Forecasting the expenditure on household appliances in SA
Exponential smoothing is relevant for South Australian expenditure on household appliances between 19830-2019. Exponential smoothing would best suite this data due to the time series forecasts methods are weighted averages of past observations, with the weights decaying exponentially as the observations get older.
This time series line chart is both a trend and a seasonality time series. It is a trend because of the consistent and underlying increase from period to period. This can be explained by the increase in the South Australian population and housing growth. It also has seasonality in it due to repeatable fluctuations in the time series. This line chart also has a structural break between 2012-2019 which seems to be level and seasonal. This structural break can be explained by the increase in renters in the housing market, compared to a decrease in the home owners in the market.
The smoothing here predicts that the level trend will continue into 2020. The orange (forecast) line is very close to the actual (blue) line. This indicates that seasonality is strongly connected to the data as well as the weight leaning towards to the more recent data instead of historical data. The forecasts for the winters exponential smoothing model seem reasonable as they account for the trend and seasonal components in the original time series. The trend upward every December can also be seen in the forecast. The forecasts reasonably extrapolate these components into the future to provide the forecasts.
Critique
Due to the recent events of covid-19, the economic and environmental factors for this data is enormous and couldn’t have been predicted for. To project anything beyond one year would be unrealistic due to the virus. Due to the trend of more renters in the market will affect the expenditure of household appliances due to these individuals may have lost their job to the virus and may have intended of renting a property in the next 12 months, cannot not due to their income be diminished. If the model were to continue, more weight would have to be given to the recent data instead of historical.
Regression
Critique
I would change the amount of observations in the regression, by increasing the amount of observations, we can get a more accurate regression model and increase the R squared. The regression analysis seems to underestimate the expenditure for 2020 compared to report WES method. This could be due to the changes in the alpha, beta and gamma. Regression analysis though allows for strategy and scenario based predictions. Through regression assessing the alternative scenarios, it can provide a range of the likely values for the forecasting time, and helps executives makes a more knowledgeable decision.