Making Institutional Microeconomics Playable

Inside a health-system strategy game where rival institutions act on different incentives, information, and clocks.

Three Months at Riverside

What follows is the opening of an actual playthrough: the 24-month competitive campaign on Normal difficulty, using seed 42. I played Riverside’s executive team. Each month, the game gave me an institutional report and a limited action budget to spend on staffing, capital projects, monitoring, payer negotiations, or public commitments.

Riverside Community Health began my January turn with 60 units of cash, three action points, and a nursing vacancy problem. Its reported access index was 68. State Medicaid officials were increasing scrutiny of access reporting. Six critical-care patients were boarding in the emergency department because the system had no effective ICU capacity. Northlake Health and Summit Care, the two rival systems in the regional market, had made no public moves.

The executive report offered four immediate priorities: protect capacity, recruit staff, investigate a rival, or make a public access pledge. Each addressed a recognizable problem. The action budget allowed three points, enough to begin a serious strategy but too little to pursue all four.

I chose an outpatient project and nurse recruitment:

1

project kind=clinic_network budget=18; recruit role=nurse headcount=4

The two commands formed a plausible access strategy. The clinic network would eventually add outpatient capacity, while the nurses could relieve a staffing constraint. Their timing differed. Recruitment would resolve after a delay. The clinic network was a nine-month project: its first draw occurred at authorization, leaving eight months after January and a completion date in Month 10. Riverside was committing cash to capacity that did not yet exist.

January ended with 46 cash, down from 60 even after an operating margin of 8. The system had treated all 24 modeled demand units, but the six boarded critical-care patients remained. The four nurses had not arrived. Workforce trust fell by four points, although the next report still described it as moderate. The clinic project sat in the queue with a monthly draw of 2.

The market had also moved. Northlake invested 25 in beds. That action was publicly visible in Riverside’s February report. Summit’s activity remained private.

By February, the original plan had become harder to interpret. The full treatment count suggested that Riverside was meeting modeled demand, while ED boarding showed a service-specific constraint that the total concealed. The clinic project might improve future outpatient access, but it would not create ICU beds. Northlake was expanding the capacity Riverside had declined to build. Riverside still had three action points, yet some of its financial flexibility already belonged to January’s decision.

I used the month to learn about Summit and to make the access strategy public:

1

monitor target=summit depth=1; commit pledge_type=access level=3

Monitoring produced no beds, staff, or revenue. It spent part of the month’s capacity to improve the evidence behind a later decision. The pledge raised Riverside’s reported access index from 68 to 74 and gave state regulators a visible commitment to track. Meanwhile, the delayed recruitment resolved and brought the four nurses into the system. Operating cost rose from 34 to 36, leaving a monthly margin of 6. Cash recovered to 50, though the clinic project still had seven months to run.

The monitoring result changed the March briefing. In Month 1, Summit had used an aggressive posture in a private CarrierA negotiation. Riverside could now see that payer bargaining was part of the competitive field, alongside Northlake’s bed expansion. The information arrived after Summit had acted. It did not rewrite what Riverside knew in January, when its managers had to commit without that evidence.

I responded with a neutral negotiation:

1

negotiate payer=carrier_a rate_posture=neutral

The move was deliberately limited. Riverside had better intelligence, but copying Summit would have required assumptions about bargaining leverage, payer priorities, and organizational risk that the monitoring result did not answer. March ended with 55 cash, a margin of 7, 28 nurses, an access index of 74, and the same six patients boarding in the ED. The clinic network remained six months from completion. Northlake invested in beds again.

This is what the competitive campaign currently feels like to play: read an incomplete institutional picture, commit scarce resources, watch other organizations respond, and reconsider the strategy from a changed position. The opening was neither disastrous nor a clean success. Riverside remained solvent and treated all modeled demand. The strategy improved staffing and reported access while leaving a conspicuous critical-care constraint untouched. It committed future cash before producing new clinic capacity. It gave Riverside better information about one rival as another accumulated beds.

The run had not established whether the access strategy was correct. It had made the cost and timing of that strategy visible.

A Game About Choosing a Theory of the Institution

A dashboard can report Riverside’s cash, staffing, access, and operating margin. The Health Policy Strategy Game asks the player to decide what those numbers mean while the institutions around Riverside continue acting.

The distinction matters because a hospital metric rarely identifies its own remedy. Six boarded critical-care patients might support an ICU investment. Elevated nursing vacancies support recruitment. Access scrutiny creates a reason to make a public pledge. A rival bed expansion may justify defensive capacity, or it may make duplicative investment less attractive. The same report can support several theories of the problem.

The player controls one fictional nonprofit health system, not the region. This is an institutional model rather than a planner’s model. Riverside cannot allocate Northlake’s capital, choose Summit’s payer posture, relax labor-market pressure, or determine how state officials interpret a public pledge. It acts from one position inside the system and receives the consequences of decisions made elsewhere.

That makes diagnosis part of the strategy. Each month, the player has to decide which explanation deserves scarce cash, political capital, and action points. The January report supported a staffing theory, a critical-care theory, an outpatient-access theory, and an information-first theory. Choosing among them also meant deciding which evidence could wait.

The game is therefore trying to make institutional microeconomics playable. Opportunity cost appears through commands left unused. Complementarity appears when buildings need workers before they become useful capacity. Information has a price. Bargaining takes place among organizations with different outside options. Actions create effects beyond the accounts of the institution that chose them. These ideas matter because they change what Riverside can do next, not because the game presents them as definitions.

The same run can function as a reproducible case. It creates a record of what one organization knew, what it chose, what other institutions did, and how those decisions changed later options. The player can argue with that record after the monthly result has resolved.

This article follows two earlier parts of the project’s design lineage. “Fidelity of Consequence: What Capitalism 2 Taught Me About Modeling Systems” describes the standard I carried from an older business simulation. “A Management Game Where the Market Talks Back” explains the architecture I thought a health-policy game would require. The question here is what that architecture now asks a player to decide.

Consequences Travel Through Time

The competitive campaign carries much of its strategy through a compact operating loop:

• regional demand × market position -> system demand
• staffed effective capacity -> treated volume + unmet demand
• treated volume × quality and payer-pressure realization -> operating revenue
• workforce + physical footprint -> operating cost
• revenue - cost -> cash available for later decisions

The first line allocates part of regional demand to each health system according to its modeled market position. The next line asks how much of that demand the organization can actually treat. Beds and clinics are nominal assets; workers determine how much of them becomes effective capacity. Treated activity produces revenue under the month’s quality and payer conditions. Workers and facilities create operating costs. The remaining cash becomes a constraint on later action.

None of those lines is complicated in isolation. The strategic pressure comes from their clocks and connections.

Nurse recruitment gives the staffing line a concrete meaning. Riverside authorized four nurses in January and paid 20 cash immediately. The nurses arrived after a one-month delay. Their arrival improved the staffing position while raising recurring operating cost. A player who reads recruitment as an instant capacity bonus will misunderstand both the timing and the financial commitment.

The clinic project works on a longer clock. A budget of 18 produces a monthly draw of 2 across its nine-month duration. The first draw was included in the January command cost; the capacity increase remained pending until Month 10. During the opening, the project changed cash and occupied one of Riverside’s two project slots without adding a single unit of outpatient capacity. Its future benefit was real inside the model, but so was the option value Riverside gave up while waiting.

Northlake’s decisions occurred on another clock. It invested in beds while Riverside was building outpatient capacity. A project justified by January conditions would eventually mature in a market that included Northlake’s expansion and Summit’s payer behavior. Waiting for more evidence would reduce one kind of uncertainty while conceding time to competitors. Acting early preserved time and increased the risk of committing to the wrong constraint.

The treatment ratio and ED boarding expose a different problem. Riverside treated 24 of 24 modeled demand units in January and February, then 25 of 25 in March. Unmet demand was zero in each aggregate report. Six critical-care patients still boarded because Riverside lacked ICU capacity. The operational experience and the aggregate measure were both part of the state, but they answered different questions.

A team looking only at the treatment ratio might conclude that capacity was adequate. A team looking at boarding could prioritize critical care over outpatient growth. The game does not replace the broad metric with the specific warning. It leaves the player to decide which signal should control the next commitment.

You Control One Institution, Not the Region

The economic loop explains how consequences propagate within Riverside. The surrounding institutions determine why those consequences cannot be evaluated as a private optimization problem alone.

Northlake entered the run with a growth-oriented style. Its decision process favored bed investment and accepted a tighter cash position in pursuit of capacity. Summit followed a margin-oriented strategy and repeatedly favored aggressive CarrierA negotiation. Their commands were generated from their own visible information, resources, and strategy weights. Riverside could observe some actions immediately, learn others by monitoring, and discover still more only in the instructor debrief.

The game rejects a single objective at two levels. Within Riverside, cash, access, quality, staffing, workforce trust, community trust, and political capital can conflict. That is an internal managerial problem: even a coherent leadership team has to decide what the organization will preserve or put at risk.

Across the region, the conflict is strategic. Providers, payers, policy actors, and workforce interests value different outcomes and possess different forms of authority. Summit’s preferred contract may weaken another organization’s revenue position. Northlake may tolerate financial pressure that Riverside would consider dangerous. A state official may value access that no individual system can finance on the same terms. The disagreement exists among objective functions, not only among components of Riverside’s own scorecard.

Information follows those institutional boundaries. Summit’s Month 1 payer move existed in the true simulated history, but it was not in Riverside’s initial observation. Monitoring in February changed Riverside’s information set and made the action visible in March. The end-of-run debrief could later expose Summit’s rationale without pretending that Riverside had known it earlier.

Monitoring does not resolve uncertainty in the broader sense. It reveals a signal, not the payoff function. Riverside learned that Summit had negotiated aggressively; it did not learn whether CarrierA would respond to Riverside in the same way, whether the relationship cost would be acceptable, or whether a neutral posture would prove more valuable over the full campaign. Better information narrowed the decision without producing one correct command.

The action point spent on monitoring also had a visible opportunity cost. It produced no current capacity and left fewer points for staffing, capital, bargaining, or commitments. The player sometimes has to improve the next decision while accepting that the present dashboard will barely move.

Rational Institutions Can Produce Bad Systems

Once organizations act from different positions, a rational local choice can contribute to a weak regional outcome. The game is designed to represent several forms of this divergence.

Financially rational underprovision is one path to that result. Riverside could preserve liquidity by declining an ICU investment even while critical-care boarding remained visible. The regional or patient-facing value of the service need not match the return Riverside could capture in its own cash balance.

Competition can also encourage duplicative investment. Northlake’s bed expansion could improve capacity, strengthen its market position, and invite Riverside to respond with more beds. If both systems protect share by expanding the same service, the region may receive more physical capacity than staffing or demand can support. Competition can produce a response without guaranteeing that the response targets the most urgent need.

Contracting pressure creates a different form of divergence through cost shifting. A payer can protect its spending position through an aggressive rate posture while transferring financial pressure to providers. A health system can answer by delaying recruitment, reducing investment, or emphasizing services with better returns. The payer’s result, the provider’s result, and the regional access result can move in different directions without any actor behaving irrationally.

Riverside’s access pledge makes the evaluative problem especially visible. The pledge raised the reported access index and created a follow-through obligation. It did not create ICU beds, and the clinic project was still months away. The pledge could be strategically useful for legitimacy while remaining incomplete as an operational response.

No universal score can decide how to rank these outcomes. Profit matters because cash preserves an institution’s capacity to act. Access, staffing, trust, and distribution remain separate because financial survival does not tell us who received care, which service stayed constrained, or which institution absorbed the cost. Strategy consists in deciding which values to spend, defer, or put at risk while recognizing that another actor may judge the same result differently.

From Playthrough to Reproducible Case

The run does not end when the last monthly result appears. Each committed transition retains its prior state, the player’s observation, command batches, resolved external inputs, events, attributed effects, resulting state, and a stable state hash. That record turns a playthrough into something that can be replayed and inspected.

The hash has a modest but useful job. If two groups begin with the same seed and make the same decisions, they can verify that they are discussing the same committed history rather than two superficially similar runs. If their decisions diverge, the histories show the first state at which their cases separate. The hash does not explain why a decision was good, validate the model’s parameters, or establish that an actor behaved like a real health system. It keeps the mechanical facts stable enough for those arguments to be precise.

Committed history also preserves the difference between what was knowable during play and what became visible later. A debrief can show that Summit negotiated aggressively in Month 1 while also showing that Riverside learned this only after monitoring. It can distinguish the immediate recruitment cost from the nurses’ later arrival and the clinic project’s delayed benefit from an investment that never worked. A poor outcome can then be traced to a player choice, a rival action, an external realization, a delay, or some combination of them.

The game provides two layers of help for judging decision quality. A bounded automated review flags specific risks, including unsafe cash runway while projects continue drawing funds, recruitment under severe workforce stress, aggressive payer negotiation without sufficient leverage, failure to respond to rival capacity that erodes share, and repeated access pledges without operational follow-through. These checks are diagnostics over named conditions. Passing them means that the run avoided those particular warning patterns, not that the player found an optimal strategy.

The instructor review supplies the context those rules cannot. It reconstructs the player’s actions alongside the observations available at the time, then reveals previously hidden rival actions and rationales for retrospective discussion. An instructor or group can ask whether the player considered feasible alternatives, acted consistently with a stated strategy, protected flexibility, and adapted when new evidence arrived. The game organizes the evidence for that judgment; it does not settle every normative disagreement.

This division makes decision quality different from outcome quality in operational terms. An aggressive negotiation made with weak observable leverage can be criticized even if a favorable realization rescues the month. A cautious investment can remain defensible when a later shock makes it look expensive. A public access pledge can be questioned when several months pass without staffing or capacity behind it, even if the reported index initially improves.

That structure creates a plausible educational use without demonstrating learning empirically. Opportunity cost, complementarity, investment under uncertainty, information acquisition, strategic response, and externalities appear as reasons for making or challenging a decision. A class can compare openings under the same seed. A facilitator can restrict discussion to contemporaneous information before revealing the instructor view. A model reviewer can challenge the nine-month clinic delay, the payer-leverage threshold, or the access effect against a concrete trace instead of debating a general description.

The current evidence supports a narrow proposition: the game can generate reproducible strategic cases whose assumptions and consequences remain open to inspection. Whether those cases improve human learning requires a separate study with human evidence.

Play the Same Opening Differently

The current prototype contains three campaign formats. The five-turn stabilization campaign provides the shortest introduction. The 24-month competitive campaign used here adds simultaneous rival action, lagged observability, recurring operations, and an end-of-run instructor review. A six-stage regional-affiliation campaign explores partner observations, commitments, review, and integration. All three can be played through the command line; a local browser GUI currently supports the competitive campaign through the same observation, command, transition, and debrief boundaries.

The prototype’s present authority is structural. Its numerical units are documented game abstractions rather than calibrated encounters, reimbursement, expenses, or forecasts. The game is not intended for operational, clinical, financial, regulatory, or policy decisions. Behavioral validity, empirical calibration, and measured educational impact require different evidence from deterministic replay.

The Health Policy Strategy Game repository contains the source, design documentation, and play instructions. With Rust and Cargo installed, the command-line game can be started with:

1
2
3

git clone https://github.com/SaehwanPark/hs-mgt-game
cd hs-mgt-game
cargo run

Choose the competitive campaign, Normal difficulty, interactive play, and seed 42 to begin from the same regional setup.

At the start of Month 4, Riverside had more nurses, a stronger reported access index, 55 cash, new information about Summit, and six more months before the clinic network would complete. Northlake had chosen beds three times. Critical-care boarding had not moved.

I can defend the opening as a patient commitment to outpatient access, staffing, and information. I can also see the case for putting ICU capacity first, monitoring before committing capital, or preserving the public pledge until Riverside knew which access constraint it could actually relieve.

The opening is reproducible; its strategy is not compulsory. Playing the same seed with a different theory of Riverside is the most direct way to decide where its managers should have acted differently.