The autocovariance function is the sequence of autocovariances of a time series across lag lengths.
$$$$
What it looks like
For a covariance-stationary series, the autocovariance function is:
$$ {\gamma_k}_{k=0}^{\infty} $$
where each (\gamma_k) measures covariance between (y_t) and (y_{t-k}).
Why economists use it
The autocovariance function summarizes the dependence structure of a time series. Different models imply different patterns, so economists use it to diagnose persistence, identify likely models, and understand whether shocks fade quickly or slowly.
From raw dependence to normalized dependence
The autocovariance function is closely related to the autocorrelation function. The difference is that autocovariances retain the scale of the series, while autocorrelations normalize by variance.
Related Terms
Knowledge Check
### The autocovariance function is:
- [x] the collection of autocovariances across lags
- [ ] the cumulative sum of inflation rates
- [ ] a market-clearing condition
- [ ] a tax schedule
> **Explanation:** It describes how dependence changes with lag length.
### Why is the autocovariance function useful in time-series analysis?
- [x] Because different models generate different lag-dependence patterns
- [ ] Because it eliminates variance
- [ ] Because it applies only to cross sections
- [ ] Because it replaces all regressions
> **Explanation:** Economists use the function to understand persistence and to help identify model structure.
### The autocorrelation function differs because it:
- [x] normalizes autocovariances by the variance
- [ ] ignores lag structure
- [ ] applies only to nonstationary series
- [ ] measures only the sample mean
> **Explanation:** Normalization makes the dependence easier to compare across series.