miomlao | Econometrics & Markets

$IBM dropped -25.2% in a day. From a GARCH perspective this isn't just a one-off price move or a normal fluctuation - it's a volatility shock with its own dynamics I used the Parkinson estimator on today's range, $204.50 - $219.95 - daily volatility comes out to 4.4%. Options for tomorrow's call are pricing another 7% swing. The gap between these two numbers is persistence. Volatility after a shock like this doesn't disappear overnight, it decays gradually 7% for a company with a $190B market cap is a risk you can't ignore. The regime hasn't returned to pre-crisis levels yet. Tomorrow decides whether the volatility keeps decaying, or a new shock lands on an already elevated base I wrote about this exact tool almost a month ago. Now's the best time to actually put it to the test - the day before the print

Your S&P 500 index fund isn't diversified. This week's earnings from 3 companies will prove it - they're what actually moves the whole index Alphabet and Tesla report on July 22, Intel on July 23 Alphabet weighs 5.89% in the index - but delivered more than 20% of its entire return this year. Intel weighs 0.39%, practically invisible on paper, but contributed 8% of returns on a stock up +163% YTD. One company that barely moves the index by weight has, in practice, defined almost a tenth of its movement The options market is pricing a 13.5% swing for Intel $INTC versus 9.17% for Alphabet $GOOGL and 6.79% for Tesla $TSLA One disappointment and the "diversified" index drops like it holds just 3 stocks, not 500 In recent years the S&P 500 has stopped being about 500 companies - it's about a small group of names that decide where the whole index goes Diversification gets harder every year, and the market ends up depending only on giants

IBM's CEO wrote a letter to investors. Three words led to -25% in a single day: "we did not anticipate" On July 14 $IBM lost $73 per share in hours - the worst day in 115 years of company history, worse than Black Monday Clients shifted their budgets in the final weeks of the quarter from IBM's software to AI hardware. Servers, memory, chips. IBM is standing in line while all the money went to Nvidia and Micron July 22 is the full earnings call. The market is expecting a cut to annual guidance. Before the crash the stock was up 139% over three years. In one week -28% YTD with a securities fraud investigation on top The question remains open: did the market overreact or is $IBM simply done growing? Waiting for the report to get at least some clarity on where the company is headed

After WWDC 2026 $AAPL dropped to $270. I saw it coming - this pattern plays out every single year I put in $15,000 at an average of $307. Today it's $333 While everyone was calling Apple a failure, you had to understand what this company actually is - it never stays down. Steve Jobs was pushed out in 1985, the company nearly went bankrupt, and they still came back and built the most valuable company in history The stock hit 15 intraday records in 2026. Market cap approaching $5 trillion, briefly making it the most valuable company in the world. Citi raised their target to $365 Profit is still minimal because I bought at a high price, but I believe the July 30 earnings will show consensus EPS of $1.88 The real story isn't quarterly numbers - it's the products. iPhone 18 in September, which like every year will give the stock its next boost. Holding for a long time. I believe in $350 before year-end

MIT just dropped a volatility modeling lecture from a practicing quant. Fall 2024. Free. Peter Kempthorne covers GARCH, stochastic volatility models, and volatility clustering - the math that explains why two portfolios with the same average return produce different outcomes. Volatility isn't symmetric: losses cluster, and that's exactly what turns it into a hidden tax on compound returns.Why high volatility literally subtracts from your return- full derivation in my article:

A Yale professor explains why losses hurt more than gains feel good - and how that's built directly into the math of finance. ECON 251. Free. Risk aversion, the Bernoulli brothers, and CAPM - it all starts with one question: how do you measure the value of risk to an investor. That same logic is behind the Kelly Criterion - you don't maximize money, you maximize the logarithm of wealth. Why -50% feels worse than +50% feels good - and why that's not just psychology, it's math:

MIT just dropped a portfolio management lecture from someone who actually built these portfolios at real banks. Fall 2024. Free. Dr. Jake Xia covers portfolio construction and risk parity - where position size is defined not by dollars but by risk contribution. That's the applied version of what I write about in my article: volatility isn't just a metric, it's a capital management tool. if you're still sizing positions by intuition - the math behind that decision is in my article:

A university professor just gave you a full lecture on the Kelly Criterion - the formula for optimal position sizing. Free on YouTube. Kelly is not "risk 2% per trade." it's a mathematical optimum derived from excess return and variance. On SPX historical data it gives 3.52× leverage - which is exactly why nobody runs full Kelly in real trading. Why half-Kelly survives where full Kelly blows up the account - full derivation in my article:

MIT just gave away a free lecture on portfolio management from someone who actually built portfolios at Goldman Sachs and Morgan Stanley. Dr. Jake Xia breaks down risk parity -the strategy behind Bridgewater's $150B fund. Not equal weight across assets, but equal risk contribution from each. Volatility as a management tool, not just a metric. If you're not managing risk through volatility - you're just holding assets. the math behind it - in my article:

This MIT professor just taught you the exact statistical foundation behind every quant strategy on Wall Street 18.650 by Prof. Philippe Rigollet. 25 lectures. Free Maximum Likelihood. Hypothesis testing. Regression analysis. Bayesian inference. The math behind every econometric model that actually predicts markets. Bookmark the course

Bridgewater uses Hamilton's regime-switching to run All Weather. Goldman runs daily VaR on 400+ risk factors through GARCH. The Fed has modeled monetary policy transmission through VAR since 1993. Citadel built its stat-arb pairs book on Engle-Granger cointegration. JPMorgan quantifies earnings surprise impact through event study methodology These aren't academic exercises. These are the five econometric tools that move trillions every single day The same five tools - explained from first principles, with the math behind each one

The core mechanism driving structural statistical arbitrage is the phenomenon of cointegration. This fundamental mathematical concept serves as the bedrock for any institutional spread-trading framework The comprehensive academic foundations and rigorous analytical principles of this methodology were originally presented by Professor Peter Kempthorne at MIT The framework synthesizes cointegrated vector autoregressions (VAR) with linear state-space formulations, utilizing dynamic Kalman filtering algorithms to isolate and track equilibrium relationships between financial assets asset pairs continuously

This Chicago Booth textbook is the only book that covers every major tool quants use for financial time series in one place. Goldman, Citadel, Two Sigma analysts all studied it. ARIMA and ARMA from chapter 2, GARCH and EGARCH from chapter 3, Markov Switching model from chapter 4 - the exact same framework Hamilton built in 1989. Plus high-frequency microstructure, continuous-time models, and multivariate VAR with cointegration. 600+ pages. one author. Wiley. Ruey Tsay spent decades at Chicago Booth building the bridge between econometric theory and financial practice. This book is that bridge like + bookmark. Read how five tools from this book work in real market prediction

#d=gs_qabs&t=1781691242730&u=%23p%3D9cIyvt2BIYEJ" target="_blank" rel="nofollow noopener">scholar.google.com/scholar?q=MacK…

Текст к посту: This Wharton paper from 1997 is the bible of event study methodology. 12,000+ citations. Every hedge fund uses it to measure market impact of earnings, M&A, and macro announcements Abnormal return = actual return minus expected return without the event. Estimation window - 120 days before. CAR aggregated across the event window with statistical testing. The paper proves that if markets are rational, any event's economic impact shows up in prices within days - no need for months of productivity data Pre-event abnormal returns detect information leakage before announcements. That single finding reshaped how regulators and traders approach insider activity. like + bookmark

In 2003, Clive Granger won the Nobel Prize for cointegration - the concept that two non-stationary series can share a hidden long-run equilibrium. This is the lecture where he explains how it was discovered Granger tried to prove a colleague wrong. in doing so he proved the opposite - and invented cointegration. Key results in the lecture: if two series are cointegrated, at least one must Granger-cause the other. Standard regression on non-stationary series without cointegration testing produces spurious relationships - a finding that forced economists to recheck decades of published papers The Federal Reserve and central banks worldwide use error-correction models built on his framework for policy simulations to this day like + bookmark

#VARtools" target="_blank" rel="nofollow noopener">princeton.edu/~sims/#VARtools

n 2011, Christopher Sims won the Nobel Prize for proving that VAR models can identify the real effects of monetary policy on markets. this is his Princeton page with everything he built → Nobel lecture on VAR methodology → VAR software tools in R and MATLAB - free, ready to run → Error bands for impulse responses → Full Time Series course with ARMA, structural VAR, Bayesian inference One key result from his 1996 paper: monetary policy shocks explain less of output variance than most economists assumed. the rest is noise - and VAR identifies exactly how much. like + bookmark.