1) The pohcy of subsldizmg institutions ralher than students may hake perkcrse effects on equalIt> of opportunity as other ureters have pi)intcd out.$$

270

S A HWNACL P D. M~AGHFR, W. C. WEILEK and R. A. ZILLGIT~

Policies to ameliorate these effects are also possible. with the result that a larger number of qualified students from low-income families might choose a college education than would otherwise be the case.* (2) Institutions often use a nearly fixed tuition fee system which usually has no reliable relationship with program costs. Since teaching costs vary considerably among departments, the result of a fixed fee system is that some students subsidize others. Analyses of average mstructional costs usually reveal that tuitions for liberal arts programs are a higher percentage of costs than tuition in most graduate and professional programs. The result is that, relative to their perceived benefits. whatever they may be. there are fewer students in the lowcost programs (where the present value of the benefits net of tuition IS relatively too low) than there should be. If tuition for students m the various programs varied directly in proportion to differences m their unit costs, students would be distributed among a university’s programs more closely according to the perceived benefits of the education received.? It is possible to regard a cost-based tuition policy as an incentive acting on departments and colleges as well as students. Units which want to attract students will be encouraged to find ways to keep costs down. On the other hand. the policy may have a number of serious problems in performing adequately as an incentive system. Departments and/or colleges often differ in the amount of political power they wield among their col-

leagues. Some, because of their large size relative to the rest of the parent institution, or because of the emmence of their faculty. have monopoly power of sorts over ccntral administrative decisions atrecting them. Dcpartmental power is also increased by existing subsidy practlces which force students to attend particular mstltutions in order to receive the subsldq. Thus. an indl>ldual department offering strongly demanded spcclahzed training can exert very substantial monopoly power within its university and generally wlthm its state. Such units may be able to extract special consldcratlons and larger budgets from central administrators m spite of any incentive system. However, any unit, even one without much monopoly power. may be ablr to USC the incentive system to extract a larger budget by chnosmg

Instruction in order to inflate them. The planmng procedure In this paper does not overcome this problem because it lets umt heads define the categories of costs they beheve arc most rclekant to their resource allocation decisions. However. it does make flagrant exercises of monopoly power easlcr to detect because it makes choices of cost categorlcs explicit and because It requires unit heads to sa! hoa they would meet future demands on then umts.S This opens the matter of marginal costs of incremental output to negotiation between operating unit heads and administrators. Any particular Incentive formula that might bc used must be analyzed for Its perverse as well as bun&la1 effects. Consider, for example, an incentive formula m which a unit with a 4yr program 1s rewarded for teaching by means of a fixed annual amount per unit of ‘output’, where output is the sum of students m the first 3 yrs of the program plus the number of current * See Hoenack [IO]. graduates. Such a formula approach subkcrts three economic realities: ( 1) that margmal costs change sigt However. the effects of such a cost-based tultlon pohcy nificantly with demand: (2) that the price of an cducaon mdlvldual department or college enrollments and hence on the total costs of a umverslty can be complex and posse tlonal program affects the number of students choosbly surprising. In this regard. see Hoenack and Weiler [ 1’71 ing it; and (3) that the reward structure. rather than Any mstltutlon embarkmg upon such a policy needs to be student demand. determines output. A5 ;I result. censensitive to the tmtlon elasticity of student demand for each tral admimstratlon would have the dilficult anal) tIcal of the programs. This does not necessarily mean that in task of improvmg Its incentive formulas by obtammg every case university admmlstrators must possess precise good estimates of student demand. and departments estimates of student demand elasticities, but that they would attempt to add politlcal input to this consldershould flexlblq adjust according to observed excess demand atlon m proportlon to their monopoly poacr. and supply conditions. The most valuable general dlscusIf student demand for a unit’s program \ICK butSlon of tlus topic for many types of bureaucracies 1s provlded by Lange. m Lange and Taylor 1151. It is possible that cessfully estimated and used m conjunctIon with cost the elasticltles can be so much greater m the low cost prodata. it would be possible to charge tuition for the programs than m the high cost programs that the lowermg of gram which would leave output at borne desired lsvcl. tuition fees m the low cost programs increases the number However. this approach requires a reward structure of low cost students sufficiently that total umvcrsity costs mwhich does not determine output. but which does rccrease faster than addItIona revenues. ward the infra-marginal unit of quality teaching. 1 WIthin the system of discretionary central admmlstraFinally, consider the case N here studrnt economic tlve declslonmakmg authority assumed here to be m rxlstdemand for a unit’s programs IS estimated and tuition encc. and m wluch departments often provldc the only subis based on the costs incurred for the level of output sldlzed services for state residents. the present authors are determined by the mtersectlon of the unit’s marginal unable to suggest specific means of curbmg departmental cost curve and the estimated marginal revenue of an monopoly power However. m a different financmg context. additional student. In practice. this may requtre a unit there seem to be ways to accomplish this. In this regard, see Hocnack and Norman, op. clr. [I I]. to operate at somethmg other than an eficicnt lcvcl of

Umvers~t~ planrung.

decrntrahzatlon

if the level of student demand is great enough. These, of course, are intended to be interestmg cases, not insurmountable difficulties. Nowhere, to the knowledge of the present authors, cdn one find theoretical or empirical analyses of such problems. output

Formerly, new monies were available each year which were not necessarily committed to currently pursued activities With these new momes institutions were able to fund the new activities that were an essential part of simple physical growth. as well as intellectual development. Because of demographic trends the need for universal growth is past, but certainly not the need for selective growth and general intellectual development. Thesenew moniesconstltuted the Institution’s venture capital. However. the new financial cnvironmcnt. coupled with the commitments mhercnt in faculty tenure, implies that the supply of new venture capital will be severely curtailed. if not cut off. In the absence of new monies, a umversity budgetary process should provide ways ofproviding collegiate deans and central administrators with a regular supply of funds for developmg new programs and mamtammg existmg excellence. Other than endowments, special appropriations, and grants from government (an uncertam source). the only possible regular source of funds IS reallocation from existing activities. There would appear to be essentially two ways of raising venture capital from existing activities. (a) Whole units or programs can he cut back or ehminated. Such a course requires decisions on the part of central administrators that are likely to be made only with extreme reluctance. (b) Proposals for new or expanded programs may also be dealt with m the following manner: (1) A department or collegrate head proposing a new or expanded program might be required to mdicate which of his prssent activities he would be willing to operate with fewer faculty and other resources (that IS, at a planned level of super-efficiency) to free a portion of the capital needed to get the new program off and running and earning a fair share of its own costs. The dean or central admmistrator who would have to finance the remaining portion of the costs would also make a careful analysis of the proposal, of the proposmg unit’s current activities, and of the activltles of other umts from which the remainder of the resources needed would have to be extracted. If there seemed to be no way to raise the proposing unit’s share of the venture capital, or the balance. it would Imply that the expected ‘payoff’ of the proposed activity is smaller than the (imphcit) costs of reallocatmg the necessary resources. (2) A unit might be required to indicate which of Its actlvities It would be willing to cut back or eliminate in order to raise a portion of the necessary additional funds. If it is unwlllmg to do so, it may be

and resource

allocation

771

assumed that the unit 1s engaging currently in those activities which are most important to it. A similar situation confronts the administrator responsible for raising the balance of the needed funds. Only when new ventures are more important than existing actlvlties would the old be dlscarded in favor of the new. An important question in operating such a system, of course. IS the imphcatlons It would have for the structure of admmlstrative authority in a university since a dean or central administrator might also propose new programs. Can. for example, a central admimstratlon propose a new or expanded program and o&r an unwilling dean or department to provide a portion of the funds’? Again, data developed as described in the planmng sectlon could be used by all concerned to explore alternatives. The existence of such data acts as an incentive for openness and the use of facts rather than secrecy and rhetonc m dealings among university admimstrators. (3) In some cases it may bc possible to raise venture capital m the form of loans from private money markets. An example might be health care professlonal programs where there IS great demand for the graduates. In such cases students are now able to obtain loans which, m turn, are substantially used to pay tuition. It would be a relatively small step to use part of such revenues to pay Interest and principal on debts incurred to set up a program.

The proposals which have been discussed here would help universities achieve both dollar efficiency and optimum size. The planning procedures organize and condense data at the departmental and collegiate levels. Such procedures, if properly applied, promote dollar efficiency whatever the level of output because they permit larger numbers of alternatives to be consldered with the limited time available to administrators. The procedure would also promote better resource allocation decisions because units would describe alternative actions before the fact: budgets would no longer be Justified by pointing to existing workloads The common practice of basing departmental or collegiate requests for funds on existing workloads seems to be an inherent property of the recurring budget system. As has been seen. recurring budgets tend to promote inefficiency: they tend to perpetuate previously determined cost-output positions and they tend to treat existing budgets as a ‘base’ to which extra funds and extra activities can only be added. The replacement of the recurring budgetary system with a properly designed incentive based budgetary system could move a umversity toward efficiency Although properly designed incentives enforced by central administration can go a long way toward promoting efficiency m university operations. the fact that

273

S A HOI.NA~,L P D. MLAGHLK, W (

teaching units are often not actively competing with each other for the attention of scarce students and their tuition has the effect of mitigating the desire to mmimaze costs at the departmental level. In such a situation vigorous negotiation between central admmistration and operatmg units over admissible categories of costs and their appropriate levels is the only remedy the authors can suggest at present. In conclusion. proposed incentives can bc described and analyzed, although It is no easy task to do so. The design and analysts of Incentives for universities and other non-profit institutions is an Interesting and almost cntlrely unexplored field for further research. .4C~~lo~c.l~~~l~o,IC’lltS-Theauthors would hke to thank David S. P. Hopkms for valuable critlclsm of an earher draft of this paper They also gratefully acknowledge the many discussions with David J. Berg at a time when numerous ldcas m this paper were first being formulated We also want to thank Faith E. Jaycox for invaluable editorial assistance and Bonnie J. Grobe for typmg numerous earher drafts of this paper.

REFERENCES

4

5

9

10.

II

12.

13.

1 F. E Baldcrston, Complementarlty. Independence and substltutlon m unlvcrslty resource allocatIon and operatIon, Paper P-39. Ford Foundation Research Program m Llmverslt> Admimstratlon Umverslty of Cahforma. Berkeley (August 1973) 15 16

3 D U’. Brencman.

An economxc theorq of Ph.D production. the case at Bcrhcley. Paper P-X. Ford Foundation Research Program in Umverslty Admmlstration. Umvcralty of Cdllfornla. Berhcley (June 1970)

17

WLILI K and R. A. ZILL (;I I I

B. Danrzlg. Llwur Pro~qr~wnrrm~ rrrul E v~~~~I.~IoJ~,~. Prmceton Umversq Press, Prmceton (1963). R Dorfman. P A. Samuelson and R Solow. ,!,/~lcilr Pw~I’UIIII~III~ ur~tf Ecomwrc 4r1ul~xs. McGraw-Hill. Ne\\ York ( 1Y58). R. B. Freeman. T/I,, hfur%~t /I~I CO//~Y+~7iat-related tultlon pohcles and unlvcrslty enrollments Managcmcnt Informatlon Dlvlslon. Umverslty of Mmnesota. Mmneapohs 119741. D. S. P. Hopkms. On the use of Iargo-xali am~ulatlon models for unlversitj plunnmg. Kc{ Educ Kc\ 41(5). 367-378 (December 197 I ) R D. Lamson and J H. Powel. E/~,IJI~vI~,Rc/~ri~,d fo t/w L)‘~rtv,,,,lll,l,~,,l ,,I C‘O\I\11lldBc~llcqrf,\01 (r,.dLcltic Sl~ll~~lli\ The Council of Graduate Schools. Washmgton D.C ( 197’). 0. Lange and F. M Taylor, 011 tlrti E~OI/