Term Paper Suggestions: Categories
This page is continually under construction, check it again
if you're looking for a topic.
Mon Mar 23 02:21:47 EDT 2009
- Program development
- Study the steps eBay takes (if any) to discourage crawling of
their site. In other words: Exactly what does eBay do to make Assignment
- Is it true that when eBay uses anonymized user IDs like e***m
they are consistent from auction to auction---at least some of time?
- Can you identify dealers in anonymized bid histories by looking
at the ratio of total feedback points to number of bids?
- Automatically assemble a catalog of current prices in
a particular subfield (Limoge China, US state spoons...).
Easily updated of course. Compare to published
catalogs in some areas?
- Track effect of news events on prices in some areas:
people in the news and autographs, deaths and photos,
- Search automatically for bidder wars on eBay. In what categories
and in what circumstances are they more likely to occur?
- Literature survey and review of a specialized field
of auction theory or experiment. The goal here is to
determine the current state of knowledge in a fairly
narrow aspect of auction theory or practice, to
review the recent work critically, and perhaps extend or
interpret creatively. Examples:
- Combinatorial auctions and complexity theory
- Budget constraints, financial constraints
- Risk aversion
- Interdependent values.
The classic reference is P. R. Milgrom and R. J. Weber,
"Theory of auctions and competitive
bidding," Econometrica, vol. 50, no. 5, Sept. 1982,
pp. 1089-1122. (Can you derive their revenue ranking more simply
using a model more restrictive than affiliated? Perhaps correlated?)
I'm interested in this for class presentation; the definition
of "affiliated" is not very transparent.
- Common-value auctions
- Efficiency in multiunit auctions
- Auctions with discrete value distributions
- Measurement study of some aspect of online auctions.
The goal here is to learn something from downloaded data,
or data you obtain yourself in experiments. Examples:
- Study the effect of eBay's change in reputation system,
which begins May 1, 2008 (I think). This may be a unique
opportunity to study the dynamics of feedback score change
as the new system is put to use.
- Effects of special features on eBay, such as featured
item listing, highlighting, bold face, etc.
- Use of Buy-It-Now feature on eBay
- Effects of listing choices on eBay, such as time of
listing, length of auction, opening bid, hidden reserve,
reputation, shipping costs, use of PayPal, etc.
- Cross-venue comparison of like items, such as Penbid
- Cross-venue study of the effect of extending deadlines
vs. sharp deadlines
- Correlation of slabbed (US) coin prices with third-party
certification companies (for example, click
here for a fourth party's comments on the third party companies.)
Are there any other categories with third-party certifications?
- Is there a correlation between bidders who have raised
their own bids and winning?
- To what extent are auctions in certain specialized areas
and closing times becoming completely determined by snipers,
and hence in some sense equivalent to Vickrey auctions?
- Do buyers and sellers in particular subareas tend to
gather at certain auction ending times? I suspect so,
and this would be analogous to like businesses accumulating
in certain sections of cities (the diamond exchange in NYC,
for example, or the fishmarket).
- Study auctions that have developed in Massively Multiplayer
Online Role-Playing Games: what formats are used, how
are the forms suited to the games, what behavior is observed?
- Compare buy-it-now prices and non-buy-it-now prices.
Can you develop a theory of the value of time?
- Laboratory experiments
- Repeat any of the classic experiments: overbidding in
first-price IPV auctions, learning in second-price,
effect of clock speed in Dutch, etc.
- Figure out a way to distinguish between risk aversion
and spite (anyone doing fMRI?) as an explanation for
overbidding in first-price auctions. Here's an ambitious
suggestion: have the same bidders participate in three
different games. One game "measures" spite; a second
"measures" risk aversion; a third is a sealed-bid
first-price auction. Then determine whether the correlation
is better between overbidding and spite, or overbidding
and risk aversion.
- Correlate risk aversion with cultural background of subjects.
- Correlate spiteful behavior with cultural background of subjects.
(I think at least one such study has been published, one that
studies the possible effect of nationality.)
- Simulation or numerical study of equilibrium strategies
in cases too complicated to analyze. Examples:
- Simulate swoopo. Or my suggested zero-price abstraction:
"throwing good Sacawagea dollars after bad".
- Equilibria and revenue ranking in hard-to-analyze cases:
Interdependent values, combinatorial auctions, risk averse bidders,
asymmetrical bidders, spiteful bidders, etc.
- Effect of market-clearing mechanisms in repeated double-auctions
- Learning in auction-based competition
- Automated fraud detection
- Build a competitive environment with two geographic domains:
one where pairwise competition determines survival, and
the other where overall fitness does. Study through simulation
the emergence of adaptive species in this environment.
- Simulate machine learning of equiibria in auctions.
- Sociological, legal, or ethical studies of internet auctions.
Analysis of the impact of auctions in these fields. Examples:
- Ethical boundaries of behavior: collusion, shilling, sniping,
- Legal aspects of internet auctions: international considerations,
libel, copyright, general intellectual property issues, etc.
- Ethical boundaries of sale items: human organs, items objectionable
on grounds of politics, race, gender, etc.
- Study the self-policing of a subarea of eBay by looking
for news groups and their possible effects on bidding.
(For example, in coins, there's a "fakes" newsgroup.)
- Study of racial bias (see Prof. Salganik, Dept. of Sociology).
Can you think of a clever way to find a proxy for race? Gender?
- In what order should I auction off items with a distribution of values?
I believe real-life auctioneers often sell the lower-value items first.
- Study sequential auctions in general. (After first reviewing
- Given an auction mechanism with expected profit functional
EP(bidding strategy), call it transitive if, in pairwise
matches, EP(A) ≥ EP(B) and EP(B) ≥ EP(C) ⇒ EP(A) ≥
EP(C), for all bidding strategies A, B, and C. Which auctions
are transitive? Any? Do you have to restrict the class of strategies?
Start with IPV auctions and uniform distributions.
- (Thanks to D. Stavens) Study Princeton's Registrar's own
"web bidding" for limited enrollment courses.
The budget constraints. These may result in NP-complete problems
for the registrar! The general subject of what are called
"combinatorial" auctions would be interesting to study and review,
- Model machine learning in auctions.
- Study auctions with no pure equilibria. Can there be auctions with mixed
- Study the variance of the revenue in various auctions.
- Analyze the all-pay auction for risk-averse bidders:
can you prove they will they shun such an auction?
Similarly for risk-averse bidders in other autions.
- Study the susceptibility of various auction forms to
bid-taker cheating, bidder collusion, and other forms
- Auctions with a continuum of bidders? (Weird.)