Description

Hypotheses:

SQ1: The importance of the state of the economy into the model increases the predictive power of Google Trends volumes on the new car model sales.

SQ2: The predictability of Google Trends volumes and the state of economy differ between India & South Korea.

SQ3: The presence of seasonality will impact the predictability of Google Trends volumes and the price of car sales across countries.

SQ4: The incorporation of minimum time lag into the model impact the predictive power of optimal time lag in new car sales between India and South Korea.

SQ5: There is a noticeable difference between the time lag between India & South Korea.

           Sub: In both countries, there is a longer time lag for high price cars than low price cars.

The definitions in the questions merit some comments.

. - The state of the economy can be measured through the GDP or the Industrial Production Index of the countries included in the research. The state of the economy should be expected to be correlated with both new car model sales and Google Trends volumes, for example, if the economy is booming one would expect higher car sales and also an increase in web searches for information on cars for sale (similarly one would expect the opposite behavior if the economy is declining).

- The minimum time lag would be defined as a lag greater than zero and less than the optimal lag that exceeds one of the confidence limits of the cross-correlation function between new car model sales and Google Trends volumes. The idea is that this lag represents those buyers who do less web search before making the purchase, by including only the optimal lag you are omitting information about this market niche. It is assumed that incorporating the optimal lag improves forecasting[1] and we want to evaluate whether the additional inclusion of the minimum time lag increases the predictive capacity of the model. For example, in the graph of the cross-correlation function on page 44 of the sample thesis, the minimum time lag would be 1.

- The maximum lag is defined as the more distant lag greater than the optimal lag that exceeds one of the confidence limits of the cross-correlation function between new car model sales and Google Trends volumes. The idea is that this lag represents those buyers who require many web searches before deciding to buy, as in the previous case this is a market niche whose information is omitted if we incorporate only the optimal lag. For example, in the graph of the cross-correlation function on page 44 of the sample thesis, the maximum lag would be 10.

. - In models whose main purpose is to predict, seasonality plays an important role, in some cases its effect is positive and in others negative. For this reason I propose to evaluate how its presence affects and for this purpose we would use as a benchmark the model including the Google Trends volumes and the state of the economy with all the seasonally adjusted series and we would compare it (using R2 or MSFE)[2] against the model that includes all the previous variables without seasonally adjusted. The seasonal decomposition method would be used to construct the seasonally adjusted series.

MSFE = Mean Squared Forecast Error