Economics of organisations - theory

what are the 5 theories of why firms exist?

• transaction costs

• property rights theory

• incentives

• adaption

• teams

why do we bundle steps of production inside a firm instead of having separate suppliers for each step?

• because price/contracts are costly

• there are transaction costs

  • firms centralise the tasks and directs production, eliminating the transaction costs

  • ie employees don’t make contracts with one another, the manager will decide on any dispute between employees

examples of transaction costs found in markets

• search and price discovery

• set contracts between individuals

• enforcing contracts

• renegotiate/resolving disputes

why do contracts break down?

• according to Williamson (1975):

  • bounded rationality and uncertainty

    • can’t foresee all contingencies or specify all actions ex-ante

      • ie we can’t predict every single action that could happen and write it into the contract

    • contracts are there incomplete

  • opportunism

    • parties exploit contract gaps: shading, renegotiation threats, misrepresentation

    • asset specificity creates hold-up problems

• overall, costly ex-post bargaining and safeguards are unavoidable

what does hierarchy (present in firms) do well compared to the market?

• adaption to shocks (no full renegotiation)

• authority for dispute resolution

• coordination across interdependent tasks

what does hierarchy (present in firms) not do well compared to the market?

• weaker incentives (‘low-powered’)

  • firms don’t replicate the same changes as in the market

  • ie if the price of a good goes down, the firm will decrease profits but won’t decrease the wage to the workers

    • this shows that rich information is not so easily absorbed compared to the market

• monitoring/communication limits due to bureaucracy present

• loss of market discipline/competitiveness

  • workers on the same task won’t act competitively in the same firm as they would in the market setting

example of appropriable quasi-rent

• a bicycle company was developing a new model which required special wheels,

• the wheels cannot be produced by wheel producers who don’t have a special machine that is specific to the wheel design

• the wheels cannot be sold to any other producer

what is the subgame perfect equilibrium?

• using backward induction, the wheel producer prefers to accept the deal (-20) to selling to another bike producer (-20)

• the bike producer prefers to reduce the price (150) to keeping the price (100)

• the wheel producer prefers to not invest (0) to accepting the deal (-10)

  • therefore, the subgame perfect equilibrium is not invest

• therefore, the hold-up problem results in the most efficient outcomes (100,100) not being achieved due to pareto efficiency

are there solutions to solve the ‘hold-up’ problem?

• could write a contract to avoid not achieving the most efficient outcome

  • there’s always ways-out/loopholes

  • ie claim quality was not good

  • ways out may be desirable, especially if the asset doesn’t sell

  • producer could claim ‘unexpected’ costs and raise the price

• integrating the transactions into a single firm would reduce the rent-seeking opportunistic behaviour

key idea of transaction cost theory

• make or buy

  • integrate when transaction hazards dominate, ie high appropriable quasi rents, uncertainty, etc

  • outsource to market when bureaucratic/monitoring costs dominate

• all issues are ‘ex-post’ transaction

  • problems due to rent-seeking after transaction

property rights theory

• developed by Grossman, Hart, and Moore

• parties combine assets to create value

  • as ownership gives right to exclude/redeploy the assets, cooperation requires compensating owners enough to participate

  • ownership determines how the transaction’s value is allocated

• problems arise from ex-ante decisions

  • payoffs not affected by ex-post issues, only by ownership (ie ex-ante non-contractible investments)

  • this is because investment is needed before a transaction can take place such as training which is not contractable

example of property rights theory

• input producer owns asset 1 and final producer owns asset 2

• asset produces V1 and V2 separately but V >/= V1 + V2 when combined

• party i can invest Ii (eg training) that increases the payoff but not the asset price

  • ie payoff is higher if ownership is higher (similar to stocks)

• they can’t write a contract setting the level of investment

• each party i gets what is in the picture

• the share of AQR is fixed ex-ante (ie there are no opportunistic issues)

  • AQR here represents the surplus of the transaction generated

why does the free rider problem occur within the property rights theory?

• efficient bargaining results in parties sharing the surplus from their specific investments

• party i bears the full cost of investment but receives only a share of the surplus

• this results in suboptimal investment due to free riding

• however, if party i has a higher level of ownership (eg higher Vi), this increases the party’s incentives to invest

key takeaways from the property rights theory

• if V = V1 + V2, there are no incentives to integrate

• if I has a large impact on V, then i should own more assets

• if V1 + V2 = 0 but V > 0, the assets should have a single owner

incentives theory of firms

• firms exist to solve incentive issues (eg moral hazard)

  • some tasks/transactions may be too costly to integrate

  • firms implement pay systems, task allocation, technology, etc to mitigate these issues

  • availability of these tools determine firm boundaries

• integration is only optimal if there is good technology to monitor effort

• two ways to structure the problem:

  • if the agent owns the asset, they receive a payment based on measured performance and the asset’s value after production occurs (two incentives)

  • if the agent does not own the asset, the incentives come from being paid on performance

adaptation theory of firms

• collective decisions are costly and slow

• delegate authority into a ‘boss’ who can make fast decision

  • potential problem: bosses may have different objectives to the firm

team theory of firms

• teamwork requires supervisors that prevent free riding

  • who control the supervisors?

  • need a whole hierarchy just to solve the free riding problem

ownership trade-off in trucking industry (Baker and Hubbard, 2004)

driver

• good: incentives to drive carefully which preserve truck’s value

• bad: discretion to take external jobs or reject trips

  • carrier has less incentives to find good ‘backhauls’

carrier company

• good: incentives to find backhauls

  • force drivers to not accept external trips and to accept profitable trips

• bad: drivers may not drive carefully, drive fast to have more breaks, blame traffic for delays, etc

how is the trade-off in ownership related to theory? (Baker and Hubbard, 2004)

• asset ownership gives different incentives to invest

  • eg preserving the truck vs coordination efforts, haggling, etc

• driver’s actions are hard to monitor

  • carriers can’t enforce contracts with careful driving, how to drive, etc

  • carriers can’t specify what to do in every single scenario

how does the trade-off in asset ownership depend on trip/cargo? (Baker and Hubbard, 2004)

• long hauls: more likely to drive fast, make up time, etc

  • more valuable to have a backhaul or else it is very costly

• specialized trailers: backhaul is unusual

  • lowers potential for haggling or bargaining on trips

  • eg would not expect chilled goods and fuel to have backhaul

what do Baker and Hubbard (2004) study?

• the study the trade-offs between asset ownership in the trucking industry

• carriers work with a mix of drivers; some own their trucks and others drive company’s owned trucks

• study driver vs carrier ownership in two ways:

  • how it differs for different trips/cargos

  • what happened after the introduction of a technology (on board computers) that allowed measuring driver’s behaviour accurately

• data is not a panel of trucks but instead a repeated cross-section

  • define: cohort = state*product*trailer*distance

    • eg long-distance hauls of food in refrigerated vans based in California

    • each cohort observed before (1987) and after (1992) OBC introduction

      • minimises within-segment hetereogeneity in haul characteristic that would otherwise tend to bias estimates

what does this table of results show us? (Baker and Hubbard, 2004)

• the share of trucks owned by employees increase as distance increases

  • consistent with the hypothesis that carrier-owned drivers have less of an incentive to drive carefully

what does this table of results show us? (Baker and Hubbard, 2004)

• the share of owner-owned trucks is lower for backhaul trailers

  • consistent with owner-drivers’ rent seeking behaviour (eg reject trip, bargain commissions)

  • this means carriers have low incentives to secure backhauls for owner-drivers

predictions about on board computers (Baker and Hubbard, 2004)

• more adoption for long hauls and backhauls trailers

  • most commonly carried out by carrier-owned drivers who are more likely to speed, not drive carefully etc

  • on board computers makes it easier to monitor drivers and enforce careful driving with contractual clauses

• adoption of on board computer should increase carrier-ownership

  • this is because there is a higher benefit to the carrier company in having the on board computers compared to employees who own their trucks

• carrier-drivers should change their driving behaviour more than owner-drivers

what do these results show us? (Baker and Hubbard, 2004)

• company-owned trucks adopted more on board computers

  • this is because the on board computers has a higher benefit for the carrier company

• adoption is higher for longer distances

what do these results show us? (Baker and Hubbard, 2004)

• owner drivers are more likely to adopt the on board computers with messaging capability (eg EVMS) as they benefit from receiving traffic updates

  • benefit from on board computers for coordination rather than for solving ‘careless driving’ issues

  • still lower adoption rates compared to company owned trucks

what does this equation show us? (Baker and Hubbard, 2004)

• S_r = ratio of carrier vs driver owned trucks in cohort r (triangle = change)

  • the higher the ratio, the more carrier-owned trucks in that cohort compared to driver-owned trucks

• OBC_r = share of cohort r that adopted OBC (between 1987 and 1992)

• X_r = some characteristics of the cohort

• this equation helps us to answer whether the adoption of the computer machines changes the structure of ownership

what do these results show us? (Baker and Hubbard, 2004)

• on board computers adoption increases carrier ownership

  • it solves the issue of carriers previously not being able to monitor drivers

  • solves the main incentive issue

• however, this effect is only significant for long trips

  • this is because the incentives for driving carefully are more problematic for long hauls

what do these results show? (Baker and Hubbard, 2004)

• trip recorders encouraged carrier-drivers to drive more efficiently

  • ie to consume less fuel

  • fuel consumption is used as a proxy for careful driving as if the driver speeds and breaks harshly, they will use up a lot of fuel

  • no effect on owner drivers as they already have the incentive to drive carefully

• EVMS (on board messaging computer) improved fuel consumption for both employee-owned and carrier-owned drivers

  • better ‘coordination’ can help with fuel consumption (eg carrier can tell you to avoid a busy route)

key messages from Baker and Hubbard (2004)

• ownership of assets affects incentives of parties

  • depends on the ability to measure and monitor the actions of parties

• can be different for different tasks (eg long vs short trips)

• new technologies can affect this trade-off and induce changes in ownership and firm’s structure

background into Hansman et al (2020)

• setting: Peruvian fishmeal industry

• many independent fishing boats supplying processing plants

• key friction: the quality of fishmeal (ie the protein content) is hard to contract on

  • protein content depends on how long fish has been out of water

    • a matter of hours can make huge differences

  • only observed ex post, making it difficult to trace quality back to individual boats

• questions:

• does quality premium increase integration?

  • integration = boats are owned by processing firm

  • quality premium = huge price differentials between high and low protein content fish

• do integrated suppliers prioritize quality?

• data: observe prices and production of each processing plant and over time

what do these results show us? (Hansman et al, 2020)

• we are regressing integration (share of fish bought from owned suppliers) on quality premium

  • comparing the same plant when the quality premium is low and high

• export markets pay large premia for high quality

• when quality becomes more valuable, firms are more likely to vertically integrate fishing boats

  • market contracts fail to induce quality

  • integration allow direct control of boat behaviour

    • have GPS to track boats to tell them to return immediately once caught fish

• link to theory: firm boundaries shift when markets cannot provide the right incentives

what do these results show us? (Hansman et al, 2020)

• we are comparing the same boats before and after integration

• including plant and year fixed effects

• the higher the share of inputs, the higher the quality of fish

• after integration, the same boat will deliver less fish but higher quality fish

  • under market contracts, suppliers will focus on quantity as quality cannot be observed so cannot have contracts on quality

  • quality therefore can only be achieved through integration

when does integration become more important?

• when contract imperfections are high

what happens when appropriable quasi rents are large?

• it makes integration more likely

  • AQRs makes haggling more likely/costly as it provides a source of private monetary gain

assumptions of rent-seeking/transaction costs theory of firms that are not clear

• assumes that integration can stop the haggling induced by the AQRs but it requires an implicit focus on certain types of haggling

  • eg if haggling was completed through the manipulation of alienable capital, then integration could remove the control rights form the haggler

  • if haggling was completed by manipulation of inalienable capital, then integration could not stop rent-seeking

• suggests that ownership could stop haggling that is undertaken via alienable assets

reasons why hold-ups between firms may not result in integration (according to rent-seeking/transaction costs theory)

• the hold-ups use inalienable instruments so the observed hold-ups are unavoidable

  • integration would do nothing to solve it

• the observed hold-ups are actually better than what integration would have provided

comparison of property rights theory and rent seeking/transaction costs theory

• rent-seeking theory predicts socially destructive haggling

  • property rights theory assumes efficient bargaining

• rent seeking is consistent with contractible specific investments ex ante

  • property rights requires non-contractible specific investments

• in rent seeking theory, the integration decision determines ex post haggling and hence total surplus

  • in the property rights theory, the integration decision determines ex ante investment and hence total surplus

• rent seeking theory does not say anything about internal organisation

  • property rights theory defines and evaluates life under integration and non-integration to give benefits and costs of integration

why does owning more assets result in stronger investment incentives, according to the property rights thoeyr?

• efficient bargaining results in parties sharing the surplus from their specific investments

• surplus share is determined by each party’s investment incentive

• each party’s asset ownership determines the party’s surplus share

  • therefore, owning more assets guarantees a bigger surplus share and creates a stronger investment incentive

  • if it is more important to maximise one party’s investment, then that party should own all of the asset

  • if all the parties’ incentives are important, then dividing the asset between parties is efficient

downside of integration

• after integration, if the producer can now hold up the supplier, this may reduce the supplier’s incentive to invest

  • reduced initiative for the acquired firm

how does the incentive theory differ from the property rights and transaction costs theories?

• incentive theory: focuses on an incentive problem between a principal and agent

• transaction cost/property rights theory: focuses on make-or-buy problem

what is the optimal incentive contract?

• depends on ownership structure

  • if the agent owns the asset, they have the incentive to increase the asset’s value

  • the optimal contract for an agent who does not own the asset (employee) provides weaker incentives than does the optimal contract for an agent who does own the asset (contractor)

why should we care about vertical integration?

• frictions: transaction costs, information asymmetries

• overcomes the problem of imperfect contracts

  • easier for firms to acquire the inputs themselves rather than going to the market to avoid the market frictions present

  • firms vertically integrate to ensure the quality of their inputs

3 main hypotheses of Hansman et al (2020)

• should see a correlation between the proportion of inputs coming from a vertically integrated supplier and the quality of output

  • when the incentives of the market are larger (ie when the quality premium is larger), firms should source more of their fish from vertically integrated suppliers

    • ie they should be reacting to prices

  • vertical integration increases when quality becomes more important

• vertically integrated suppliers take more actions to ensure high quality inputs

• when firms source more from a vertically-integrated supplier, this has a causal effect on quality of output

how does diseconomies of scale arise?

• arises not from variations in scale but from the inability to vary all factors

• eg management

  • it is fixed and cnanot be replicated

advantages of spot markets

• efficient adaptation to changes in demand and cost

  • equilibrium prices and quantities adjust to reflect changes in demand and cost and realize maximum total surplus

• cost minimization

  • a firm will invest in cost reduction until the marginal benefits of cost reduction equals the marginal cost

• realisation of economies of scale

• these advantages arise as there is not a relationship between a firm and its suppliers, meaning firms can substitute away to a cheaper supplier

what do the terms in the regression represent?

• beta = the effect that X has on Y

  • causal effect

• u = error term

  • captures everything that is not included in x

what does the hat above parameters mean?

• means they are ‘estimated’ from the data

what does the bar above parameters mean?

• average

OLS formula (with one regressor)

• if we divide the numerator in the first equation by N, it becomes the covariance

• if we divide the denominator in the first equation by N, it becomes the variance of x

• therefore, beta measures the correlation between y and x

  • this does NOT imply causality though

do we really need the ‘true’ model to be linear?

• no

• for any two variables (y and x), we can write what is in the picture

• u = a residual uncorrelated with x

• E(y I x) = conditional expectation function which captures the ‘average’ relation between y and x

  • tells us how x affects y on average

  • can take any form, including non-linear

  • looking at what the average of y is when x is held constant and we do this for several values of x to create a best fitted line

even if the true relation is not linear, can we use OLS?

• yes

• this is because we can show that the best linear approximation to the conditional expectation function is a line with slope cov (y, x) / var(x)

  • this is just OLS !!

reasons why beta will not measure the ‘true’ causal beta

• omitted variable bias / selection

• measurement error

• reverse causation

what is the omitted variable in this?

• temperature/ weather

OVB formula

• the problem is the beta picks part of the effect of z on y

• there is no OVB if gamma = 0 (ie x and z are uncorrelated)

what could be the omitted variable in this?

• the type of organisation

  • ie more horizontal firms may be able to implement working from home but there may be less promotional opportunities

what is the omitted variable when we we regress profits on CEO pay?

• CEO experience

• we would expect high experienced CEOs are paid more

• we would also expect experienced CEOs increase profits

how would we solve this problem where there are two identifiable groups with different outcome values?

• solve this by controlling for the omitted variable

  • ie have a dummy variable to differentiate the two groups

  • if we don’t do this, we would conclude that the effect of X on Y is much higher than the true one

• results in an extra constant

  • results in two different intercepts

notes about OVB

• in a multiple regression, beta is interpreted as the effect of X on Y conditional on other controls

  • can be interpreted as the effect of X on Y after netting out the effect of other controls

• need to control for a variable if it affects Y and is correlated with the variable of interest (ie X)

• can’t interpret the correlation as causal if there is OVB

beta formula with measurement error

• don’t observe x but observe a noisy version of it

  • noise is uncorrelated with any variable

• the higher the noise (ie the higher variance of the noise), beta is biased towards to zero

  • ie it observes attenuation bias

  • can’t fix it but if it is very small, we can ‘ignore’ it

• if the attenuation bias is high, then the regression line is flatter

what is the problem with this graph?

• looks like when there are more police, crime is higher so the police are not good at their job

  • they are actually capturing when crime is higher, more police will patrol that area

• OLS cannot distinguish in which direction the relation goes

  • there could be a causal effect but OLS will pick the total effect so we may overestimate the effect we are wanting to study

  • this happens when the reverse relationship is very large

Which theory of the firm is consistent with this setting?

• the property rights theory

• claims that the issues determining the boundaries of the firm (which transaction should be internalized) are due to ex-ante investments that parties can make to increase the value of the transaction

• due to contracting problems (and the fact that asset ownership confers the right to decide on the asset’s use), the parties can’t commit to make these investments

• in this context, integrating transactions (eg a common owner of assets) can be efficient

What is the optimal level of investment if producers could commit to it with a contract?

Discuss why the owner will not hire the chef if they have to pay a fixed wage

• as effort is not observable (thus no penalties can be imposed on it), the chef has no incentives to put any effort

• on average, R = 1, while the paid wage has to be at least equal to 1.5 (so the chef would accept it as this is equal to the outside option)

Based on the example of the semi-skimmed milk vs yoghurt production, explain why incomplete contracts are problematic (Hansman et al, 2020)

• when producing something very standard as semi-skimmed milk, different parts of the production process can be done by separate firms

  • eg the farmers and the milk processing

• it is easy to specify in a contract what are the requirements of the transacted products and related terms

• when you need to produce something much more specific such as a certain type of yoghurt, it is more difficult to establish the requirements of the products into a contract

  • particularly, things related to quality that is difficult to observe and measure (in a short period) become difficult to contract

• we expect more integration when unobserved quality of the inputs are important for the final value of the product

  • what we see in Hansman et al (2020)

why is it difficult to monitor the behaviour of the boats in order to specify in a contract the way in which they should manage catchment? (Hansman et al, 2020)

• although boats have GPS, this is not observed by the firms

  • observed only by regulators

• also it is not obvious there is a unique strategy for fishing efficiently

  • eg it may be better sometimes not to do short trips but to stay in a sea area where you would find a lot of fish

• thus, a contract about how the boat should behave would have to be contingent on a large number of sea/weather/etc conditions

how does the fact that quality is unobserved affect the decision to integrate? (Hansman et al, 2020)

• by integrating vertically (the input and final good producer), they can avoid many transaction costs related to problems with quality checking

• as the firm producing the input (responsible for making effort to keep quality standards) and the output producers are both affected by the quality of the inputs, there is no need to rely on complicated contracts to set quality standards

why is it possible (particularly in the context where the study takes place) that problems preventing quality upgrading with market transactions are not solved even if parties can write a very detailed market with all the necessary specifications? (Hansman et al, 2020)

• even if quality is easily observable and can be specify into contracts, these problems may not be solved

• main reasons:

  • the poor institutional quality (eg non-efficient judicial system)

  • high cost of enforcing the contact (eg if your input provider breaks the contract, you need to hire lawyers, go to a court, wait years before court decision, etc)

(Hansman et al, 2020)

• if we switch VI from 0 to 1, the share of high-quality production will increase by 0.311 ie by 31.1 percentage points

• can scale the effect to say that a 10-percentage-point increase in vertical integration raises the share of high-quality output by about 3.1 p.p

(Hansman et al, 2020)

• VI = dummy (important to note for log-level regressions)

• ignoring the approximation error, this implies that increasing vertical integration from 0 to 1 (ie the boat becomes fully integrated) raises the number of trips by 19%

• if we account for the approximation error, this corresponds to an increase of exp(0.19) - 1 = 0.21 ie a 21% increase in the number of trips made by the boat

This regression includes boat (ie supplier) fixed effects. Explain intuitively what the estimated coefficient (beta) captures. What comparison does the regression implicitly make? (Hansman et al, 2020)

• boat fixed effects imply that we are comparing changes within boats

• ie we are looking at the number of trips of the same boat, before vs after integration (and then, we get an average of the effect across all boats)

• if a boat is always integrated or always not integrated, it does not contribute to the estimation of beta

Consider the baseline regression used by Baker and Hubbard (2004) in the paper about the trucking industry (specifically the effect of OBC adoption on ownership). Discuss what are the main threats to interpreting the estimation as causal?

• omitted variables, eg the demand for a particular type of produce at some locations may have increased

  • carrier companies may have the incentive to buy more trucks to exploit these potential profits

  • at the same time, they may use these higher profits to invest in new technology which includes OBC

    • the effect on ownership is not driven by the OBC but the higher profits potential affected both ownership and OBC adoption simultaneously

  • reverse causality, ie the fact that carrier-ownership expansion may affect the OBC adoption

    • OBC producers may track, advertise and send their sale agents to talk to carrier companies that have just expanded

  • measurement error

    • could happen when an individual lies in a survey or variables are not measured correctly

    • results in attenuation bias (results are biased towards zero)

Why is Baker and Hubbard (2004)’s specification better than a single cross-section using a single survey year?

• their specification is first differences (similar to fixed effects)

• first differences means it is controlling for any unobservable time-invariant differences between the different cohorts

  • eg the price of OBC may differ across states and in states where OBC are cheaper or where carrier companies are bigger so they can buy more of them

  • any time-invariant characteristic that makes ownership different across states, trailer type, or distances that is also correlated with OBC adoption will show up as OVB in a cross-section regression

Explain the IV strategy Baker and Hubbard (2004) use and discuss the conditions to solve the causality problem

• instead of using a single IV, they use 19 dummies indicating the transported product as instruments

• intuition behind instruments: they only use the variation in the OBC adoption coming from the fact that OBC is more valuable for some products (independently of the ownership)

  • ie we would expect OBC to be adopted more for truck hauls with dangerous cargo such as petrol or chemicals

• relevance: the dummies are not jointly insignificant

• exogeneity: products shouldn’t influence the shift from using an owner-operator to a company driver

  • however, different products could also be associated to changes in incentives to adopt one type of ownership, ie drivers may be less willing to use their truck for dangerous cargo

Discuss the advantages and disadvantages of having the sales agent vs the insurance company owning the client network

• advantage of sales agent:

  • has the incentive to work in order to get commission

• advantage of internal agent:

  • easier to monitor as they can bypass contracting issues such as the hold-up problem

• disadvantage of sales agent:

  • she may sell products from other company, suggesting to invest in packages that are not best for the company

  • hard to contract them to do exactly what the company wants them to do

• disadvantage of internal agent:

  • wage is fixed so they don’t have the incentives to make much effort in keeping clients happy