2020-2 Contract Design.fex

Introduction
===

lecturers:
	alexander stremitzer (prof law at bonn)
	nate atkinson (postdoc law & economics)

target:
	how to structure deals strategically
	(off-topic is the legal language to implement this)
	strategically read contracts:
		when cost-plus -> expect cost control clauses
		when fixed price -> expect quality control clauses
	strategically write contracts:
		choose appropriate overall structure
		align incentives to reduce transactions cost

why state enforces contracts:
	well functioning system increases societal health
	as voluntary transactions enhance welfare
	bc seller sells only if price >= valuation
	bc buyer only buys if price <= worth
	worth - valuation = welfare increase
	other arguments:
		companies do not have to prepare for other scenarios
	limitations:
		duress,	misrepresentation, fraud, uninformed decisions
		state must not enforce in these cases
		like parties benefiting by hurting others (hitman)
		like anticompetitive contracts (preisabsprache)
		like externalized cost (selling drug)
		like unknown interests (children, incapacitated)
	enables economic growth:
		specialization generates value, but needs division of labour
		the cheaper trade, the more fragmented division possible
		hence the richer society gets
	transaction cost:
		might make deal more expensive than what its worth
		by this, may also prevent some deals to happen
		need to optimize transaction cost by optimal clauses
		contract writers as "transaction cost engineers"

contractible effort
===

contract motivation:
	coordination:
		agreement about technicalities without conflict of interest
		"cheap talk" is enough (no sanctions needed to enforce)
		like on which street side to drive on generally
		like at which time to deliver so buyer/seller meet
	spot transaction:
		immediate exchange of goods
		like buying an apple & paying for it at the store
	sequential transactions:
		goods/services exchanged, potential conflict in the future
		like sale (need guarantees product meets expectations)

conflict of interest:
	ex-ante (before) and ex-post (after) incentives might be different
	rational choice could be to defect later on
	as other party knows, might prevent contract
	solution options:
		use ex-ante interests to commit to value
		to settle potentially conflicting ex-post incentives 
		repeated transactions:
			smaller contracts / reputation systems 
			works if payoff defect < future transactions benefit
		prevent problem:
			agree on technical solution that prevent problem
			like pulling teeth of viper, smart contracts, ...
		compensation:
			compensate for damages
			prevents problem if compensation cost higher than defection payoff
		criminal sanctions:
			penalize injurer 
			prevent problem if penalty cost higher than defection payoff

value to commitment:
	commit tools:
		governance systems and institutions
		contracts (use formal language to clearly commit)
		promises (willingness to not break a promise, being honest)
		moral institutions (willingness to be fair to others)
		technical solution (like smart contracts)
		enforce by law, continued relationships, reputation, social sanctions
	smaller commits:
		result in smaller temptations to break it
		work with milestones (smaller units of work & payment)
		transforms one-shot to repeated transaction

contract structure:
	define payoff resulting from different clause of action
	depends on social model, contracts, institutions, ...
	general approach:
		outcomes result from human interaction, constrained by institutions
		model behaviour to find out likely outputs
		motivates institutions to be constructed
	model human behaviour:	
		assumptions need to be realistic
		like "more money better", "less work better"
	institution viewpoints:
		evolutionary hypothesis (existing structures exist bc they work)
		social engineering ("triumph of reason")
		piecemeal social engineering (look at existing contracts)
		utopian social engineering (mechanism design, game theory)
	analyse existing contracts:
		be humble (may contain concealed wisdom)
		hence might not be best to simply rewrite contracts
		but game theory / mechanism design helps to improve

contract types:
	care about risk, quality and cost
	choose appropriate type so interests align
	cost-plus:
		seller invoices direct cost + some profit margin
		fit if quality important / monitoring difficult, small companies
		need to control cost, to avoid expected cost overruns
		with cost ceiling, approval for bought products, ...
	fixed price:
		seller invoices full cost in advance
		fit for large / diverse operator
		need to control quality, to avoid expected quality underuns
		with warranties, define product quality, certification
		
contract value:
	work done:
		seller is able to produce goods at some cost
		which is lower than benefit of those to buyer
	risk-shifting:
		shift risk from risk-averse to risk-neutral party
		like large operator can diversify / take risk better
		shift risk to party controlling it
		like seller might has superior information about the product
		price rises due to the risk shifting, but worth to other party
		like insurance contracts

contractible effort examples
===

before/after incentives:
	"the farmer and the viper" (fable by aesop)
	setup:
		viper is in a ditch, and cold / about to die
		farmer wants to buy the viper, but does not want to get bitten
		viper promises to not bite
		farmer takes viper home and warms it up
		viper bites farmer, farmer dies
	trust game model:
		farmer/viper payoff
		(1) 0/-20 not saving viper; viper dies
		or (2) saving viper, then
		(a) 5/5 viper does not bite farmer
		(b) -20/10 viper bites farmer
		hence rational choice of viper to bite
		hence rational choice of farmer not to save
	save the viper:
		need to change the game so viper does not bite any more
		prevent bite by taking tooth out or mouth muzzle
	interpretation:
		at (1) interests align (viper live, farmer save)
		but at (a or b), conflict of interests
		hence need to agree at (1) to solution
		at (2), an agreement is impossible

conflict of interests:
	inventor vs VC:
		inventor / VS payoff
		(1) 1/1 not revealing idea
		or (2) revealing, then
		(a) 6/6 VC cooperates
		(b) -20/20 VS steals idea
		hence VC will steal the idea
	buyer vs producer:
		if payment first, buyer risks low-quality product
		if product first, producer risks missing payment
		more drastic if tailored product 
	insurance vs customer:
		if house burns down, insurance does not want to pay
	asymmetric information sale:
		when selling something other has not full information about
		need assurances, else transaction likely fails
	conclusion:
		client mistrusts if no warranty given as seller has information superiority
		seller needs to grant warranty or else sale will no happen
		hence used common interest ex-ante
		to agree on terms which resolve different interests afterwards

value to commitment:
	warranty clause:
		seller sells only high quality goods, valuation at 60
		buyer values low/high quality at 20/80, but unable to detect quality
		as buyer values product at 50, no sale happens
		by including warranty clause (pay 60 if low quality)
		sale happens, and surplus 20 is created (at buyer or seller)

moral hazard
===

when agents actions are unobservable
need to compensate on something different than effort
creates incentive issues

terminology:
	incentive constraint:
		if incentive exists to act as expected
		for example contingency clause motivating high effort
	participation constraint:
		if deal has expected positive payoff for party
		for example payoff higher than cost of production / risk
	first-best:
		exchange reached if everyone acts in the common interest
		unreachable in practice (ignores private interests of actors)
		first-best risk allocation when company takes all the risk
		is only possible if agent is honest
		contracts needed bc in reality agents use freedom for self advantage
	second-best:
		as close as possible to first-best but including private incentives
		hence the realistic contract actually feasible in practice
		lose some contractual pie to transaction cost, distortion
	zero sum game:
		if no value created (only allocated)
		game-theory approach, but not always true in contracts
		bc mutually beneficial clauses increase contractual pie
	mutually beneficial term:
		introduce clauses where costs of p1 < benefit for p2
		because the contractual pie increases
		hence can compensate other party for loss

contingency contract:
	payoff depending on result of project
	legal remarks:
		by law, some contingency may apply automatically
		"auftragnehmer muss sorgfältig tätig sein"
		=> by law required to deliver up to some quality
		needs not to be spelled out in legal document
		legal doctrine applies automatically

non-contractable effort:
	if effort not directly measurable contingency clause impossible
	when missing:
		describability (high-effort not known in context)
		like travelling sales person; boss is not present
		observability (monitoring might not be possible)
		like software engineer; boss does not understand work
		verifiability (effort not provable in court)
		link employee performance; low effort might not be documented
	alternatives:
		contract on signal:
			contingency clause on signal of effort
			like commission, bonus, stock options, hours worked
		invest in monitoring:
			expand contractable contingencies to be observable
			invest in monitoring, specialized courts / dispute boards
			like cameras on construction sites, ...
		self-enforcing mechanism:
			reputation mechanism for repeated observable interaction
			helps to reduce reliance on verifiability
			needs observability to be useful
			like not extending contract if performance bad

contract on signal:	
	signal types:
		"perfect signal":
			if output deterministically on effort
			simply use contingency contract on signal
		"stochastic signal":
			if output influenced by other factors than effort
			signal is noisy, has some error term
			like general market health, colleagues in the office, ...
			if output reasonable influenced by effort then still useful
	strategies:
		step function:
			payoff as soon as some signal strength reached
			if noise is known, can compensate all high effort equally
			like shifting support scheme
			but if noise not exactly known / error terms
			then may lead to changed incentives throughout the project
			if goal become unreachable or is already reached => low effort
		linear:
			payoff scales with signal strength
			but if noisy, may gives different payoff to same real value

risk sharing:
	create value by shifting risk from risk averse to risk neutral party
	principle:
		"who controls the risk should bear risk"
		like producer controls quality
		like regional firms controls regional law changes
	risk types:
		risk neutral pays expected value
		risk averse pays less than expected value
		risk seeking pays less than expected value
	assumptions:
		risk averse individuals assumption:
			decreasing marginal utility of money
			if low income rather than millionaire
			lifestyle changes more radically with same amount
		prospect theory:
			expected payoff usually not used by individuals
			prefer lower but secure payoff than larger but insecure (risk-averse)
			prefer higher but insecure cost than lower but secure (risk-seeking)
			value own property higher than economic value
		special case lottery participant:
			actively risk-seeking
			bc probably only way to earn that much money
	risk diversification:
		helps to reduce risk
		can take higher risks, but still be risk neutral
		example:
			if owner of buildings & construction company at the same time
			risk of earthquake close to neutralized
		law of large numbers:
			probability that car crashes unclear, risky
			but overall number of crashed cars in some area similar
	risk taking party:
		less risk-averse party should take all risks
		viewpoint conflicts:
			from risk-sharing point of view "first-best risk sharing"
			party which can take higher risks should bear it
			from incentive point of view "first-best incentives provision"
			party which controls risks should bear it
			will lose welfare through transaction cost
			for example manager more risk averse than employer, but controls risk

uncertain output:
	optimal incentive scheme not necessarily increasing in output
	like fast delivery is good, but do not want recklessness
	uncertain signals:
		could be complex / non-linear
		use probability density functions for good/bad measurements
		use likelihood ratio of these curves to determine signal
	mathematical foundation:
		derivation exists on how optimal contract c depends on input/output
		if effort contractible, c independent on output
		if effort not contractible, c dependent on output
		if likelihood ratio high, then agent should be rewarded
		if likelihood ratio low, then signal must be higher for same reward

effort measurement:
	if effort contractable, then contract directly on it
	if effort non-contractable, then optimal contract depends on output
	principle:
		relationship between optimal contract & output
		works through information content of the outcome
	valuable information:
		information must be a signal rather than noise
		(like slow delivery time could just mean full streets)
		signal correlated to other contains no value
	comparisons to other agents:
		do relative performance evaluation to neutralize noise
		like two investors in same market filter out noise

distortion:
	measured goals are targeted by actors if known
	but potential problem if real target is not measured (only proxy)
	"if measurement becomes a target, it ceases to be a measurement"
	could try to be intransparent concerning what is actually measured
	output factors:
		relatively undistorted (like measure sales)
		but very much noise (why is it high / low?)
	input factors:
		very distorted (how to measure effort?)
		but not much noise (can observe single person)

shifting support schemes:
	assume curve known for low/high efforts
	if supports (=ends) of distributions do not overlap
	then can punish where only low effort was a possible input
	no risks for compliant agents 
	like claw back clause (some money not payed out at some signal)
	robustness to error:
		harsh punishments can  start as soon as no overlaps anymore
		but if error in expected curves then could punish good agents
		like incorrectly estimated market risk
		linear contracts perform better with errors

path dependence:
	effort is chosen not only once
	information & incentives might change over time
	like if goal unreachable then step function motives low effort

monitoring:
	output monitoring:
		more prone to error; increasingly when more "downstream"
		external factors can affect output ("noisy signal")
		but easy & inexpensive
		like measuring worked hours, ...
	input monitoring:
		some signals very accurate
		like "hours worked"
		but might be expensive/hard to observe (supervisors, monitoring technologies)
		like "working hard" is again hard to observe
	distortion:
		monitoring might not be able to catch all it needs to
		agent will focus only on tested outputs
		possibly better to not create incentives at all
		like measure testing ability but not actual skill
	reduce monitoring cost (becker):
		make audits probabilistic & increase punishments
		archive same deterrence with same expected cost
		but limited wealth (simply cannot pay punishment)
		but proportionality to crime lost (morally wrong)
		but chilling effect (simply will not do task anymore)
		but behavioural (small probabilities difficult to comprehend)
	rules vs standards:
		fixed rules define how agents have to perform
		easy to enforce but distortion likely, inflexible
		standards define general ways how to behave
		by standards committees & other knowledge authorities
		undistorted & flexible, but vague & difficult to enforce
	incentive rent-extraction tradeoff:
		low commission for distributor results in high profit per sale
		but likely results in only low effort of distributor
		vice-versa with high commission

probabilistic monitoring:
	monitor probabilistic while increasing punishment 
	keeps same deterrence with lower monitoring cost
	conflict of view:
		breaks with traditional view of punishment proportional to crime
		in favour of making it more efficient (mechanism design)
	assumptions (which are likely wrong):
		humans are rational with small probabilities
		no wealth constraints
		no monitoring mistakes
		no "honest" mistakes (need to be able to observe intent)
	conclusion:
		simplifying assumptions make process look rational
		but recommendation not that great when relaxing them
		wrong assumptions introduce chilling effect
		which prevents trade as risk of wrong punishment too great

limited wealth changes risk behaviour:
	limited liability / bankruptcy (can only pay what is available)
	the less wealth, the lower potential downside
	hence higher willingness to gamble upside
	bc do not have to pay cost of downside
	asymmetry in risks and rewards distorts behaviour 
	willingness to game:
		if utility does not decrease
		like take money of company rather than personal reserves
	prevent:
		input monitoring for low-wealth individuals
		safety monitoring 
		capital requirements for banks
		mandatory insurance	

moral hazard examples
===

moral hazard:
	firm (principal) introduces CRM to increase sales
	consultant (agent) is payed to help introducing it
	model:
		consultant can choose high/low effort
		for high effort, cost is 100, company benefit 200
		for low effort, cost is 10, company benefit 20
		high effort joint gain 100, low effort gain 10
	fixed fee contract proposals:
		(1) for fixed fee contract for 150 & low effort chosen
		then principal loses 130, hence no deal
		"participation constraint" violated
		(2) for fixed fee contract for 15 & low effort chosen
		then principal/agent both earn 5, hence contract entered
	(3) contingent contract:
		payment as function of condition
		"if high effort, then 150 payed"
		principal/agent both earn 50, hence deal valid
		agent agrees to constraint bc else no contract
	mutually beneficial term:
		from (2) to (3) cost of agent increased by 100-10 = 90
		but benefit for principal increased by 200-20 = 180
		hence contractual pie increased by 180-90
		use surplus money to pay agent

non-contractable effort:
	performance of manager depends on effort & market condition
	setup:
		manager spends low/high effort (10 / 100 cost)
		company benefit for high effort 50% for 400 (success), else 0
		company benefit for low effort 5% for 400 (success), else 0
		best case output is hence 0.5*400 + 0.5*0 - 100 = 100
	fixed salary:
		company bears all the risks
		manager has no incentive any more for high effort
	step function:
		300 bonus in case of success else 0
		then manager chooses high effort
		hence manager expected payoff 0.5*300 + 0.5*0 - 100 = 50
		hence company expected payoff 50
		best-case reached (as 100 is best-case joint surplus)
		(depending on bargaining power, company or manager earns more)
		like bonus depending on reached goals
	sell shop:
		manager buys company for 50
		then manager chooses high effort
		both again get expected payoff of 50
		like management buyout
	risk sharing:
		for step function; both bear risk
		for sell shop; agent bears all risk
	risk averse manager:
		step function compensation valued at 30 (-20)
		sell shop compensation valued at 20 (-30)
		step function preferred by manager
		but joint surplus only at 80 (-20)
	
risk sharing:
	cell phone insurance by apple:
		able to diversify risks & enough money
		but user controls risks if cell phone is destroyed
		=> insuring increases risk of damage ("moral hazard")
	landlord / tenant agreements:
		risk neutral rich landlord 
		values risk of loosing 100k close to expected cost
		risk averse low income tenant
		values it much higher bc of resulting lifestyle impact	
	insurance:
		risk neutral insurance
		deductible (some low amount beared by the insured)
		limits (max amount payed out)
		co-insurance (some percentage beared by the insured)
		exclusions (exclude some damages, reductions for recklessness)
	mortgage crisis:
		combine mortgage claims in packets (risk diversification)
		distribute to many agents (low price -> risk neutral agents)
		mortgage granted if background check passed for applicants
		and if vehicle score > 620, then selling mortgage in the bonds possible
		observed more defaults > 620, bc banks skipped background checks
		as risk sold out, no incentive to evaluate risk (moral hazard)
		resolve by "skin in the game" -> banks need to keep some part
		many institutions bought these bonds bc AAA ratings & high interests

production / distribution risk sharing:
	distributor D earns 11% + expenses, rest for producer P
	D high effort:
		through payment structure, D only incentive to cover own cost
		introduce monitoring through P (meaningful consulting, ...)
		include non-discriminatory clauses to movies produced by D
		(expected to be non-distorted, bc interests align)			
	P high efforts:
		P has requirements to involve key persons, minimum budget
		P also has exclusivity clause (no different distributions)
	no-cross-collateralisation:
		(movies are accounted separately)
		prevents D taking unnecessary risks after successful movie
		prevents P to stop investing into films after bad movie
	no sell-the-shop:
		pricing unclear, effort unobservable

uncertain output:
	truck driver delivery speed
	fast & intact delivery suggests good job
	but too fast delivery suggests reckless driving
	probability density function:
		deliver speed curve c_good of good driver more to right
		deliver speed curve c_bad of bad driver more to left
		calculate likelihood ratio c_good / c_bad
		as probability density gets higher, more likely good driver
	include reckless driving:
		curve of bad driver increases for very fast speeds again
		hence likelihood ratio decreases again at the end

path dependency:
	bridge construction:
		bridge designed for 5 year construction
		in the final phase cable & damper technology installed
		base payment & big bonus if some quality reached
		but potentially during the project goal reached or unreachable
		hence low/high effort incentive change
		
monitoring:
	sales company:
		unclear if contingency should be placed in sales or on monitoring results
		sales could be too noisy or monitoring too expensive
	southwest airlines:
		employees with large profit sharing for employees
		efforts of single employee has no real effect (signal is noisy)
	wristband amazon:
		allows to calculate how effective employees perform
	teacher bonus:
		paying teachers based on PISA test results
		teachers will shift priorities to actually tested skills
		might leaves out social skills, critical thinking, ...
		might motivate to cheat
	football:
		get bonus with each goal
		players argue over who shoots, rather than the best does
	sears:
		commission based on repairs
		unnecessary repairs / disgruntled customers
	salefites:
		commission for each windshield installed
		but if its broken, then had to fix for free
		worked well
	manager pay based on accounting:
		managers might reduce R&D for short term benefit
		use stock options, claw clauses, long-term payout to combat
	agency marketing:
		could contract to spend fixed amount on marketing specific product
		but distortion possible, potentially nonsense amount
		likely better to agree on "best effort" standard of industry
	distributor:
		with commission-based sales difficult to balance commission
		with directly sold goods, distributor needs enough money to do so
		could add reputation / repetition system

probabilistic monitoring:
	rent car:
		forgot to pay for the gas at gas station
		able to pay later + potentially a fine
		with probabilistic monitoring, high fine / jail time required

limited wealth problem:
	manager bonus:
		manager gets bonus of 10% of companies benefit
		investment 1 has 50% that payoff 100, else 0
		company expected = 50, manager expected = 5
		investment 2 has 50% that payoff 100, else -200
		company expected = -50, manager expected = 5

legal regimes
===

how background rules affect contract design
courts may not enforce contract as written

tasks of contract designer:
	economic problem (mechanism to minimize transaction cost)
	legal problem (get court to enforce mechanism as intended)

consideration:
	to form contract consent is enough in civil law
	else need "consideration" (own behaviour somehow restricted)
	contracts without exchange potentially not enforceable
	could ask for small monetary amount or "good faith"
	but courts could also imply good faith without it being stated

contract enforcement:
	formation:
		no contract exists
		requirement as specified in contract unmet (like physical signature)
		like amount > 500, duration > 1y need written contract (USA)
		like sale of land without notary (CH)
		like only promise exists (bc promise != contract)
	defenses:
		contract not enforceable
		like misrepresentation, unconscionability
		like public policy
	interpretation:
		contract means something else
		most legal problems happen in this area
		like price unreasonable for informed parties
	excuses:
		non-performance is excused
		impracticability / frustration of purpose
		like rented concert avenue burning down & owner not responsible
		then can not sue him for lost money
	remedies:
		desired remedy (payoff) not available
		under-compensatory damages (for example for immaterial damages)
		like immaterial harm (band misses performance, privacy violations)

remedies:
	punitive:
		additional to compensation get punishment
		like additional money or material payout
		less common in civil law
	compensatory:
		specific (breaching party forced to perform)
		substitutional (other remedy payed out)
	substitutional remedy amount:
		expectation (as much as performance would have resulted in)
		calculate like promised value - received value
		reliance (restore pre-contract state)
		calculate like value before - value now + damages incurred
		restitution (reverse all exchanges; damages are not reimbursed)
		litigated damages (contract predetermined amount payed)
		default is expectation damages		
	litigated damage clauses:
		only enforced if not considered excessive, else "penalty clause"
		but what may seems excessive ex post might be reasonable ex ante
		convince court and/or declare explicitly in contract

courts:
	may not always enforce contracts such as it is written
	will try to interpret contract to intended meaning
	invalid clauses:
		when unconscionable or against public policy
		if appearing unfair (but might only be the case ex post!)
	constructive conditions:
		when conditions obviously depend on each other
		does not always has to be specified explicitly
		like when customer does not pay in sell contract
		then do not have to deliver good
	in court:
		understanding the deal structure (ex-ante incentives) 
		helps to make arguments & counterarguments in litigation

legal regimes examples
===

substitutional remedy:
	serious injury on hand results in work defects
	doctor persuades patient for experimental hand surgery
	doctor promises 100% restoration of hand function
	but surgery went terribly wrong
	model:
		values always pre-op, post-op and promised
		value of hand 5000/2500/10000
		cost of operation 0/1000/1000
		pain from surgery 0/1500/1500
		assuming doctors experience benefit worth 2000
	substitutional remedy amounts:
		expectation damages 10000 - 2500 = 7500
		reliance damages 5000 - 2500 + 1000 + 1500 = 5000
		restitution damages 1000 + 2000

hard to enforce:
	NDA:
		source of information is hard to determine
		hard to enforce liquidated damages ex-post
		hard to show actually incurred damages

ex-ante remedies to enable contract:
	inventor wants to work with VC for more success
	but risky bc VC could disclose idea
	model:
		if inventor does not disclose idea to VC (1/1)
		else inventor discloses to VC
		(1) if VC performs, then (6/6)
		(2) or VC does not perform (-20/20)
		not performing two possible outcomes
		(2a) if VC simply does not perform (50% chance) (0/20)
		(2b) else VC additionally discloses idea (50% chance) (-40/20)
		no deal bc VC will choose not to perform
	contract + expectation damages:
		if VC simply does not perform (50% chance) (6/14)
		else VC additionally discloses idea (50% chance) (6/-26)
		hence VC does not perform payoff (6/-6)
		deal bc VC will choose to perform
	assume (2a) and (2b) not distinguishable / (2b) not quantifiable:
		if VC simply does not perform (50% chance) (6/14)
		else VC additionally discloses idea (50% chance) (-34/14)
		hence VC does not perform payoff (-14/14)
		no deal bc VC will choose not to perform
	assume liquidated damages of 16 or higher:
		court only needs to observe breach (not quantify damage)
		if VC simply does not perform (50% chance) (16/-4)
		else VC additionally discloses idea (50% chance) (-24/4)
		hence VC does not perform payoff (-4/4)
		deal bc VC will choose to perform
	assume liquidated damages considered excessive:
		then in case (2a) only remedy of 6, hence only (6, 12)
		hence VC does not perform payoff (-9/9)
		no deal bc VC will choose not to perform

plain language vs intended meaning:
	beanstalk (marketing) vs AM general (hummer cars)
	deal:
		beanstalk markets brand
		gets 35% of all deals including hummer
	conflict:
		AM general sells brand to GM motors
		but does not want to pay 35%
	question:
		what was the mutual intent of parties at time of written contract
		plain language (as it was written) or would that be absurd?
	argument:
		plain language needs to be interpreted 
		as how reasonable parties would have agreed to deal
		beanstalk is in the trademark business 
		and hence gets commission for sales, merch, ...
		but not entitled to commission for sale of whole company
		as hardly intended in clause, as not contributed to it
	counter arguments:
		parties did not want to clear up all cases 
		to prevent conflicts & sign fast
		but possibly were aware of this potential outcome

asymmetric information
===

in contrast to complete information yielding efficient outcome
with asymmetric information only second-best possible
happens when one party knows something the other does not

phenomena:
	(low value) seller gets information rent or
	high value goods are not traded
	in continuous quality case, both happen at same time
	buyer tradeoff:
		allocative efficiency (good goes to party valuing it highest)
		extracting information rent (overpaying for good)
	choosing a price:
		if low price, then information rent
		but no allocative efficiency (high quality units not sold)
		if high price, then allocative efficiency 
		but information rent (low quality sold for price of high)
		if intermediate price both at the same time

unravelling of the market:
	model:
		seller can observe quality, buyer can not
		continuous value range
		buyer value = seller value + 10
	first-best:
		with complete information, all goods are sold
		as each good is worth more to buyer than to seller
	offer 50:
		average value to seller is 50, hence average value to buyer is 60
		any seller value lower than 50 will sell
		50% chance of trade, average buyer value is 25 + 10 -> payoff = -7.5
		allocative inefficiency (because only 50% trade)
		information rent payed for (many goods value lower than 50 sold)
	offer 25:
		same problem as offer 50
		25% chance of trade, average buyer value is 12.5 + 10 -> payoff = -0.625
	offer 5 (chosen by buyer):
		5% chance of trade, average buyer value is 2.5 + 10 -> payoff = 0.375

eliminating information asymmetry:
	costly, sometimes impossible
	why:
		payoff when selling low value goods (information rent)
		but high quality trades will not be possible anymore
	third party verification:
		have expert / rating systems examine good
		but costly & risk of collusion
	require disclosure:
		presumes information is verifiable at some point
		background legal rules which enable credible communication
		(misrepresentation / fraud laws)

screening:
	when uninformed party creates contract
	customers will all behave the same ("pooling utility")
	but want to charge different prices ("differential utility")	
	pooling equilibrium:
		customers all take same contract
		but contractual pie smaller than it has to be
		due to allocative inefficiency (not all goods bought)
		or information rent payed (too high a price)
	separating equilibrium:
		customers take different contracts
		use differently appealing aspects to sell customer types
		based on differential utility, design multiple types of contracts
		maximizes joint surplus (both low / high quality goods sold)
		like support service to frequent users
		like excessive working hours for high quality employees

signalling:
	when informed party creates contract
	good types might accept inefficient terms to signal high type
	like to sell warranty as a premium product seller
	offer generous warranty:
		expensive, but even more for low quality
		distortion at the top (as behaviour of top producers changes)
		no distortion at the bottom (low quality producers will never offer warranty)
		but creates moral hazard
	invest in brand:
		costly commit to high quality (marketing, actually produce high quality)
		generate high brand value at customers
		then very costly to defect, as brand value is lost instantly
		for unknown brand less a problem (as not much to lose anyway)
	education & signalling:
		education more costly for low-types, hence signal for high-type
		hence invest in education independently if employee benefits
		same reasoning for low-salary internship which high-type accepts
		as reasonable probability afterwards for job offer
		in contrast to low-type which will be detected

distortion:
	behaviour changes suboptimally based on contract setup
	with asymmetric information:
		can no longer fully trade in all cases (first-best unreachable)
		design distortion to include private incentives
		to make the contractual pie as large as possible
		might be necessary to reach optimal second-best contract 
	bottom vs top:
		top distortion when high quality goods do not trade 
		like excessive education only to signal value
		bottom distortion when low quality goods do not trade
		like chip producer corrupting cheap chips

contract negotiation:
	good contractual solutions might exist
	but difficult to negotiate
	inefficient terms:
		inefficiency is accepted
		because else negative signal might be sent
		like prenuptial agreement, tenant contract
	default terms:
		when contractual incompleteness likely
		create default terms to replace negotiations

asymmetric information examples
===

asymmetric information:
	overlapping prices for high/low quality car:
		low /high quality undetectable for buyer
		buyer simply assumes with 50% probability if high/low quality
		setup:
			high quality car1 (buyer value 100, seller value 50)
			low quality car2 (buyer value 50, seller value 20)
		first-best:
			sell car1 for 50 < price < 100
			sell car2 for 20 < price < 50
			bargaining power determines exact price
			all cars are sold
		price 20:
			payoff buyer 15, payoff seller 0
			only low quality cars traded; 50% no trade happens
		price 50:
			payoff buyer 25, payoff seller 30
			seller of low quality car archives "information rent" of 30		
	non-overlapping prices for high/low quality car:
		low /high quality undetectable for buyer
		buyer simply assumes with 50% probability if high/low quality
		setup:
			high quality car1 (buyer value 100, seller value 70)
			low quality car2 (buyer value 50, seller value 20)
		price 20:
			payoff buyer 15, payoff seller 0
			only low Q cars traded; 50% no trade happens
		price 70:
			payoff buyer 5, payoff seller 25
			seller of low quality car archives "information rent" of 50

unravelling markets:
	insurance:
		for high risk people (conditions, dangerous lifestyle) very valuable 
		if single price set, unattractive to low risk people 
		but then prices for remaining high risk people go up 
		moral hazard problem (cost of dangerous activities sink)
		adverse selection problem (high risk behaves differently as low risk)
		=> mandate insurance
	borrower rates:
		good borrowers will reject bad rates
		banks left only with bad borrowers

screening:
	selling software:
		frequent users to which program is essential
		infrequent users to which program is nonessential
		frequent will disguise as infrequent
		if high price, then infrequent will leave
		if low price, then frequent will gain information rent
	after sales service:
		service more valuable to frequent users than to infrequent
		offer high price & after sell service -> frequent have to choose this
		offer low price & no after sell service -> infrequent will choose this
		but distortion that infrequent users do not get service anymore
		even though it might still be efficient to provide
	hire for indifferent employees:
		want to hire employees but can not differentiate ability
		if high wage, then low quality get too much information rent
		if low + high wage, then low quality will pretend to be high
		assume high types work more efficiently than low type
		offer high wage + excessive hours, low wage + normal hours
		low types pick normal hours (as rest is not doable)
		high types pick excessive hours (to get good wage)
	other screening examples:
		phone company tries to sell expensive abo
		airline makes first class nicer, second class worse
		chip producer sabotages chip to sell functioning version for higher price

signalling:
	excessive education:
		as low-types much harder to finish
		invest in education even if non-beneficial to employer
		as it signals high-type person
	waiving access to bankruptcy:
		costly for anyone, but much more to low-type company
		hence good signal; without it loans might not be granted
		but in many cases law prevents waiving access

distortion:
	prenuptial agreement:
		both spouses have considerable assets
		in case of divorce, lawyers could be costly
		hence create contract to avoid dispute
		but suggesting the contract raises doubts
	tenant contracts:
		include clauses that are invalid
		but if spoken about, then contract will not happen

"cover all your bases":
	the more the contract covers the easier to handle in court
	includes already enforced legal doctrines / implicit claims
	like copyright, implicit trust assumptions
	representation:
		assert something to be true in the moment
		to enforce, have to show it was untrue in the moment
		useful if sometimes fact unknown; allocates risk to claimer
	warranties:
		give concrete warranty to outcome or effort
		easier enforceable in front of court
	indemnify (=reimburse):
		"should anything I say be untrue, I will indemnify"
		spell out misrepresentation law
		claimer will pay for damages caused by misrepresentation

elicitation mechanisms
===

how to get private values ("elicit") of other party

asset specific investment:
	asset which is more valuable in specific relationship than else
	hence not protected by the market, lower resell value
	like non-adaptability to other purposes / customers
	with high specificity, need long-term commitment for other party

buy-sell agreements:
	work well with equal partner (similar ownerships / financial capabilities)
	protect parties in case can no longer do business together
	put-call provision (texas shootout):
		one party proposes price p
		other must buy all shares or sell its own for price p
		in general works to buy/sell at reasonable price
		if A offers p, then B payoff is abs(v_b - p)
		if A knows private value v_b, can choose unfair offering price p
	unequal partners:
		some value might be relationship specific
		more valuable party can threaten to leave

auctions:
	take advantage of buyer's private value
	flat price either too high or too low
	auction process:
		sealed bid when placing without other knowing value 
		public auction when others learn value
	revenue equivalence theorem:
		if bidders are risk-neutral & values independently drawn
		then all auctions lead to same expected revenue for seller
		& are efficient (person with highest value gets item)
		assumes "independent" value (own valuation does not depend on other's)
	model:
		item worth some value to me
		others have independent, uniform distributed value 0-100
	first price auction:
		highest bidding buys for price of highest bid
		bid value - value/n
		might lose bid, or might make profit = value/n
	second price auction:
		highest bidding buys for price of second bid
		bid honestly
		not higher (bc maximum price want to pay)
		not lower (bc do want item)
	split the difference scheme:
		highest bidding buys for price between highest/second bid
		bid value - value/2n
	all-pay auction:
		all bidders pay, highest bidding gets item
		bid value^N - value^N/N
	common values:
		if value of items depends on what others value is
		then winners curse (winnner more optimistic than all others)
		like IPO (bc want to sell stock later like others)
		like collection items (bc want to resell)
		like diamond sights (box full of diamonds, very short time)
	defend against winners curse:
		take into account what others bid
		reevaluate / self-reflect in case own bid highest
		choose different mechanism (like second price auction)
	collusion:
		likely if few participants in auction (2-4)
		prevent designing contract only few parties can actually take

impossibility theorem (myerson-sattlerwhaite):
	if one party owns good, the other wants to buy
	but uncertainty about gains from trade
	there is not efficient way to trade 
	human element:
		in models, would never reveal true values
		but in real work, human element outperforms game theory
		parties engage in cheap talk
		parties reveal their values, coordinate on price
		face-to-face communication improves behaviour (compared to written)

elicitation mechanisms examples
===

goethe:
	wanted to know own value
	unclear bc no copyright, sales unknown, ...
	game:
		sends minimal price p1 to attorney
		attorney receives price of publisher p2
		if p1 < p2 then no trade
		else trade for price p1 and learn about p2
	publisher view:
		will bid private valuation as no information about goethes p1
		might understate if interested in long-term relationship
	goethes view:
		wants to learn own market value
		hence better than take-it-or-leave-it offer
		but goethe might renegotiate with learned value
	what happened:
		attorney told publisher what minimal price was
		goethe could sell, but did not learn own price
		resulted in big profits for publisher
	later auction:
		received much more of the share
		as not all publishers colluded

relationship specific investments
===

existing vs redesigning mechanisms:
	redesign with rationality (auguste conte):
		create better society by redesigning it ("social engineering)
		rationally improve over naturally evolved society
		approach of "mechanism design"
		assumptions about human rationality / behaviour
		lead to mathematically reasoned mechanism 
		to archive first-best / second-best
	accept state of the art (edmund burke):
		old institution performed reasonably well in the past
		hence constructing everything from scratch is dangerous
		underlying assumption that evolution evolves into something efficient
		(which is not necessarily true)
	hidden wisdom (friedrich august von hayek):
		presumption of hidden wisdom of institutions
		"fatal conceit" (einbildung) one could redesign all
	piecemeal social engineering (karl popper):
		iterative process; gradually improving upon state of the art
		small, evidence based changes
		archive progress without redesigning all

contract completeness:
	complete:
		condition on every possible state (if contractable)
		only moral hazard / adverse selection risk remains
	incomplete:
		might not include all contractable possible states
		might renegotiate later if enforcement inefficient
		bc bounded rationality (cannot predict all possible states)
		bc transactional costs (creating contract too expensive)
		bc non-verifiability (non-verifiable input/output)
		bc institutional comparison (use existing as a guideline)

hold up:
	if non-contractable relationship-specific investment has to be made
	but proceeds of this have to be shared with other party
	like renegotiation due to changed circumstances after investment sunk
	introduces underinvestment (trying to reduce relationship specificness)
	reasons:
		trade after production:
			parties (re)negotiate trade after production
			at point in time where investment already sunk
		relationship-specific investment:
			investment worth more inside relationship than outside
			like investment specific / customized to customer
			like specific skills, machines, software, processes
	prevent holdup:
		less sequential contracts:
			combine exchange / production; for example with milestone contracts
			like in construction contracts
		write contract before "fundamental transformation":
			before relationship-specific investment happens
			leads to first-best if investment is indeed contractable
		avoid renegotiation (sign complete contract):
			avoid having to negotiate after some investment sunk
			but very hard / impossible to write complete contract
		long-term contract:
			enter long-term contract with other party
			allow investing party to prolong contract by itself
	prevent holdup examples:
		improve BATNA of investing party
		or reduce BATNA for not investing party
		fixed price + specific performance:
			investing party is guaranteed performance
			but ex-post inefficient (goods produced even if inefficient)
			but overinvestment (as can always sell goods)
		fixed price + expectation damages:
			investing party will be payed out expectation performance
			only produce iff price - cost > expectation damage
			for expectation damage = value - price
			hence contract breached if value < cost (which is good!)
			hence ex-post efficient (goods produced only if production worth)
			but overinvestment (goods with price - cost < expectation damages)
	improve contract:
		design threat point:
			for investing (hence weaker) party as a BATNA / outside option
			like acquiring ownership of underperforming party's tool 
			like deciding whether to continue contract 
			like option to increase/decrease default quantity sold 
			like exclusivity (preventing to trade with others)
		reduce sequential structure:
			introduce milestones with deliverables on both sides
			reduces one-shot character
		assign bargaining power:
			to investing party (hence weaker) party
			then efficient investment (socially optimal) will occur
			unclear if realistic (bc of incomplete contracts/renegotiation)
			like take-it-or-leave-it offer, smart contracts

renegotiation:
	solves ex-post inefficient trade
	bargaining occurs with thread point in mind
	coarse theorem:
		besides renegotiation cost, efficient allocation occurs
		unless personal conflicts, time pressure, strategic games
	for continuous trade decisions:
		balance threats of max sales vs no sales
		by contracting in between
		then renegotiating with more infos

vertical integration:
	as way to redesign outside option / BATNA
	coarse:
		if market good, why do firms exist
		=> market needs transaction cost to trade
		if firms good, why not all done by same firm
		=> too big firms inefficient
	williamson:
		avoid holdup with vertical integration
		firms should produce specialized goods internally
		firms should buy standard products on the market
	grossman & hart:
		vertical integration allows to redesign threat points
		(hence can easily walk away from bad negotiations)
		but does not solve unobservability problems

legal solutions:
	complete contracts avoid having to renegotiate
	but renegotiation offers flexibility to avoid inefficiencies
	economic duress doctrine:
		limits that parties can exploit each other through renegotiation
		if nothing changed since agreeing to first contract
		then modifications after renegotiation can be reversed
		else likely that one party used unfair leverage to over-compensate
		damages for breaking contract might not be enough
		bc other party unable to pay, damages non-materialistic, ...

technical solution:
	must be robust to error
	doomsday machine:
		prevent attack by threatening massive retaliation
		not ex-post efficient (hence not negotiation proof)
		for threat to be credible need to build doomsday machine
		which excludes human intervention (and enemy knows about it)
	smart-contract:
		might allow to exclude renegotiation
		might allow to assign bargaining power
		anonymity critical otherwise legal system could revert effects

relationship specific investments examples
===

example incomplete contract:
	fisher body (FB) supplies general motors (GM)
	FB build new factory next to GM has joint surplus 100
	pro vs con:
		pro bc lower cost, integrate better with GM
		cont bc GM may capture part of surplus
	no contract:
		with equal bargaining power, payoff 50
		with cost 40, will build it (and should!)
		with cost 80, will not build it (but should!)
		hence underinvestment could occur
	fisher negotiation power after plant built:
		threaten to withhold does not work
		bc investment already sunk
		threaten to sell does not work
		bc investment worth more inside relationship than outside
	contract:
		predetermine price or exclusive rights
		but depending on market, price/quantities need to change

relationship-specific investment:
	fishing company:
		fishing company hires worker for short fishing season
		then processes the fish in plant with large investment
		shortly before fishing season, workers could renegotiate
		company has no choice but to increase salary, else loses investment

ex-post inefficient trade:
	wood table production:
		agree on wood table production with specific wood
		but wood not deliverable anymore
		cost for wood higher than benefit to buyer
		with renegotiation, both parties can benefit

assigned bargaining power:
	model:
		buyer could investment to increase value to seller
		seller cost is unknown until buyer made investment
		buyer investment can be 0/2/5.5, resulting in gains of 4/10/14
		seller cost is 50% 1 else 12
		for investment 0, benefit for low cost 3 else -8
		for investment 2, benefit for low cost 9 else -2
		for investment 5.5, benefit for low cost 13 else 2 
		socially optimal is investment 2 with expected benefit 2.5
		but trade (ex-post) only occurs if cost lower than price
	no contract (ex-post efficiency):
		assume buyer has 30% bargaining power
		then buyers optimal investment is 0 (underinvestment)
		as 0.5 * 0.3 * 3 (archived price) - 0 (investment) = 0.45
		assume buyer has 100% bargaining power
		then buyers optimal investment is 2 (socially optimal)
		as 0.5 * (10 - 1) (gain - cost) - 2 (investment) = 2.5
	fixed price + specific performance:
		assume price must be 2 (seller only produces for low cost)
		then buyers optimal investment is 5.5 (overinvestment)
		as (14-5.5) (gain - investment) - 2 (price) = 6.5
	fixed price + expectation damages:
		assume buyer must be put in as good position as if contract performed
		then buyers optimal investment is 5.5 (overinvestment)
		as (14-5.5) (gain - investment) - p (price) = 8.5 - p
	actual resolution:
		exclusivity & cost-plus contract
		but FB overstaffed to increase costs
		GM bought FB in the end

relational contracting
===

when able to rely on relationship & reputation
in absence of formal contract

repeated cooperation:
	preconditions:
		gains from future cooperations sufficiently high
		people care about the future (discount rate sufficiently high)
		no definite end to cooperation
	discount rate:
		rate < 1 which future events are less worth
		bc uncertain about future, today preferred to tomorrow
		payoff = payoff_today + rate * payoff_tomorrow + rate^2 * payoff_after_tomorrow ...
		payoff = 1/(1-rate)
		for rate 0, only today relevant
	grim-trigger strategy:
		cooperate unless other player defects
		as soon as other defected, defection forever
		cooperation happens for rate >= 0.5
	tit-for-tat:
		mirror last action of other player
		if other cooperates, then cooperate
		if other defects, then defect
	predefined end of cooperation:
		no cooperation in last period
		but other player anticipates, hence no cooperation either
		unravelling (hence no cooperation at all)
		in practice incomplete (fail to foresee, late defection preferred)
		reputation & contracts as substitutes
		like retirement
	probable end of cooperation:
		some future event with probability p will terminate cooperation
		payoff = payoff_today + rate * (1 - p) * payoff_tomorrow + ...
		payoff = 1/(1-rate)(1-p)
	spike of payoffs:
		some future event payoffs for defection will spike
		cooperation unsustainable
	increasing payoffs in the future:
		in future event payoffs will increase gradually
		cooperation more sustainable

relational contracting examples
===

cooperation:
	price wars amongst gas station in vancouver
	after demand shock (likely some player defects)
	price goes up/down & is more or less chaotic
	then after some months, again cooperation on some high price
	collusion can not be contracted, hence had to relay on cooperation

reputation:
	diamond exchange in new york by orthodox jews
	much trust amongst diamond merchants
	defection increases reputation cost

business relationship:
	business partner & self
	both may choose to cooperate & defect
	payoffs:
		both cooperate (1,1)
		both defect (0,0)
		one cooperates, one defects (2, -1)
	one-shot:
		same as prisoners dilemma
		both will defect, anticipating defect of other party
	repeated:
		if discount rate sufficiently high
		then cooperation chosen

predefined end (centipede game):
	parties can choose whether to give other party some amount or not
	if money received, get additional reward from bank
	game stops if both parties have 100 received
	but predefined end, hence unravelling expected
	in practice:
		5% end at the end (both get 100)
		5% do not event start (as parties predict defect)
		rest ends up in the middle
	reasons:
		players do not foresee 100 steps
		players want to delay defection

externalities
===

parties enter contract which sets them better
but may hurt third parties / social welfare
government imposes regulation to prevent inefficient outcome

exclusivity contract:
	agree on price p with firm to be dealt with
	agree on damages d on deals with other firm 
	problems:
		while increasing joint payoff of involved parties
		decreases societal welfare as market entered late by competition
		only after efficiency gain bigger than damages
	benefit:
		prevent free-riding
		like advertisement benefiting whole industry
		mitigate hold-up problem
		like stronger partner abuses other
		align incentives
		like partner can focus on other
		improve certainty
		like knowing supply for next years

externalities examples
===

exclusive dealing:
	buyer can buy from incumbent (existing) or entrant firm
	buyer accepts exclusive dealing contract with incumbent
	contract specifies price p and damages d (if dealing with entrant)
	model:
		cost of incumbent 50 and entrant c = uniform(0,100)
		buyers value of good 100
	order:
		buyer accepts contract with incumbent (price, damages)
		entrants cost is determined (between 0 - 100)
		entrant decides whether or not to enter market
		production and trade take place
	social welfare:
		if entrant cost < 50 then entrant produces
		else incumbent produces
	no contract:
		if entrant cost lower, then prices lower than incumbent 
		else entrant stays out of market (bc incumbent undercut)
		but then incumbent monopoly, could charge 100
		buyer welfare 25 = 0.5 * 50 (if entrant sells) + 0.5 * 0 (if incumbent sells)
		incumbent profit 25 = 0.5 * (100 - 50) (sell at price 100, cost 50)
		entrant profit 12.5 = 0.5 * (50 - 25) (average cost 25)
		social welfare 62.25 = 25 + 25 + 12.5
	exclusive dealing for p = 75 and d = 100:
		buyer welfare = 25 (always buys for 75)
		incumbent profit = 25 (always sells for 75)
		but entrant profit = 0 (no sale)
		social welfare 50 = 25 + 25 + 0
		buyer indifferent bc same payoff than before
	exclusive dealing for p = 75 and d = 50:
		buyer welfare = 25 (always buys from 75, either from incumbent/entrant)
		incumbent profit 31.25 = 0.75 * (75 - 50) + 0.25 * 50 (damages)
		entrant profit 3.125 = 0.25 * (25 - 12.5)
		social welfare 59.375 
		buyer / incumbent accept as higher payoff

tort victims:
	two victims with claim of 300, lawyer costs 400
	together would sue firm and get full 600
	but single victim accepts settlement 100.1 (> (600-400)/2)
	then other victim unable to sue anymore (400-300 < 0)

selling your business:
	worth of 50, two buyers with 90 / 100 valuation
	bargaining power is equal
	order:
		negotiate first with buyer1
		then with buyer2
	no contract:
		agree with buyer1 to price of 70 (90-50 / 2)
		then agree with buyer2 to price of 85 (100-70 / 2)
		payoff seller 35, buyer1 0, buyer2 15
	breakup fee:
		agree with buyer1 to price of 85, breakup fee 10
		then agree with buyer2 to price of 97.5 (100-95 / 2)
		payoff seller 37.5, buyer1 10, buyer2 2.5

observable but not verifiable
===

informational environment where performance observable
but not verifiable (hence proofable in court)

discretion clauses:
	assigns strong power to one of the party 
	like to terminate contract whenever see fit
	but invites opportunism of such party
	reputation:
		abusing clause will prevent further exchanges
		works if reputation of accessible & important of stronger party
		like mc-donalds reputation under franchises
		but franchise reputation irrelevant (can just switch provider)

up and out:
	company policy that after some time promoted or let go
	to fulfil commitment to promote hard working employees
	removes any temptation not to promote

giving bonus to winner of tournament:
	verifiable for all that some winner have been chosen
	removes temptation not giving bonus
	will give bonus to best employees as no other incentive

shoot the liar:
	transform what is observable to verifiable
	ask observers separately to report on what has happened
	reward if reports match, punish otherwise
	hence if no no way to coordinate beforehand
	then will report the truth as highest probability reports match
	
observable but not verifiable examples
===

not verifiable:
	employee performance
	group member performance 

franchise business:
	charge fee for concept & get marketing, design, procedures
	owner of restaurant has to invest into building
	includes clause to terminate freely without cause (discretion)
	reasons for clause:
		free-riding (save money but reputation cost payed by brand)
	advantage:
		can use discretion to terminate underperforming franchise
		might be hard to prove underperformance in front of court
	disadvantage:
		could abuse discretion to terminate well-performing franchise
		to take over restaurant and operate it self
		for franchise discretion clause like having no contract
		hence franchise will underinvest

behavioural
===

contract are way to give commitment
to reach first-best need to align incentives
but might need additional behaviours

shading:
	perfunctionary performance (follows contract to the letter)
	has to be delivered, enforced by court in any case
	consummate performance (follows contract to the spirit)
	followed by parties if their interest ("sharding")
	avoid shading by including more specific conditions in contract
	but too fix contract may deliver inefficient outcome

promises:
	use promises to get commitment
	where central enforcement costly / impossible
	relying vs expectation:
		"expectation" when only set internally 
		"relying" when basing decisions / investments off it
	why promises are kept:
		third party enforcement
		like begin forced to perform or having to pay damages
		reputational concerns
		like value of future relationship is lost
		moral force
		like value system motivates to keep promises
	observed moral effects:
		expectation per-se effect
		want to deliver expectation (irrespective of promise)
		promising per-se effect
		want to keep promise (irrespective of expectation)
		interaction effect
		if promise given even higher effort to fulfil expectation
	overreliance:
		to guilt-trip others into performance
		introducing legal enforcement makes investments more stable
		less underinvestment (as enforcement exists)
		less overinvestment (as guilt-trip can disappear)
	punishment by third party:
		if third-party observes interactions
		will likely punish promisor
		increasingly so if promise given, expectation high
	irrationality:
		need to be able to understand / predict other party
		but rationality assumption might not be suited
		might results in additional negotiation power 
		like vietnam/kissenger, fighting witches

implications for contract design:
	include difficult to enforce clauses (unverifiable, ...)
	encourage explicit promise making (oaths, ...)
	explicitly invite expectations (advertisement, ...)
	make expectations visible (link expectations / promises)
	undesirable effects:
		even efficient breach is avoided
		might circumvents public policy/unconscionability defenses
		parties feel compelled to fulfil invalid clauses

realistic goals:
	might be desirable to set realistic goals
	lower performance observable with too high / low threshold 
	spillover from aspirational goals to realistic goals observable
	low variance from low threshold
	it matters if people are in win / loss frame
	impacts constitutional design, international agreements
	
legal theory:
	expectations / reliance relevant for legal enforcement
	legal enforcement solves over and underinvestments

questions:
	50% do not follow moral objectives when unpunished
	=> how to generalize

behavioural examples
===

trust / guilt:
	kindergarden example:
		parents had to pay fine being late
		but now can pay-out of moral obligation
		parents came even more often late		
	50% suspectible to guilt:
		in study could choose to keep / break promise
		no reciprocity / reputation effects
		50% broke promise for monetary benefits
		25% traded off money against morality
		25% did keep integrity not matter cost
		might be dependent on social circles / cultural effects 

irrationality:
	scorpion and the turtle:
		scorpion asks turtle to transfer over shore
		but scorpion stings turtle in the middle of the river
		turtle asks "why", scorpion responds "bc its in my nature"
	kissenger:
		pretended in negotiations US president is crazy
		and would drop nuclear bomb on vietnam
		if unable to return with concessions
	bar fight:
		would be likely to get caught after violent dispute
		but if actor irrational might not risk an encounter
		puts actor in better state
	goya's fighting witches:
		witches are fighting over who is the bigger witch
		can get witches together if shared interest
		but will never have trust / friendship (because evil)

source of commitment:
	third-party observation:
		third party observes interaction between promisor/promisee
		higher/lower reliance investments, choose keep/break promise
		third party can then decide to pay to punish promisor
		punishment higher when promise was given
		and when reliance investments were higher
	delegation with low-powered incentives:
		like manager of company with flat salary
		then might feel obligated by own ethics to keep promise
		hence contracts with such delegates could be much simpler
		as you would not expect this party to defect
		observed in civil service, family owned enterprises
		but then need to solve moral hazard problem differently

trust game (charness):
	like farmer (trustor) vs viper (trustee)
	truster puts itself at mercy of other
	model:
		trustor can opt out (3/3) else can trust
		if trust honoured (10/10), else (14, 0)
		=> nash solution would be expected
	observation:
		25% cooperation in simple case
		50% cooperation with enabling communication
	interpretation:
		communication enabled cooperation
		even if only cheap talk (non-enforceable)
		reputation impact zero as one-shot game

promises study:
	keep word to not disappoint promisees expectation
	set expectation game:
		seller/buyer learn deliver has to be over device
		promise made that deliver will be made
		then both learn if delivery with high/low probability possible
		then seller can decide whether to deliver the goods
	observations:
		reliable device motived higher performance

loss frame:
	bridge:
		goal of suspension was unreachable
		needed to pay damages depending on goal miss		
	factory workers:
		some workers got payment + deductions
		others got base payment + bonus
		loss frame vs win frame
		workers of loss frame performed better
	contract variants:
		vanilla (payed by element)
		threshold with bonus / malus (target given)
		best performing if threshold realistic
		worse than vanilla if threshold aspirational (unrealistic)
		low variance with low threshold (nice for safety / coordination)
	realistic vs aspirational goals:
		if some goals unreachable
		then spillover effect to other goals observable
		like UNO climate commits to 2°, while 1.5° would be great
		like sustainable development goals too ambitious