# Attention or Returns – What’s First? A Granger Causality Analysis

What came first? The chicken or the egg? In financial markets, we sometimes ask a similar question. Does investor attention affect market returns? Or do extreme market returns draw investor attention? Nadia Vozlyublennaia tries to solve this question in her 2014 Journal of Banking & Finance paper “Investor attention, index performance, and return predictability”. She considers different asset classes: stocks, bonds, and gold. Google Search Volume (GSV) is the proxy for attention.

### Chicken or egg? Granger causality helps!

Vozlyublennaia then runs a vector autoregression (VAR). This model allows you to explore the relation between two variables *X* and *Y* by estimating the impact of lagged values of *X* on Y and vice versa:

*X(t)=a*X(t-1)+b* Y(t-1)*

*Y(t)=c*X(t-1)+d* Y(t-1)*

You can interpret the results as follows: if b is significantly different from zero, variable *Y* affects *X*. If c is significantly different from zero, *X* affects *Y*. Assume that c is significantly different from zero, but that b is not. Then, you can conclude that *X* affects *Y*, but *Y* does not affect *X*. The econometric term for this relation is “Granger causality”. By using higher-order lags of *X* and *Y*, you can capture slower adjustment processes.

### Asset returns and investor attention

Vozlyublennaia uses asset returns R and GSV as *X* and *Y*. Table 1 presents a snapshot of the estimation results:

Stocks | Bonds | Gold | ||||

R(t) | GSV(t) | R(t) | GSV(t) | R(t) | GSV(t) | |

R(t-1) | -0.145 |
-6.361 |
-0.106 |
1.487 | -0.047 | 1.129 |

R(t-2) | -0.023 | -1.078 | 0.123 |
-.483 | -0.045 | 1.765 |

GSV(t-1) | -0.004 |
0.728 |
-0.001 | 0.496 |
-0.001 | 0.804 |

GSV(t-2) | -0.001 | 0.138 |
-0.002 | 0.019 | –0.004 | 0.086 |

Table 1: VAR for search probability and index return. Information taken from Table 3 of Vozlyublennaia (2014). Bold font indicates statistical significance at the 5% level or higher.

The results in Table 1 show that attention affects stock returns (positive or negative) and vice versa. For bonds, there is no relation between attention and returns. For gold, returns affect attention, but attention does not affect returns. So, Granger causality runs from returns to attention. This difference is economically sensible: transaction costs for bonds and gold are high. Hence, retail investors play a smaller role. And, as discussed in this post, GSV measures retail rather than institutional investor attention.

### Not all attention is equal

In summary: In stock markets, attention affects market returns and vice versa. But markets where institutional investors play a larger role do not show an impact of attention on returns.