- 102
- 253 105
The Logic of Risk
Switzerland
เข้าร่วมเมื่อ 10 มี.ค. 2017
Hi there, I am Pasquale and "The Logic of Risk" is my channel containing videos about risk management, extreme value theory, probability, and applications in economics and finance.
I will try to share my knowledge in a clear and hopefully funny way.
For comments and suggestions, please do not hesitate to contact me.
It would be nice if you could also follow me on Twitter: @DrCirillo.
If you want to have an idea of my teaching, you can have a look here: www.edx.org/course/introduction-credit-risk-management-delftx-tw3421x-2 (free and open)
More info about me: www.pasqualecirillo.eu
I will try to share my knowledge in a clear and hopefully funny way.
For comments and suggestions, please do not hesitate to contact me.
It would be nice if you could also follow me on Twitter: @DrCirillo.
If you want to have an idea of my teaching, you can have a look here: www.edx.org/course/introduction-credit-risk-management-delftx-tw3421x-2 (free and open)
More info about me: www.pasqualecirillo.eu
Low Default Portfolios (Part 2)
We continue the discussion on LDPs, focusing our attention on the beta-binomial model, and the results of Tasche (2013).
The link to the paper is at the end of this description.
Which estimator should I use? Answering this question is not easy.
From a theoretical point of view, the Bayesian estimators (with uniform, Beta, and the "conservative" prior of Tasche, 2013; just have a look) are to be preferred, because of their sounder probabilistic background.
Moreover, they also allow for the estimation of the PD even when k is not zero, but we do not trust the simple MLE.
These are also the estimators commonly discussed by experts and (partially) regulators (again, have a look at Tasche, 2013), even if the Bayesian approach is not always (and this is wrong!) considered as reliable as the frequentist one.
From a practical point of view, the Bayesian estimators with uniform and conservative prior give similar results, and they are easier to "sell".
The one relying on the Beta distribution is much more flexible, but we need to well motivate the choice of the parameters, for example with sound economic evaluations.
The other estimators we have considered, despite their simplicity, are very quick to compute and generally do not require the elicitation of any prior. They can be used for quick evaluations.
In particular, the simple p=1-C^(1/n) shows interesting performances for C around 0.4 and n below 100.
Tasche (2013): arxiv.org/abs/1112.5550
The link to the paper is at the end of this description.
Which estimator should I use? Answering this question is not easy.
From a theoretical point of view, the Bayesian estimators (with uniform, Beta, and the "conservative" prior of Tasche, 2013; just have a look) are to be preferred, because of their sounder probabilistic background.
Moreover, they also allow for the estimation of the PD even when k is not zero, but we do not trust the simple MLE.
These are also the estimators commonly discussed by experts and (partially) regulators (again, have a look at Tasche, 2013), even if the Bayesian approach is not always (and this is wrong!) considered as reliable as the frequentist one.
From a practical point of view, the Bayesian estimators with uniform and conservative prior give similar results, and they are easier to "sell".
The one relying on the Beta distribution is much more flexible, but we need to well motivate the choice of the parameters, for example with sound economic evaluations.
The other estimators we have considered, despite their simplicity, are very quick to compute and generally do not require the elicitation of any prior. They can be used for quick evaluations.
In particular, the simple p=1-C^(1/n) shows interesting performances for C around 0.4 and n below 100.
Tasche (2013): arxiv.org/abs/1112.5550
มุมมอง: 692
วีดีโอ
Low Default Portfolios (Part 1)
มุมมอง 1.4Kปีที่แล้ว
A Low Default Portfolio (LDP) is a portfolio characterized by a low number of defaults. Too simple? Citing the BCBS (Basel Committee on Banking Supervision): Several types of portfolios may have low numbers of defaults. For example, some portfolios historically have experienced low numbers of defaults and are generally-but not always- considered to be low-risk (e.g. portfolios of exposures to s...
Pseudo-sums, contingent claims, and a generalized memoryless property
มุมมอง 304ปีที่แล้ว
This is a talk I gave at the Peter Carr Memorial Conference last June 2022 in New York. In the talk, I discussed some joint work with Peter Carr. Those interested, can find details in our paper: papers.ssrn.com/sol3/papers.cfm?abstract_id=4081193
Brain Teasers: 13. "A drunk man on a bridge"
มุมมอง 2.4K3 ปีที่แล้ว
This is a nice exercise that can be solved using Markov Chains, Random Walks and Martingales.
Brain Teasers: 12. A simple symmetric random walk
มุมมอง 2K3 ปีที่แล้ว
Very easy exercise about the first moments of a symmetric random walk.
Brain Teasers: 11. Expected number of tosses
มุมมอง 3.9K3 ปีที่แล้ว
This is another exercise relying on Markov chains. It is strictly connected to Episode 10, which I suggest to watch.
Brain Teasers: 10. Winning in a Markov chain
มุมมอง 1.5K3 ปีที่แล้ว
In this exercise we use the absorbing equations for Markov Chains, to solve a simple game between two players. The Zoom connection was not very stable, hence there are a few audio problems. Sorry.
Brain Teasers: 9. Tossing a die and American options.
มุมมอง 1.3K3 ปีที่แล้ว
A nice exercise that introduces the reasoning behind American Options (discrete case).
Brain Teasers: 8. A little Bayes.
มุมมอง 9003 ปีที่แล้ว
A simple application of the law of total probability and of Bayes' formula.
Brain Teasers: 7. Do we meet?
มุมมอง 9303 ปีที่แล้ว
An interesting exercise combining geometry and probability.
Brain Teasers: 6. Observing or not an event and its probability.
มุมมอง 9753 ปีที่แล้ว
In this short video we discuss two exercises that look quite similar, but that have different solutions. The reason? It all depends on what we know and what we observe. Formally, we are dealing with two applications of Bayes' formula (not covered here).
Brain Teasers: 5. Do you like anagrams?
มุมมอง 7323 ปีที่แล้ว
Let's consider permutations with and without repetition, to see how we can create anagrams of words like One and Mississippi.
Brain Teasers: 4. Is a series convergent?
มุมมอง 1K3 ปีที่แล้ว
In this exercise we quickly review some tests to verify whether a series is convergent or not. More info about these tests (and others) can be found in any standard book on calculus like the one by Stewart.
Brain Teasers: 3. Pigeons, ants and occupancy
มุมมอง 1.6K3 ปีที่แล้ว
In this video we see an application of the (generalised) pigeonhole principle (en.wikipedia.org/wiki/Pigeonhole_principle).
Risks and fat tails
มุมมอง 3.5K3 ปีที่แล้ว
This is the video of my talk at the Conference on Complex Systems 2020 (CCS2020), in the satellite event organised by Alfredo J. Morales (MIT) and Rosa M. Benito (Technical University of Madrid). For privacy reasons, I have cut the video, not to show the pictures of the other participants via Zoom. More about the event here: sites.google.com/view/complex-systems-applications. We speak about ris...
Solution of Exercise 8 using Markov Chains
มุมมอง 1.5K4 ปีที่แล้ว
Solution of Exercise 8 using Markov Chains
QRM 9-1: Market risk and historical simulation
มุมมอง 7354 ปีที่แล้ว
QRM 9-1: Market risk and historical simulation
Fin Math L11: Numeraire, T-forward measure and interest rates
มุมมอง 2.4K4 ปีที่แล้ว
Fin Math L11: Numeraire, T-forward measure and interest rates
QRM 8-3: R lesson on extremal index, records and Garch
มุมมอง 9474 ปีที่แล้ว
QRM 8-3: R lesson on extremal index, records and Garch
I like the way you speak, slowly and easy for me to listen and understand
Glad you like it.
Complimenti, lezione chiara ed interessante, grazie.
There's no word to express my gratitude for the knowledge you share and how you share it
My pleasure
Innovative and Informative Lecture.Thanks
Ĺ'uccello padulo nero però , che è quello che fa più danno...
Thanks for the lesson. Given the unpredictable nature of this risk, would you consider it controllable, hence a good measure of a business managers performance?
u explain this very well brother
sir the best video i just found probably after 5 to 6 i got this one thnqu
Congrats, really intuitive explanation
thanks for your explaination
Amazing lecture
THANK YOU!
Thank you Pasquale, I'm learning a lot from your videos. They complement beautifully with RWRI.
Is stretched exponential in the MDA of Gumbel distribution as well ? merci
Thank you very much for such well explained lecture.
Hi ! The lecture is great but haven't found the book anywhere... do you have an idea where I can find it ?
Great
Thanks
Great exercise to illustrate markov chains, and well explained. Thanks!
Thank you sir
Hi Dr Cirillo, two questions: 1)with FRTB, banks continue to use VaR for backtesting, Is It correct? If yes, where Is It written? 2)how to do backtesting for a VaR calculated with historical simulation? In both questions i refer to market risk. Thanks many
You easily find all answers in every recent risk management textbook. Have a look at Hull's, for example.
@@TheLogicofRisk Is risk management and financial institution a good book?
@@vincenzocardone510 For a broad, basic treatment, definitely yes.
Your content is seriously amazing. Very much appreciated.
Glad you enjoy it!
Grazie! Puoi lasciarmi la tua email?
Hi, could you share the material with us? Thank you.
Check description
Thanks mate for the content. What book are you using?
My own lecture notes (linked in the descriptions when needed).
Thank you so much.
You're welcome!
Great video, thank you!! I did the crypto assignment. My results suggest that empirical VaR & ES estimations are not inferior to the GPD approach & do not underestimate the risk linked to Cryptos. Is that correct??
Yes for VaR, for ES it is a little more subtle.
great video. This helped my research. Thank you sir.
Glad to hear it!
Still at this point in the course, just a noob question: if I want to let’s say gamma hedge and I apply the formula you’ve shown using the gamma of an option, isn’t this gamma calculated assuming the underlying asset to follow a gaussian distribution? Isn’t there a problem in case this isn’t true? Anyway thanks for this material, very interesting
No, there is no assumption on the underlying distribution.
I am trying to run some of the r code and it is not working. Are there additional packages that are needed for it to work?
specifically regarding the stuff using ineq package/GINI calculation. For me it just returns an empty plot. I was able to get the other stuff to work.
Just the basic installation and the ineq package with its dependencies.
The pigeonhole principle in that problem don’t make much sense cause you have no information about the distribution of ants on the plane.
You don't need the distribution.
The lecture states N(d1) is the probability of St > K using the D measure My understanding has been that N(d1) is the mathematical equivalent to the expectation of the stock conditioned on St > K using the risk free Q measure, discounted to today. Are these two statements both correct?
N(d1) is a probability, it cannot be an expected value, unless you consider the expectation of a random indicator.
Thanks a lot for this video and the whole series, I'm really enjoying it. I was wondering, in ARCH (and GARCH) models, why do we use lagged squared returns, and not lagged variances? We try to forecast variance, yet we use returns as inputs. Are they considered equivalent under the assumption that mean returns = 0 ? I understand that that we cannot compute the daily volatility on 1-day data, but could we use rolling windows for instance?
You introduce lagged variances for GARCH. In ARCH you just look at the second order moments (not necessarily centered). Then, it is true that E[R]=0, but it is not strictly necessary.
@@TheLogicofRisk Thanks a lot for your answer, and again for providing this valuable content online !
What's the probability the man will realize his position, know he's on the edge & change how he moves?
Given the text, 0.
@@TheLogicofRisk That would be an interesting self-awareness, physics type of question
Thanks for putting this up. What material would you recommend as an intro to EVT for an undergrad with a fairly basic grasp on statistics?
I would first improve my basic knowledge of statistics.
@@TheLogicofRisk Thanks for answering. Are there some core concepts outside of ML Estimators, confidence intervals and hypothesis testing I should look into before I move on? That's all the stuff the course I am taking right now is covering
@@philippweisang Order statistics, basics of robust measures, LLN, CLT and, if possible, a bit of Poisson processes.
@@TheLogicofRiskThis is so helpful, thanks again. I look forward to learning all this stuff
Hi there. How can I have all the plots show when using "plot(fit)" without using the menu? Do you have to manually use the menu, or can you somehow turn off the menu? Thanks for your videos.
You can edit the function as you prefer, with edit().
Great video! You are a superb teacher. Enjoying your channel a lot.
You are really the best
Thanks for watching.
Wow.. so little views for that kind of useful content.. I'm glad I found this channel. I hope you won't quit making videos cause I really hope to learn from you
There will be something new in a few weeks.
great content, thanks so much for sharing this online.
Thanks.
quindi va bene ancora per comprendere la povertà....scusami se la domanda può sembrarti stupida
Da solo no.
Brilliant 👍🏼
This is really good material. I'd love if you could recommend other books on QRM. Greetings from Mexico.
Thank you very mush fo this video professor.
Hi as I am searching for Credit risk tutorial this is only best channel I will say. Can u please suggest any source or material to build credit PD model for mortgage loan using python or R
This is great, watching from New York. Is it possible to receive an emailed copy of exams to practice?
We cannot share the texts of the exams. It is a policy of the university. Sorry.
this video draws some great links between related concepts, fantastic. thanks for making this available for all.
i use distortion risk measures a lot and have been trying to understand the link with RN derivatives, this video really helped me to understand the relationship
Dr Cirillo First of thank you for this excellent series. Quick Question @2:46 when you talk about VAR, you mention some links for learning basics of quantiles and practice problems, is that video still available somewhere ?
Have a look at my risk management course: th-cam.com/play/PLgCR5H4IzggGihtfhTtA0fxGiBU8DMWHq.html
@@TheLogicofRisk yes found them in that series. thanks very much.
Great series of talks. Thank you! @4:19 the call function for getSymbols has been changed recently to getSymbols("AAPL", from="2016-01-01", to="2021-01-06", src="yahoo", auto.assign = getOption('loadSymbols.auto.assign',TRUE))
Hi Pasquale, this is Evellyn from RWRI 15. I have been following your lecture here on youtube, you did excellent work with the explaining. I did the exercise for ethereum, is there a way i can send my work to you for feedback? thank you so much :)
You find my contact details on my personal website.