# Event Studies and Intraday Research Estimate event-associated behavior while accounting for selection and timing. Use this workbook alongside the course. Write your answers before opening the solutions. Practical work is self-reviewed; scored knowledge checks are in the Academy. ## 1. Event definition ### Scheduled versus unscheduled events Scheduled events have a known release plan; unscheduled events may have uncertain first-publication times. Define the event source and timestamp standard before collecting market responses. A later article's time may not mark the first information arrival. ### Release versus observation time Observation period and release time differ. A report about last month becomes public at its release, not during the month it describes. A market study must align returns with information availability. ### Surprise measures A surprise measure compares actual information with a specified expectation, such as a recorded consensus. Preserve the expectation as known before release. A later updated consensus cannot be substituted silently. ### Revisions and overlapping news Revisions and overlapping announcements can affect interpretation. A headline value may be accompanied by changed prior data or another release. Record these rather than attributing every price move to one selected number. ### Worked example A release beats consensus but revises prior data downward. A one-variable label positive surprise may omit information that participants received simultaneously. ### Independent exercise Design an event record containing actual, prior, revised prior, pre-release expectation, source and release timestamp. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 2. Study design ### Pre-event and post-event windows Define pre-event and post-event windows before inspecting outcomes. Different horizons answer different questions. Choosing the best-performing interval afterward is another form of model selection that must be disclosed. ### Control observations Control observations help describe ordinary behavior under comparable conditions. Matching by time of day or other predetermined features can reduce obvious differences, but it does not automatically establish causal identification. ### Time-of-day effects Time-of-day effects can influence volume, spread and volatility independently of the event. A comparison between event mornings and quiet nights can confuse session patterns with event-associated changes. ### Exclude rules defined before results Exclusion rules should address known data defects or overlapping conditions symmetrically. Removing only events with inconvenient responses produces a biased sample, even if each removal has a plausible retrospective story. ### Worked example An analyst examines one-, five-, fifteen- and sixty-minute responses and reports only the most favorable horizon as if it was chosen in advance. The horizon search is part of the experiment. ### Independent exercise Write a protocol that either preselects one primary horizon or explicitly treats multiple horizons as exploratory. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 3. Measurement ### Returns and volatility Returns require a defined price stream and timestamps. Use quotes or trades appropriate to the question and distinguish observed returns from executable strategy P&L. A midpoint movement is not necessarily a tradable round trip. ### Spread and liquidity Measure spread and liquidity changes where available. An event with a large price response can still be difficult to trade because costs and fills deteriorate. Price response and implementation feasibility are separate results. ### Outliers Outliers should be investigated and reported. An extreme response may be genuine, a data error or a timestamp problem. Do not automatically delete it or assume it represents a repeatable typical event. ### Missing tick coverage Missing tick coverage can bias the measured first reaction and extremes. State whether the data can support the requested horizon. A coarse bar cannot be used to claim millisecond timing or a precise first executable price. ### Worked example A dataset contains only one-minute bars, but the report claims the price moved within 200 milliseconds of release. The data resolution cannot substantiate that timing statement. ### Independent exercise Rewrite the claim at the resolution supported by the data and list what additional data would be needed for the finer claim. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 4. Interpretation ### Multiple horizons Multiple horizons and subgroups increase the search space. Report the full analysis plan and distinguish primary tests from exploratory comparisons. An attractive subgroup discovered after inspection needs new evidence. ### Overlapping samples Overlapping windows can reuse the same market movement across events or observations. This creates dependence and can overstate precision if treated as independent data. Identify overlap explicitly. ### Alternative explanations Alternative explanations include simultaneous news, changing expectations and general liquidity conditions. An event-associated response does not uniquely identify a causal channel without stronger design assumptions. ### Net-of-cost implementation limits Translate any research finding into an implementation test separately. Include signal availability, order delay, spread and risk. A statistically clear historical reaction may still be too small, too fast or too costly to exploit. ### Worked example Two announcements occur minutes apart and their response windows overlap. Assigning the full move to each and treating the observations as independent double-counts shared information. ### Independent exercise Explain how to report this overlap and what conclusions should remain limited. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## Course project Submit a preregistered event-study protocol and a reproducible worked sample. ### Self-review rubric - Concepts and reasoning: 25% - Calculations, data and evidence: 30% - Process and risk controls: 25% - Limitations and communication: 20% Record one correction and one next practice task. This rubric is not automatically graded. ## Worked solutions ### Exercise 1 The record should retain every field separately and identify when each became available. It supports a more complete interpretation without proving that any one field caused the observed price response. ### Exercise 2 Name the primary horizon and secondary analyses before evaluation, or disclose all tested horizons and validate any selected conclusion on later events. Do not rewrite the study history around the best result. ### Exercise 3 Describe the observed one-minute interval and its limitations. Millisecond claims require appropriately timestamped higher-resolution observations, synchronized clocks and a reliable release time. ### Exercise 4 Flag overlapping events, use an appropriate grouping or exclusion rule defined in advance, and assess dependence. State that attribution to a single release may remain unresolved even if the joint window shows a clear move. ## Further reading - https://www.bls.gov/schedule/news_release/ - https://www.itl.nist.gov/div898/handbook/