Associative Memory Flashcards
(17 cards)
What matters for associative memory and similarity based recall
- Context
- Memory Database
- Agents only use subset of information
What gets cued determines recall
Framework of similarity based recall
What gets cued is based on similarity of historic context with current context over a similarity function
-> Attributes only affect decisions if experienced in a context similar to today
Modeling approaches (Effects)
Memory based overreaction
- Direct effect, updating of beliefs following news
- Indirect effect, recall past news, which also update beliefs
-> Beliefs and behavior look like they overreact to news
Similarity and Interference
- Events are more likely to be recalled if they are more similar to what is considered today
- Events are less likely to be recalled if they are more similar to alternative hypotheses
Benefit over Learning from experience
Within-Situation variation and within-subject variation
Beliefs and Market Interaction, Design (cue)
Subsequent presentation of news about a company. They randomize over number of cues in the second period and the signal (positive, negative).
Signal is embedded in a context (story, image) which is uninformative.
How does the lab study identify effects? Within/between
Within Subject variation
Between subject variation
What are the results of cued context about company quality
- Second period beliefs overreact to second period signals in cue but not nocue
- Magnitude of overreaction increases with #cue
Field Findings, Setup
Associative recall in stock return expectations
Survey Chinese retail investors with open ended questions on a past market episode that comes to mind and expected future returns
Field, Result
- Investors retrieve both recent and distant events feat. dramatic episodes
- Retried memories are similar to todays performance of the market
- Strongly predict return expectations
- Retrieved memories increase forecast error
Beliefs about economy @ household level, (income changes and expectations)
- Matched survey data from Denmark
- Hh expected inflation neg. varies with hh income changes
- also perception of realized inflation
-> Selective recall cued by affective associations as a mechanism leading to non domain specific experience effects
Beliefs about economy @ household level, Results
Neg. correlation between income and inflation forecast
-> biased recall
Attention to the Macroeconomy, Setup
Measure attention allocation with open ended questions
Two measures:
- Cohort level: Having lived through oil crisis
- Idiosyncratic pre shock income or wealth losses
Attention to inflation, Results
People that lived through oil crisis pay more attention to inflation
De-Anchoring, Setup
Check whether associative memory or learning from experience provides better model for building beliefs and decisions
- Post-pandemic surge in inflation
De-Anchoring, Theoretic differences
Learning from experience:
- Older cohorts have larger memories and put less weight on past events (update more)
Associative memory:
- Older cohort´s experience of high inflation cued by recent inflation (update more)
De-Anchoring, Results
Older cohorts update more strongly despite facing lower actual inflation
-> Consistent with associative memory
