Organizational economics studies why activities are organised inside firms and other managed organisations, and how their internal arrangements affect performance. It examines the allocation of decisions, information, incentives and resources rather than treating an organisation as a single person that automatically knows what to do.
The gold standard is a comparison between workable alternatives. A dependable explanation identifies the task, follows the decisions that produce the outcome, includes the costs of coordination and asks what would change under another arrangement. This guide uses original fictional cases to make those comparisons explicit. The numbers illustrate reasoning; they are not observations about a real employer or prescriptions for restructuring one.
Choose your organizational economics route
Follow choices through firm boundaries, local knowledge, shared resources and adaptation.
Route 1 · Chapters 1–3: Understand scope and firm boundaries
Route 2 · Chapters 4–6: Follow investment, authority and wider effects
Route 3 · Chapters 7–9: Allocate resources and coordinate information
Route 4 · Chapters 10–12: Evaluate influence, adaptation and evidence
Route 5 · Chapters 13–14: Transfer the method and continue learning
Chapter index
Understand scope and firm boundaries · Chapters 1–3
1. Open the organisation before judging its performance
Follow investment, authority and wider effects · Chapters 4–6
4. Follow dependence after an investment is made
Allocate resources and coordinate information · Chapters 7–9
7. Use internal prices carefully when capacity is scarce
8. Coordinate choices that work only together
9. Examine information acquisition as well as information transfer
Evaluate influence, adaptation and evidence · Chapters 10–12
10. Include influence costs and distorted reporting
11. Treat adaptation as an organisational task
12. Evaluate performance without mistaking selection for causation
Transfer the method and continue learning · Chapters 13–14
13. Transfer check: change the spillover and rebuild the choice
CHAPTER 1 OF 14 · Understand scope and firm boundaries
1. Open the organisation before judging its performance
Imagine a repair business with two workshops, a purchasing team and a central scheduling office. Customers see completed repairs or delays. Inside the business, technicians diagnose faults, purchasing orders parts, scheduling assigns capacity and supervisors approve exceptions. Each person can perform their own task competently while the whole service disappoints. A part may arrive after the reserved repair slot, or an urgent job may wait for an approval that nobody present can give.
Organizational economics asks how those arrangements shape choices and outcomes. The field overview by Robert Gibbons and John Roberts describes an economic approach to the nature, design and performance of organisations. Its scope includes what happens within an organisation as well as why its boundaries take a particular form.
Start with a unit of analysis smaller than the company name. A repair, a purchasing decision, a request for shared equipment or a proposal to change a workflow can be followed from beginning to end. Identify the people who possess relevant information, the people permitted to act and the people who bear the consequences. These may be different groups.
Put the question beside its neighbours
Industrial Organisation studies competition and market structure. Organisational Behaviour examines people and groups through a broader behavioural lens. Institutional Economics investigates rules and enforcement. The focus here is how economic reasoning explains choices among organisational arrangements, including arrangements inside the same firm.
These boundaries are useful routes, not impermeable walls. A workshop's internal incentive can affect customer outcomes and competitive performance. A legal rule can constrain which decisions may be delegated. The aim is to keep the immediate question clear while recognising the surrounding conditions.
For our repair business, replace “the organisation is inefficient” with a specific question: should each workshop order its own urgent parts, should central purchasing handle every request, or should the two share a bounded exception process? Each alternative needs a mechanism and a comparison, not a management slogan.
CHAPTER 2 OF 14 · Understand scope and firm boundaries
2. Compare feasible systems rather than ideal labels
A proposal to centralise purchasing often promises lower prices and consistent standards. A proposal to decentralise often promises speed and local knowledge. Both can be true in some circumstances. Neither tells us which arrangement works better for the actual mix of tasks, information and constraints.
Define what must remain comparable. The same quality requirement, demand, service horizon and available resources should be used unless changing one is explicitly part of the intervention. Comparing a fully staffed central team with an understaffed local system would confuse staffing with authority. Comparing annual costs with one month's savings would confuse time horizons.
Suppose the business handles 100 urgent requests per month. Under local ordering, administrative effort costs two units per request and average purchasing expenditure is 18. Under central ordering, administrative effort costs three units per request but purchasing expenditure falls to 15. Ignoring all other effects, local ordering costs 2,000 per month and central ordering costs 1,800.
Add the missing consequence
Now assume central processing creates a one-day delay on 30 requests, costing ten units per affected request in lost service value. Its full modelled cost becomes 1,800 + 300 = 2,100. Local ordering is cheaper on this expanded measure. If the delay instead affects only ten requests, central cost becomes 1,900 and the ranking reverses.
The arithmetic does not prove that central purchasing is good or bad. It identifies a threshold. Central purchasing saves 200 before delay costs, so it remains cheaper only when the added delay cost is below 200. At ten units per affected request, the threshold is 20 delayed requests. Equality means the two arrangements tie on the chosen measure.
This is a comparative organisational explanation: state the alternative, identify the changed cost and expose the assumption that drives the ranking. It also directs evidence collection. The manager now needs reliable information about delays and their consequences, not another generic survey asking whether staff prefer centralisation.
CHAPTER 3 OF 14 · Understand scope and firm boundaries
3. Separate production costs from arranging the exchange
Making a component uses materials, machinery and labour. Obtaining that component from another party also requires finding a supplier, communicating requirements, evaluating quality, arranging delivery and resolving disagreements. Moving production inside a firm changes these activities; it does not make coordination disappear.
The Royal Swedish Academy's account of Williamson's work explains why difficulty in contracting and mutual dependence matter when comparing market transactions and hierarchy. It also emphasises that hierarchical authority has costs. The relevant comparison is between imperfect arrangements, not a flawed supplier relationship and a frictionless internal department.
Consider a fictional annual requirement for 1,000 identical parts. An outside supplier charges eight units per part. Selection, specification and monitoring cost 600 per year, and expected disruption costs are estimated at 900. The resulting total is 8,000 + 600 + 900 = 9,500.
Internal production has variable cost six per part, annual equipment and staffing commitments of 2,000, and coordination costs of 700. Its total is 6,000 + 2,000 + 700 = 8,700. Under those assumptions, internal production costs 800 less. But if demand is only 400 parts and the commitments remain fixed, external cost becomes 3,200 + 600 + 900 = 4,700 while internal cost becomes 2,400 + 2,000 + 700 = 5,100.
Check the model's range
Holding disruption cost at 900 when volume changes is an assumption requiring scrutiny. A disruption may have a fixed incident cost, a per-unit consequence or both. Similarly, internal capacity might be redeployed, and an external supplier might change its price with volume. The calculation is useful precisely because those assumptions can be named and replaced.
Do not automatically classify every monitoring expense as waste. Inspection can prevent costly defects. The useful question is whether the resource spent improves the exchange enough to justify its opportunity cost. A low purchase price can be expensive to use; a high administrative cost can sometimes protect valuable cooperation.
CHAPTER 4 OF 14 · Follow investment, authority and wider effects
4. Follow dependence after an investment is made
A supplier that can sell the same part to many customers has alternatives. A supplier that builds equipment usable only for one customer's unusual design may become dependent on that relationship. The investment can create substantial joint value while changing what happens if the parties later disagree.
The existing Hold-Up Problem article explains this connection between specific investment and future bargaining. For organisational analysis, the next question is which arrangement preserves the incentive to invest while allowing adaptation. A long agreement, a different ownership arrangement or a less specialised design can change the comparison in different ways.
Suppose a dedicated tool costs 60 and saves 100 over its useful life. A general tool costs 40 and saves 65. Before considering bargaining, their net benefits are 40 and 25 respectively. Now assume the investing supplier expects to retain only half of the dedicated tool's saving, because the subsequent price adjustment is not settled beforehand. Its private gain is 50 − 60 = −10, so it declines the investment.
If the general tool's savings can be retained because it serves many customers, its private gain is 65 − 40 = 25. The supplier chooses the less productive design for the relationship, given the stipulated incentives. This is not evidence that it misunderstands arithmetic. It faces a different return from the joint return.
Explain a remedy's mechanism
A proposed buyer contribution of 40 reduces the supplier's net investment outlay to 20. With an expected retained saving of 50, its private gain becomes 30, exceeding the general tool's private gain of 25. The buyer retains the other 50 of savings and pays 40, leaving ten. The combined gain remains 40: the contribution redistributes value rather than creating it. That can change the investment choice in this small model. But now ask why the contribution would actually be delivered, whether investment quality can be verified and what the buyer receives in return.
The example is not a complete contracting solution. It identifies the missing incentive and shows how a transfer might alter it. For a fuller treatment of payment, participation and control, follow Contract Theory. Ownership is one instrument among several, and changing it can alter more than one party's incentives.
CHAPTER 5 OF 14 · Follow investment, authority and wider effects
5. Place decisions where information can be used
An organisational chart allocates positions. A working decision process must also specify which information reaches those positions, when it arrives and whether it can be acted on. Giving authority to a central office is not equivalent to giving that office accurate local knowledge. Giving information to a technician is not equivalent to authorising a purchase.
The distinction between the right to decide and effective influence over decisions is developed in Aghion and Tirole's Formal and Real Authority in Organizations. It cautions against assuming that formal reporting lines alone reveal where decisions are actually determined.
For an original decision exercise, suppose a workshop must choose a replacement component suitable for either a common fault or an unusual fault. The common fault occurs with probability 0.8. Choosing correctly yields 50 units of service value; choosing incorrectly yields zero. Without any diagnostic information, choosing the common-fault component yields expected value 0.8 × 50 = 40.
A local diagnostic test identifies the fault perfectly and costs four units. Using the result yields 50 − 4 = 46, an improvement of six. If transmitting the diagnosis and waiting for central approval costs another eight units, the centralised process yields 38. The information is valuable, but the route through which it is used can consume more than its benefit.
Authority needs a defined scope
This model supports local action only for the stipulated choice and consequences. It does not establish that every technician should decide every purchase. If the component affects other workshops, a shared inventory or a safety requirement, those consequences belong in the model. Local knowledge and organisation-wide effects can point toward different decision arrangements.
A bounded delegation might allow specified replacements within an approved range while escalating unfamiliar cases. Such a design should identify the boundary, evidence required and review route. Otherwise, “use your judgement” can leave staff responsible for an outcome without knowing whether they have permission to produce it.
CHAPTER 6 OF 14 · Follow investment, authority and wider effects
6. Count the effects that cross departmental boundaries
A department can choose well for its own accounts and poorly for the whole organisation. The issue is especially clear when its decision changes another department's workload. The economic question is whether the decision-maker sees the relevant costs and benefits, rather than merely whether everyone has a performance target.
Suppose purchasing can buy a batch of components for 900 instead of 1,000 by accepting a less convenient packaging format. The workshops then spend an extra 180 on unpacking and sorting. Purchasing records a saving of 100; the organisation incurs a net additional cost of 80. Both accounts can be correctly calculated while the purchasing decision is poorly aligned with total cost.
Now suppose a different packaging change saves purchasing 100 and adds only 40 of workshop work. Total cost falls by 60. A blanket prohibition on inconvenience would reject this beneficial option. The relevant measure is the complete consequence of the change, subject to quality and workload constraints.
Make the omitted effect visible
One repair is to include an estimate of handling cost in the purchasing comparison. Another is to let affected teams review changes before commitment. A transfer charge between departments may also change incentives, but the number must represent the relevant opportunity cost rather than an arbitrary penalty.
Check whether an apparent internal saving simply changes the accounting location of the same expenditure. Moving an employee's hours from one budget to another does not free those hours for the organisation. Conversely, an internal transfer payment can alter behaviour without directly changing total resources. Distinguish the accounting entry from the action it induces.
This is a useful place to connect to the principal–agent guide. Delegation can put decisions and consequences in different hands. The repair should reconnect the relevant consequence while preserving enough authority and information for the work to proceed.
CHAPTER 7 OF 14 · Allocate resources and coordinate information
7. Use internal prices carefully when capacity is scarce
A shared resource creates an allocation problem. Consider a testing machine with ten available hours. Workshop A can use six hours to produce a benefit of 30 per hour. Workshop B can use six hours for a benefit of 20 per hour. Assume divisible hours, constant benefits within these limits and no other costs or interactions.
The highest total benefit comes from allocating six hours to A and four to B: 6 × 30 + 4 × 20 = 260. An equal split gives five hours to each and total benefit 250. Equal shares may be defensible for other reasons, but they do not maximise the stated benefit in this example.
If machine time is free to departments and requests have no consequence, both may ask for six hours. An internal price can help represent scarcity, but here a single price does not automatically solve the exact allocation. A price strictly between 20 and 30 leads A to request six and B zero, leaving four hours unused. At price 20, B is indifferent and some allocation rule is still needed.
A price is part of a mechanism
This small example shows why “charge for the resource” is incomplete advice. We need to know how requests are submitted, how ties are resolved, whether unused reservations are released and whether departments have incentives to report benefits accurately. Administrative allocation and prices can be complements rather than exclusive choices.
Now change B's benefit to 35 per hour. The efficient allocation becomes six hours to B and four to A, giving 330. A standing rule that always prioritises A would miss that change. The allocation process therefore needs current information, not merely historical status.
Benefits may also be difficult to quantify. A machine slot needed to investigate a serious quality defect should not be reduced casually to an invented revenue estimate. The organisation must define constraints and priorities before optimising within them. A calculation makes a stated objective transparent; it cannot decide which legitimate objectives should be ignored.
CHAPTER 8 OF 14 · Allocate resources and coordinate information
8. Coordinate choices that work only together
Some organisational changes are complementary: the benefit of one depends on whether another is also made. Introducing a new scheduling system without changing the information supplied to it may produce little value. Training staff on a process that supervisors never permit them to use can have the same problem.
Consider two workshops choosing whether to adopt a shared parts classification. Staying with current systems gives each payoff zero. If both adopt, each receives a benefit of 12 and pays an adoption cost of four, leaving eight. If only one adopts, it pays four but receives no interoperability benefit, while the other remains at zero.
Both adopting is better for both than neither adopting. Yet neither adopting is stable against a unilateral change: the first mover would lose four. Both adopting is also stable, because abandoning the shared system would remove the adopter's benefit and leave it at zero. The model has a coordination problem, not simply a shortage of aggregate benefit.
Make the transition credible
A shared implementation date, a tested interface and a commitment of training time can help make the other workshop's adoption credible. But an announcement alone does not change the payoffs if staff doubt that the other side can complete its preparations. Check the dependencies that make coordinated adoption feasible.
Suppose a transitional adapter costs one and lets a lone adopter receive benefit six. Its unilateral payoff becomes six − four − one = one, so adoption can now be worthwhile before the other workshop changes. The adapter changes the transition problem, not just the appearance of the final design.
This example is an original coordination game, not an estimate of software returns. Its value is diagnostic: a good end state can remain difficult to reach because the route contains a temporary disadvantage. Evaluate migration, compatibility and support costs separately from the long-run benefits. Otherwise, a sensible reluctance to move first may be mistaken for irrational resistance.
CHAPTER 9 OF 14 · Allocate resources and coordinate information
9. Examine information acquisition as well as information transfer
An organisation does not merely receive information that already exists. People decide whether to investigate, diagnose, document and share it. A process can discourage those investments if the resulting knowledge is unlikely to affect any decision. The relationship between authority, information and initiative therefore needs attention before assuming that more reporting will solve a problem.
Suppose a technician can spend three units of effort to identify an improvement worth 20 to the organisation. Assume discovery succeeds with probability one half. The expected gross benefit is ten, so the investigation has net organisational value seven. But suppose the technician bears the effort and expects only one unit of personal benefit if the improvement is implemented. Their expected benefit is 0.5, below the effort cost of three.
This is a deliberately narrow model. Professional satisfaction, recognition, future opportunities and concern for colleagues could change the technician's payoff. The point is to identify what the model includes, not to assert that people care only about immediate money. If the explanation relies on a motive, name it and seek evidence.
Investigate why useful knowledge stays unused
Another possibility is that the technician happily investigates but expects proposals to wait months for approval. The information problem then lies in the response process. Paying for more proposals may produce a larger queue rather than more implemented improvements. A repair could require decision capacity, a clearer scope of local action or a prompt explanation when a proposal is rejected.
Record the stages separately: ideas considered, investigations started, useful findings produced, proposals submitted, decisions made and changes implemented. A drop between two stages suggests a different mechanism from a low total count. Do not reward the last visible number without examining the route that produces it.
Gibbons's Inside Organizations distinguishes approaches centred on pricing, control and persistent differences between organisations. That perspective is useful here: some problems are poorly represented by a missing price alone, and the same formal structure can operate differently depending on its established practices.
CHAPTER 10 OF 14 · Evaluate influence, adaptation and evidence
10. Include influence costs and distorted reporting
When budgets or approvals depend on managerial judgement, people may spend resources influencing that judgement. Some communication provides useful information; some primarily changes who receives a benefit. An analysis should distinguish the two rather than treating every meeting as either waste or necessary coordination.
Suppose two workshops compete for one upgrade. The upgrade creates total value 100 if assigned to A and 80 if assigned to B. Each workshop prepares a presentation costing ten staff-hours, valued at two units per hour. If the presentations provide no new information and A would have been selected anyway, the organisation spends 40 without changing the allocation or its value.
Now suppose a technical assessment costing 15 reveals that B's true benefit is 140 rather than 80. Selecting B then improves the allocation by 40, exceeding the assessment cost by 25. Communication can therefore be valuable when it changes the decision through better evidence. The expense alone cannot tell us whether it was productive.
Design evidence requirements around the decision
Ask what a decision-maker needs to distinguish the alternatives. A common cost definition, a shared time horizon and an independently checked capacity estimate may be more useful than competing polished narratives. At the same time, rigid forms can suppress relevant local facts if they allow no explanation of unusual cases.
Reporting incentives deserve separate attention. If bad news automatically removes a team's resources, it may become harder to obtain early warnings. If every overspend is reimbursed without scrutiny, cost control may weaken. Neither observation establishes a universal budget rule. It identifies a response to test under the actual allocation process.
An effective review preserves the distinction between an adverse event, a reporting failure and a poor decision. A team that reports a genuine problem promptly has supplied information the organisation needs. The review should examine the event and the quality of the response, rather than making the presence of a report itself evidence of incompetence.
CHAPTER 11 OF 14 · Evaluate influence, adaptation and evidence
11. Treat adaptation as an organisational task
An arrangement can perform well under familiar conditions and fail when demand changes. Adaptation requires noticing the change, interpreting it, deciding what to alter and coordinating the alteration. These are separate tasks, and they may be assigned to different people.
Imagine the repair business begins receiving a new type of fault. Technicians recognise it first, purchasing sees rising demand for a component, and scheduling observes longer repair times. Each has part of the evidence. A central report that averages all repairs may conceal the pattern, while isolated local changes may produce incompatible fixes.
Build a concrete adaptation route. Specify what observation triggers review, who combines the evidence, who can authorise a trial and how results return to the original teams. This is more informative than saying the organisation should become agile. It describes the decisions that agility would require.
Compare adaptation costs over a suitable horizon
Suppose a revised process costs 120 to implement and saves eight per week for 20 weeks. With no discounting and no other effects, the gross saving is 160 and net benefit 40. If the relevant fault disappears after ten weeks, savings are only 80 and the same change loses 40. Expected duration is therefore part of the decision.
A reversible trial costing 30 might reveal whether the fault persists before the organisation commits the remaining resources. But the trial can also delay savings and may not reproduce full-scale conditions. Its value depends on whether the information can change the subsequent choice, not simply on its being labelled a pilot.
Preserve a record of the decision and its assumptions. If the environment changes, revisiting an earlier choice should not require reconstructing why it was made from memory. An organisation learns more from “we expected 20 weeks and observed ten” than from a retrospective claim that someone should have known the outcome all along.
CHAPTER 12 OF 14 · Evaluate influence, adaptation and evidence
12. Evaluate performance without mistaking selection for causation
Observing that decentralised workshops perform better does not prove that decentralisation caused their performance. Managers may delegate to workshops that already have more capable staff, simpler cases or better equipment. Conversely, they may centralise troubled operations precisely because performance was poor. The direction of causation matters.
Suppose local-authority workshops complete 90 of 100 jobs on time, while centrally managed workshops complete 80 of 100. The raw difference is ten percentage points. Now discover that the first group handles mostly routine repairs and the second mostly unusual faults. The original comparison has not separated the effect of authority from the effect of case mix.
State what would make the comparison informative
A stronger evaluation defines eligible jobs, records task difficulty and follows outcomes over comparable periods. Where feasible and appropriate, a planned trial can allocate the change in a way that reduces selection bias. When that is not possible, explain which differences remain uncontrolled and avoid presenting an adjusted estimate as certainty.
Measure consequences beyond the chosen headline. Faster completion could coincide with more repeat repairs, higher staff fatigue or less documentation. A useful outcome set includes quality, resource use, access, reliability and effects on other teams. The appropriate measures depend on the service; a longer list is not automatically a better evaluation.
Also distinguish private performance from wider value. A business may improve its accounts by transferring delay costs to customers or uncompensated work to staff. Those effects belong in a welfare assessment even if they are absent from the firm's immediate accounting objective. State whose outcomes the analysis is meant to improve.
The Institutional Analysis root offers another route for examining rules in use, participants and outcomes. Together, the approaches encourage a disciplined claim: the observed result, the proposed mechanism, the credible alternative explanation and the evidence still needed.
CHAPTER 13 OF 14 · Transfer the method and continue learning
13. Transfer check: change the spillover and rebuild the choice
Return to the diagnostic decision in Chapter 5. Without a test, the workshop's expected service value was 40. A perfect local test costing four raised net value to 46. Sending the result through an approval process costing eight reduced it to 38. On those assumptions, local action looked best.
Now add a consequence that was previously absent. Choosing a locally suitable replacement can create compatibility trouble for another workshop. Assume local action causes an expected additional loss of 12 elsewhere. Central review prevents that loss completely. Keep every other number unchanged. Which arrangement now creates the greatest total value?
Local action yields 46 − 12 = 34. Central review yields 38. The ranking reverses because the local decision has an effect beyond the original boundary. The technician's information remains valuable; the new problem is coordinating its use with another unit's needs.
Test a third arrangement
Suppose a shared compatibility rule can be checked locally at a cost of two per decision and prevents the same loss completely. Local diagnosis plus that check yields 50 − 4 − 2 = 44. It outperforms both unrestricted local action and the original central review under these assumptions.
If the check prevents only half the expected loss, the value becomes 50 − 4 − 2 − 6 = 38, tying central review. If it prevents less than half, central review is better on this measure. This threshold tells us what evidence matters before adopting the simpler process: how reliably does the local check catch the incompatible cases?
The exercise does not recommend delegation in general. It demonstrates a transferable method: preserve the original model, add the changed consequence and recompute all feasible alternatives. A new option may combine local information with shared constraints rather than choosing either complete centralisation or complete autonomy.
This is the kind of changed-context reasoning developed in The Gold Standard of Education. The learning is secure when a reader can explain why the answer changes, not merely repeat the earlier answer with a different organisation's name.
CHAPTER 14 OF 14 · Transfer the method and continue learning
14. Build a practical learning route
Begin with one decision you can describe without confidential information: allocating a shared room, approving an ordinary purchase or assigning a routine task. Draw its sequence in words. Identify what each participant knows, which actions they can take and which consequences they experience. Keep observed facts separate from assumptions about motives.
Next construct two genuinely feasible alternatives. Compare the same task and time horizon, and include at least one consequence that crosses a team boundary. Use a small numerical example to expose the threshold at which your preference would change. If every imaginable number supports your preferred arrangement, check whether you have silently excluded its disadvantages.
Is organizational economics just cost cutting?
No. Cost is one part of performance. The field examines how arrangements affect information, cooperation, investment and adaptation. A more expensive process can produce greater value or protect a required standard. Define the objective and constraints before deciding which expenditure is avoidable.
Does a market always outperform an internal department?
No universal ranking follows. Supplier competition can be useful, while specification, monitoring and dependence can make exchange costly. An internal department can coordinate adaptation but also introduce administrative and incentive problems. Compare the actual alternatives under the same requirements.
Does centralisation mean that managers know more?
No. Authority and knowledge are distinct. Central decision-makers may see cross-unit effects while local staff know the immediate situation. The design problem is to bring relevant information and authority together at an acceptable cost, including the cost of delay and the consequences for other teams.
Can every coordination problem be solved with a price?
No. The shared-machine exercise shows that prices may still require allocation and tie-breaking rules. Some consequences are difficult to measure, some decisions face rights or safety constraints, and some interactions depend on complementary choices. A price is an instrument whose operation must be specified.
How do we know whether a reorganisation worked?
Define the expected behavioural change before implementation, track comparable outcomes and examine side effects. Separate improvements caused by the arrangement from changes in staffing, equipment, demand or case mix. Report uncertainty rather than allowing a new organisational chart to serve as evidence of success.
Continue through mechanisms and primary research
Use the How X Works hub to follow related mechanisms and the Gold Standard master to return to the subject network. Gibbons and Roberts's Organizational Economics essay provides a research map. The primary works linked above support the conceptual distinctions; the workshop calculations are original exercises with explicit assumptions.
A useful final explanation identifies a decision, compares workable arrangements and follows the consequences beyond the first department that benefits. It should also tell the reader what new evidence would change the conclusion. That makes organisational analysis a method for improving a real choice rather than an argument for a favourite structure.