Big Data Midterm Flashcards
(77 cards)
1
Q
Producer theory
A
Producer wants to maximize their profits:
- marginal cost at q is the cost of producing the qth unit.
- Go on producing as long as you make marginal profits: q* is optimal if mr(q) = mc(q)
2
Q
perfect competition
A
each producer has no effect on market value, price is given
3
Q
monopoly
A
there is a unique producer, which determines prices and quantity on the market
4
Q
Welfare analysis
A
demand can be linked to consumers utility:
- a consumer willingness to pay (WTP) is the maximum price at which the consumer would still buy the product
- for every product quantity, demand gives you the willingness to pay of consumers for the marginal unit of product
5
Q
Consumer surplus indicates:
A
How much the consumers benefit from the existence of the market for this product:
- think of it as profits, but for the consumers
- the surplus generated by each unit is value of this unit (WTP) - price paid
6
Q
The total surplus is defined as the surplus of all economic agents
A
- the producer surplus is simply the profits generated by producers
- the total surplus is CS + PS (+ government revenues if they exist)
7
Q
Total surplus as an indicator of social welfare
A
- total surplus does not say anything about distribution, inequalities, fairness
- total surplus is only an indicator of (Pareto) efficiency
- an outcome is Pareto efficient if nobody can be made better off without making another agent worse off
8
Q
Game theory
A
- consumer and producer theory study constrained optimization, game theory studies strategic interactions
- game is any situation in which agents (players) interact in a strategic manner. Game is associated by possible actions and corresponding payoffs
9
Q
Bayesian game
A
a game with incomplete information. Some of the players are uncertain about the other players’ type. Type could be their payoffs, or some other private information
10
Q
Bertrand competition
A
game theory were firms compete over prices
11
Q
Few examples of behavioral traits and biases
A
- non-bayesian updating: agents’ beliefs do not follow the probalistic rules and properties
- framing effect: choices vary with the way the outcome is presented, while the outcome is the same
- present bias: in intertemporal choices, present is overweighted with respect to all future periods
- inequality aversion: agents act in a way only consistent with a preference for overall fairness
12
Q
The right to privacy and right to data protection
A
The right to privacy is a human right according to the delaration of human rights by the UN and the European convention on human rights.
However, the right to privacy is not the right to data protection!
13
Q
Difference between right to data protection as a human right between EU and UN
A
- European council recognizes right to data protection as human right (since 1981): “everyone has right to protection of personal data. Such data must be processed fairly, for specified purposes and on bass of consent. Compliance to these rules shall ne subject to control by an independent authority.”
- UN does not recognize this as a human right. 2013 resolution on the right to privacy in the digital age, is not binding
14
Q
Right to data protection often interacts with other rights:
A
- right to data protection can interfere with other rights, but also ensure its respect. For example: freedom of expression
- when there are tension between different rights, the courts’ role is to balance them
- it is the duty of the member states to reconcile with freedom of expression and information. For example: public access to official documents and obligation of professional secrecy
15
Q
GDPR
A
General Data Protection Regulation is a single set of data protection rules to be applied by all EU members.
- it applies to all organizations that process data of anyone in EU territory, both public and private sectors, for profuts and non profits.
- it empowers EU members by allowing sanctions and fines
16
Q
Personal data
A
any data relating to an individual (data subject) who can be identified, directly or indirectly
- anonymized are not considered personal data: pseudonymized data re.
- some personal data are sensitive: race, religion, biometrics, sexual orrientation
17
Q
Data processing
A
any operation performed on personal data automated or manual
- the data controller decides why and how personal data will be processed
- the data processor processes personal data on behalf of a data controller
18
Q
7 principles of GDPR
A
- lawfulness fairness and transparency: processing must be lawful, fair and transparent to the data subject
- purpose limitation: data must be processed for the legitimate purposed specified explicitly to the data subject when collected
- data minimization: data should be collected and processed only as much data as absolutely necessary for the purposes specified
- accuracy: personal data must be accurate and up to date
- storage limitation: personally identifying data may be stored only for as long as necessary for the specified purpose
- integrity and confidentiality: processing must be done in such a way as to ensure appropriate securuty, intergrity and confidentiality
- accountability: controllers and processors must actively and continuously implement measures to ensure compliance
19
Q
GDPR subject data’s rights: the data subject have the following 8 rights
A
- the right to be informed
- the right to access
- the right to rectification
- the right to erasure
- the right to restrict processing
- the right to data portability
- the right to object
- rights against automated decision and profiling
20
Q
In case of violation of data subjects’ rights
A
- data subjects have the right to receive compensation
- the controller or processor is liable for any material and non-material damage
- administrative fines for infringements can be extremely high
21
Q
EU data strategy
A
Create a single market for data where:
- data can flow within the EU and across sectors, for the benefit of all
- European rules - in particular privacy, data protection and competition law - are fully respected
- the rules for access and use of data are fair, practical and clear
22
Q
Single market for data, part of EU data strategy, is composed of many different regulations, but in particular:
A
- digital services act
- digital markets act
- data governance act
- data act
23
Q
Digital services package is composed of
A
- the digital services act (DSA)
- the digital markets act (DMA)
24
Q
Goals of the digital services package that is composed of DSA and DMA
A
- create a safer digital space in which digital services users’ rights are protected
- establish a level playing field to foster innovation, growth and competitiveness
25
Difference between how digital services act (DSA) and digital markets act (DMA) concern online platform services
- DSA applies to most online intermediaries and platforms (social networks, app stores)
- DMA only applies to so called gatekeepers (Facebook, Youtube, Apple store)
26
DSA regulates olbigations of digital services pro that act as intermediaries
- measures surrounding moderation, illegal content and untrustworthy sellers
- transparency on recommender systems and restricition on deceptive designs
- transparency on advertising and ban on targeted advertising using sensitive data
- enhancements of user rights
- for VLOPs and VLOSEs, special diligence regarding systemic risk (Very Large Online Platforms and Search Engines > 45 mln users in EU)
27
DMA target so called gatekeepers by:
- large digital platforms that are important gateways for business and end users
- size thresholds to establish rebuttable presumptions under 3 criteria: 1) impact on internal market (EU turnover > €7.5 bln a year), 2) gateway to reach end users: EU end-users > 45 mln + active EU businesses > 10K and 3) entrenched, durable position: thresholds met in the last 3 years
- the DMA imposes some obligations towards both end users and businesses
- the DMA also prohibits commonplace platform practices (for example: parity clauses, usage restrictions and using business data for competing)
28
data governance act: goal of DGA is to foster data availability by:
- regulating the re-use of publicly held, protected data
- boosting data sharing through the regulation of data intermediaries
- encouraging the sharing of data for altruistic purposes
This concerns both personal and non-personal data
29
Data act
- is concerned by the so called internet of things
- applies to both personal and non-personal data
- aims at creating a fairer digital environment for all sectors in the economy
- specifies rules on who can access what data for which purposes
30
Data act includes measures to
- allow users to access to data they generate and to share such data with 3rd parties
- shield SMEs from unfair contractual terms imposed by "big players"
- allow public sector to access and use publicly held data in exceptional circumstances
31
Convention 108 is the first legally binding international instrument in the data protection field
- opened by signature on 28-01-1981
- propsed by the council of Europe
- revised and modernized in 2018
- very similar to the GDPR: based ont the same ideas and principles and accessible to EU members and non EU members
- 55 member states
32
Cost of withdrawing information: Sitgler (1980) and Posner (1981) argue that privacy leads to economic inefficiency:
- they define privacy in the sense of concealment of information
- concealment of information leads to mismatches, misallocations and fraud
Concerns for privacy can be seen as rules for "selling" oneself
- this summetric to selling a product so the same rules should apply
33
Lack of information will have to be compensated for by proxy information because of inefficient means
- collecting such proxy information is more costly, for example: if employment history is private, need for a longer probbation period under high scrunity
- collecting such proxy information is precise, for example: if criminal records are private, frequency of job change or long unemployment spree can be a clue
- collecting such proxy information prevents surplus, for example: if credit scores are private, lower credit limits with higher interests are initially enforced
34
The value of signals: Chicago school underlines the inefficiencies linked to assymetric information
- implicity, privacy makes observable types unobservable
- however, even without privacy, types might be unobservable
35
Privacy can be seen as preventing signals that inform about types to be transmitted, what is the value of such signals?
- value of correcting asymmetic information is ambiguous: it can be Pareto inefficient and reduce total surplus
- we consider two familiar framework to illustrate this: 1) signaling in a setup à la Spene, and 2) certifying in a setup à la Akerlof
36
Private v.s. social value of information: Hirshleifer (1971)
Shows how private value of information can exceed social value of it
- in a pure exchange economy, information is not useful for production decision
- information isn valuable to individuals because they use it against others
- information becomes worthless to everyone as soon as it is public
37
Example that shows how privat value information can exceed social value of it
Imagine a bet with someone, e.g. bet on head or tails
- privately: with opponent uninformed, knowing the outcome is very valuable
- socially: one will lose, one will win: information is worthless
38
Value of information in assymetric information games is ambiguous
- what is the cost of withdrawing information? -> it might cause inefficiencies on the market through misallocation and misbehaviour
- what is the value of signals? -> it depends, are signals allowing to correct inefficienct or only redistribute surplus?
- how does individual value of information differ from its social value? -> it generally differ: private value might be lower or higher than social value
39
Price discrimination has become particularly relevant with internet and big data because:
- big data allows to estimate individual valuation even more precisely
- personalized prices might be easier to implement online
- personalized prices can be automated through AI
40
What can you do to hide for online price discrimination
- erasing and/or blocking cookies
- multiple email adresses
- use of VPN
- temproary virtual credit cards
41
Why would a hiding technology make consumers worse off?
Consumers WTP might be unaccessible to the monopolist, either:
- because you hide, i.e. your valuation is high
- because by chance you were not tracked, i.e. your valuation is low
42
Is price discrimination desirable?
- the consumers are all better off without price discrimination
- the monopolist is better of with price discrimination
- So producer surplus is greater with price discrimination, but consumer surplus is greater without
- total surplus is usually greater with price discriminatio because it eliminates deadweight loss
43
strategic customers react to price discrimination by:
- trying to hide their willingness to pay
- being strategical about timing of purchase
44
What can you do to hide for price discrimination?
- Earsing / blocking cookies
- multiple email addresses
- VPN's
- temporary virtual credit cards
45
Identity management strategies to hide for price discrimination are typically costly becuase:
they require time, effort and money
46
technology does not necessarily benefit the consumers, we show this through a hidden cost of hiding:
- considers price disciminiation by monopolist who can sometimes track consumers
- show how a hiding technology can make consumers collectively worse off
- relies on the idea of strategic externalities
47
why would a hiding technology make consumers worse off?
- because you hide, i.e. your valuation is high
- because by chance you were not tracked, i.e. your valuation is low
48
price discrimination induces different consumers reactions when there is competition
- hiding technology can soften competition, benefit firms and lower CS
- firms might want to use dynamic targeted pricing (with repeated purchase)
- effects of voluntary disclosure might depend on the disclosure technology
- effect of competition on price discrimination has important consequences
49
strategic consumers can be hurt by privacy regulations
- in particular if regulation makes hiding easier
- hiding has strategic externalities that hurt consumer surplus
- however, naïve consumers can be better off with regulations
50
how can firms with market power benefit from committing to privacy policies
- in that case price discrimination is not used
- privacy regulation becomes, at best useless, at worst hurtful
- however, competition can make tracking more attractive
51
with repeated purchase, instead of a one shot interaction between monopolist and consumers, the possibility to hide can also have detrimental effects for consumers:
- with repeated purchase, firm would rather commit not to price discriminate
- without commitment, the firm needs to offer a lower introductory price in period 1
- when hiding is available and free, the firm is better off and consumers worse off
- when hiding is available and costly, consumers are better off
52
posibiltiy of product personalization can weaken the incentive for a monopolist to price discriminate
- a monopolist might prefer extracting surplus from higher uniform price of personalized products than from price discrimination
- either the monopolist knows type perfectly, how to personalize product and the valuation for it, or does not know anything
- by hiding, consumers forego their preferred (personalized) product in order to avoid price discrimination
- by using tracking technology, the monopolsit gets to extract a higher surplus from consumers that are not hidden and get detected
53
satistical discrimination
- due to the characteristics being informative about something unobservable
- this results from a statistical correlation between observable and unobservable characteristics
54
taste based discrimination
- due to the discriminating agent not liking the characteristics
- this is arbitrary (and inefficient). more importantly, it is also an ethical issue
55
privacy regulations have ambiguous effects on the prevalence of discrimination
- can help if taste based discrimination is at play. For instance; forbidding the employer from observing religion could help against islamophobia
- can worsen prevalance if statistical discrimination is at play. For instance; forbidding the from observing criminal records could increase racial discrimination
56
Why is price discrimination less of a problem when product personalization exists
- consumers react to price discrimination by hiding, making personalization impossible
- high value from personalization makes price discrimination less attractive
57
beyond personalized prices, other types of personalization are more prevalen online:
- personalized advertisement targeting and product recommendation
- information and content
58
targeted ads can be seen as improving match between consumers and products
- increases the total social value from matches
- reduces advertisers' spending on worthless consumers (this decreasing cost should lead to decrease in price in perfect competition)
- reduces consumers' search efforts
59
Why does market segmentation arise from direct marketing?
- Overlapping territory is costly to both firms, but at least one firm does not get any benefits from it. Cannot be an equilibrium to have overlapping territory
- when the cost of marketing is high, many devisions can arise: when cost vanishes, the market segmentation reflects the relative efficiency of the firms
60
How is targeted advertising affecting consumer surplus?
+ Targeted ads can help increase value of match, increase ad efficiency and decrease search costs
- targeted ads can increase price by helping market segmentation and softening competition
61
How do consumers feel about targeted advertisement?
- substantial evidence for consumers reactance to targeted and personalized ads
- targeting and personalization creates a tension between informativeness and intrusiveness
- restricting obtrusiveness, using non sensitive criteria to target and giving control decreases perceived intrusiveness and helps avoid reactance
62
echo chambers
environment in which agents only encounter opinions that coincide with their own, so that their views are reinforced and alternatives are not considered
63
3 biases that are relevant to echo chambers
1. selection bias: agents do not account for their contacts not being representative
2. correlation neglect: agents do not account for their contacts information being correlated to theirs
3. motivated beliefs: agents update in a way that is strategic given an objective (a belief they would like to hold)
64
How can echo chambers and polariaztion lead to suboptimal information aggregation
- directly: through agent's communication between themselves
- indirectly: through the information provided to agents in the first place
65
Consequences of polarization and echo chambers on a macroeconomic level
- affects the expectation about the future: political affiliation effects expectation about economic outcome in case "opponent" is elected
- can hinder growth, especially in the long run
- can erode trust in democracy and the political system: polarization in government trust, which leads to dysfunctional government (capitol riots)
- increases the likelihood of violent conflicts: provide theoretical and empirical insights into the relationship
66
political benefits of polarization and echo chambers
- echo chambers might help communication: they can help exchanging information
67
features of social media that make behavioral biases more biting
- online replication of news: same information copied throughout various outlets
- complexity and limited observability of networks
- opaque algoritihms and AI: extend and process of selection in information is unknown to online users
68
Does Facebook increase polarization? Link between social media and polarization is difficult to establish. But studies found 4 main findings:
1. exposure to news on social emdia affects online news consumption
2. exposure to counter attitudinal news decreases affective polarization, compared to pro attitudinal news
3. no evidence that the slant of news outlets affects political opinions
4. Facebook's algorithm is less likely to supply content from counter-attitudinal outlets, conditional on subscription
69
Why are echo chambers and polarization a problem?
- they generally affect negatively information aggregation and provision
- they weaken political competition, making policies less representative
- they affect macroeconomic expectation, and the stability of the system
70
Isues with social media algorithms
- cause negative (social) externalities, e.g. through misinformation and polarization
- shape taste, e.g. through gradual push towards extreme content
- be biased, e.g. by pushing up/down content from certain groups
71
channels for impacts, focus lies on three channels
1. exposure to persuasive content: including beneficial learning
2. facilitating coordination, in particular for political mobilization
3. shifting perception, for instance by softening stigma
72
Does social media increase political participation?
- social media plays role in amplification of political/social movements (#MeToo and BLM)
- effect might depend on democracy levels
- mixed results about effect on voters' turnout
73
does social media increase offline violence?
- substantial evidence that exposure to toxic content leads to violence
- for instance anti muslim and anti asian crimes in the US, and attacks on refugees in Germany
- But, large penalties to platform seem to help mitigate this issue
74
Data brokers' incentives rarely align with social welfare considerations:
- DB only consider the profits that data can generate
- DB have different incentives than firms: selling to one firm will affect the profits of the other
- DB helps firms capture consumer's surplus
75
How do data brokers help firms capture consumer surplus?
- they facilitate first degree price discrimination
- reducing or preventing downstream competition
- allowing higher prices
76
Different type of data relatde businesses
- data driven companies: e.g. weather app
- ad-driven companies: e.g. Meta and Google
- usage driven companies: e.g. Tinder
77
When is a platform considered a gatekeeper?
1. impact on internal market = (EU turnover > €7.5 bln a year),
2. gateway to reach end users: EU end-users > 45 mln + active EU businesses > 10K
3. entrenched, durable position: thresholds met in the last 3 years
