（1）

For the x_t variable, the p-value is 0.000 which is much less than 0.05 indicating that x_t has a
significant effect on y_t at 5% level of significance.
Since the p-value of the F-value is 0.0002, this means that the whole regression model is
significant at 5% level of significance.
R² = 0.1367 which means that x_t explains 13.67% of the variation in y_t. This low value of R²
indicates that x_t has limited ability to explain y_t, which means that the model is not well fitted.

（2）

R² = 0.9744, which means that x_t_drift explains 97.44% of the variation in y_t_drift.
The inclusion of the drift term significantly improves the fit of the model compared to the previous
regression, indicating that the drift term significantly improves the model's ability to explain the
data.
（i）

ADF test

The Z(t) statistic of -3.310 is less than the critical values at the 5% and 10% levels (-
2.926 and -2.598) and rejects the hypothesis that there is a unit root for the inf
variable.

PP test

The Z(t) statistic is -3.310, which is less than the critical values (-2.926 and -2.598) at
the 5% and 10% levels, rejecting the hypothesis that there is a unit root in the inf
variable.
Therefore, with no drift and time trend, the hypothesis that there is a unit root in the
inf variable is rejected, indicating that the time series is smooth.
（ii）
ADF TEST

The test statistic -3.310 < 1% critical value -2.399 and the p-value (0.0008) is also
less than 1%, which means that we reject the original hypothesis at 1% level of
significance

PP TEST

The test statistic -3.342 < 5% critical value -2.926 and 10% critical value -2.598 but >
1% critical value -3.573. p-value of 0.0131 indicates that we reject the original
hypothesis at 5% level of significance.
In conclusion, with a drift, inf is smooth time series data.
（iii）

ADF TEST
Z(t) value = -3.376 is less than -4.139, which rejects the original hypothesis of unit
root at 10% level of significance H0

PP TEST

Z(t) value = -3.376 is less than -4.139, which rejects the original hypothesis of unit
root at 10% level of significance H0
In conclusion, with time trend, inf is smooth time series data,at 10% level of
significance

（iv）

ADF TEST
The test statistic -1.699 > 1% critical value of -4.143, 5% critical value of -3.497 and
10% critical value of -3.178 indicates that we failed to reject the original hypothesis at
the commonly used significance levels of 1%, 5% and 10%

PP TEST

.
The test statistic -3.439 < 10% critical value -3.177 rejects the original hypothesis at
10% level of significance but fails to reject the original hypothesis at 5% level of
significance.
In conclusion, the more robust Phillips-Perron test result is favoured as it is able to
handle autocorrelation and heteroscedasticity.
The graph shows that the inflation rate (infl) has fluctuated considerably over time, especially from
the 1970s to around the mid-1980s, when it showed a significant upward trend. In the following
period, inflation has gradually declined and stabilised, especially after 2000, with some
fluctuations, but the overall fluctuations have been small.
Based on the graphical representation, infl does not appear to exhibit a typical stochastic trend
(i.e., there is no sustained non-stationary upward or downward trend). In a time series with a
stochastic trend, the values would be constantly drifting with no sign of mean reversion. However,
in this graph, although there are some large fluctuations, the overall view of inflation does not
show a clear continuous stochastic wander, but instead shows some mean reversion
characteristics, especially after 2000.
Therefore, the initial judgement from the graph is that infl may not have a stochastic trend, but
this judgement needs to be further verified by the ADF unit root test.

(i)
At the 5% significance level, the test statistic (-2.990) is less than the 5% critical value (-2.880), so
we reject the original hypothesis of a unit root, i.e., we reject the hypothesis that the variable infl
is randomly wandering. This implies that infl is likely not to have a random trend, suggesting that
it may be smooth.

(ii)

At 10% significance level, the test statistic (-3.227) is less than the critical value of 10% (-3.130),
so we reject the original hypothesis of unit root. However, at 5% and 1% significance levels, the
statistic fails to meet the requirement of rejecting the original hypothesis, suggesting that there is
still some uncertainty in the results at tighter significance levels.
Combined with the results in part (i), the inclusion of the time trend makes the results of the ADF
test insignificant, suggesting that the time trend of infl has a significant impact on the test results.
The inclusion of the time trend seems to make it more difficult to reject the unit root hypothesis,
suggesting that in this case infl may have the characteristics of a random trend.

As can be seen from the results, with 2 lag terms used, the test statistic is -2.990, which is greater
than the critical value of -3.461 for 1%, but less than the critical value of -2.880 for 5% and -2.570
for 10%, with a p-value of 0.0358. therefore, at the 5% significance level, we reject the original
hypothesis. In this result, 2 lags is a reasonable starting point, but further selection of the optimal
number of lags can be done by determining the lag order through the varsoc command
In terms of multiple information criteria (AIC, HQIC, SBIC) and LR statistics, the model with lag 3
performs best. Specifically: at lag 3, the AIC, HQIC, and SBIC values are all smaller than the other
lags, indicating that this is the optimal lag order.

The ADF test in (i) shows that the t-statistic is -2.990, which is less than the critical value of -2.880
at the 5% significance level, so we reject the original hypothesis of unit root. This means that
based on this result, we conclude that Infl's autoregressive (AR) model does not contain a unit
root.
ii.

From the results I can see that the mean value of the error of the pseudo out-of-sample prediction
is -3.060771, the mean value of the error is not zero, which shows that the pseudo out-of-sample
prediction is biased.
iii

With an RMSFE of 3.63, the large RMSFE value suggests that the model may not have sufficient
predictive power in the 1963 to 2002 sample period, and thus the pseudo out-of-sample forecasts
are not very accurate. This is consistent with the performance of the AR(2) model, suggesting that
the model may not have fully captured the changes in inflation over the period 2003 to 2017.
iv.
The price of oil has had a significant impact on inflation, particularly during the dramatic
fluctuations in the price of oil during the global financial crisis of 2008, in the following ways:
Impact of production costs: Oil is an important raw material or source of energy for many
production processes, especially in areas such as transport, chemicals and manufacturing. When
the price of oil rises, so do the costs of production, which can drive up commodity prices and thus
inflation. Conversely, when the price of oil falls sharply, the cost of production for firms is reduced
and commodity prices fall as a result, leading to lower inflation.
Impact of energy costs: Oil prices directly affect the price of fuels, such as petrol and diesel. When
oil prices fall, consumers spend less on fuel, transport costs are reduced, and price pressures on
goods and services are reduced, thus depressing the overall level of inflation.
Ripple effect: Changes in oil prices not only affect the industries that use oil directly, but also
affect many other industries through the supply chain. For example, lower transport costs lead to
lower prices for food, consumer goods, etc., which has an important impact on the overall price
level of the economy.