1. Purpose and choice of variables

In this report, the main observation is Foreign direct investment projects licensed in Vietnam. The purpose of analysis is to spot the developing pattern of overseas capital in Vietnam, how effectively the fundings are used, and see if Vietnam can keep up with this development in the future.

The data is from the General Statistics Office of Vietnam at gso.gov.vn. There are three variables: the number of projects, total registered capital and implementation capital; which are looked at yearly from 1991 to 2018. In this case, the number of projects is primarily of interest. However, to decide on efficiency, total registered capital and implementation capital are also brought up.

2. Plotting main variable of interest

The numbers of Foreign direct investment projects from 1991 to 2018 in Vietnam is presented in the graph below.

number <- invest$`Number of projects`
tsnumber <- ts(number, end = 2018)
library(forecast)
plot(tsnumber, main="Foreign direct investment projects licensed in Vietnam", xlab = "Year", ylab = "Number of projects", type="l", col="black")

As can be seen above, the number of projects seems to have an exponential growth. Next, let’s take a look at autocorrelation.

acf(tsnumber, main="Correlogram")

The correlogram shows that autocorrelation constant is significant for the first observations, and then declines slowly. Therefore, the variable can be non-stationary.

3. Differencing method

Since the variable is non-stationary, the correlogram can be plotted using differencing method.

difnumber <- diff(tsnumber)
plot(difnumber)

acf(difnumber)

4. Moving average method

Here the number of projects is forecasted using moving average method. The period chosen is 4 years, therefore the model can only predict 8 years ahead.

movnumber <- ma(tsnumber, order=4, centre=TRUE)
movfor <- forecast(movnumber)
plot(movfor)

movfor
##      Point Forecast    Lo 80    Hi 80    Lo 95    Hi 95
## 2017       2806.629 2743.610 2869.648 2710.249 2903.009
## 2018       3121.008 2980.103 3261.912 2905.513 3336.502
## 2019       3435.386 3199.612 3671.161 3074.800 3795.973
## 2020       3749.765 3404.627 4094.903 3221.922 4277.608
## 2021       4064.144 3596.826 4531.462 3349.442 4778.845
## 2022       4378.523 3777.416 4979.630 3459.209 5297.836
## 2023       4692.901 3947.320 5438.483 3552.634 5833.169
## 2024       5007.280 4107.273 5907.287 3630.838 6383.722
## 2025       5321.659 4257.876 6385.442 3694.743 6948.575
## 2026       5636.038 4399.632 6872.443 3745.118 7526.957
accuracy(movfor)
##                    ME     RMSE      MAE       MPE    MAPE      MASE
## Training set 10.76794 44.88976 36.11703 0.8666377 4.46474 0.3516341
##                   ACF1
## Training set 0.5015686

5. Exponential smoothing method

Each year, there seems to be more foreign direct investment projects in Vietnam, and because there is no indication of seasonality, the double exponential smoothing model proves to be a good choice. The graph below shows the forecasted number of project in the next 10 years.

holtnumber <- HoltWinters(tsnumber, gamma=FALSE)
plot(forecast(holtnumber))

forecast(holtnumber)
##      Point Forecast    Lo 80    Hi 80    Lo 95    Hi 95
## 2019       3421.439 3181.950 3660.927 3055.172 3787.705
## 2020       3717.959 3362.902 4073.016 3174.946 4260.971
## 2021       4014.479 3531.524 4497.434 3275.863 4753.095
## 2022       4310.999 3688.575 4933.424 3359.083 5262.916
## 2023       4607.520 3834.808 5380.231 3425.759 5789.280
## 2024       4904.040 3970.887 5837.193 3476.905 6331.175
## 2025       5200.560 4097.381 6303.739 3513.394 6887.727
## 2026       5497.081 4214.783 6779.379 3535.975 7458.186
## 2027       5793.601 4323.518 7263.684 3545.304 8041.899
## 2028       6090.121 4423.962 7756.281 3541.950 8638.292
accuracy(forecast(holtnumber))
##                    ME     RMSE      MAE      MPE     MAPE      MASE
## Training set 35.64347 186.6793 141.0209 1.087835 15.02126 0.9203686
##                     ACF1
## Training set -0.07624917