Future outlook of Sri Lankan Textile and Apparel exports: A comparison of Time Series Models

​

Student Name : J.A Rashmi Nishani

Student ID : AS2018470

Group : Standard Deviators

Module : STA 351 2.0 Research Methodology

Lecturer in charge : Dr. Neluka Devpura

​

Objectives

• To forecast the Textile and Apparel exports in Sri Lanka from Aug 2021-July 2022.
• To investigate the accuracy of time series models for forecasting Apparel and Textile exports in Sri Lanka.
• To investigate the effect of COVID 19 on Textile and Apparel exports in Sri Lanka.
• To have a foresight of the textile and apparel exports for successful operations in the industry.

​

Significance of the Study

• There are three primary groups that benefit from this study.

* T&A exporters

* Policymakers

* Investors

Data

​

• Secondary data
• Monthly data on exports(US$. Million) of textile and apparel industry from 2004(January) to 2021(July)
• Data Source : Joint Apparel Association Forum Sri Lanka (JAAFSL).

​

​

T&A Exports in Sri Lanka (2004 January - 2021 July)

​

​

T&A Exports in Sri Lanka (2009 January - 2019 December)

​

​

Methodology

Upward Trend and Seasonality

Seasonal Autoregressive Integrated Moving Average (SARIMA) model

Holt-Winters Exponential Smoothing technique

Additive

Multiplicative

Box-Ljung Test

Box-Ljung Test

Measures of Accuracy

• MAD (Mean Average

Deviation)

​

​

• MAPE (Mean Average

Percentage Error)

​

Data Analysis

Training Set(Model Fitting) : 2009 January - 2017 December

Testing Set(Model accuracy testing) : 2018 January - 2019 December

• Used R programming Language.
• Seasonal Autoregressive Integrated Moving Average (SARIMA) model.
• Seasonal Autoregressive Integrated Moving Average(SARIMA) model is projected only for stationary series.

​

Data Analysis - Original Series

Data Analysis - Original Series

Data Analysis - Log Transformed Series

Data Analysis - Log Transformed Series

Data Analysis - Seasonal Differencing

​

Data Analysis - lag 1 Differencing

SARIMA Models

ARIMA

(0,1,2)(0,1,2)12

Box – Ljung test is performed on the fitted SARIMA models to validate the assumptions of the residuals.

Holt-Winters’ Exponential Smoothing

• Holt Winters’ exponential smoothing method with multiplicative seasonality, provides a model with large AIC and BIC values.
• Variance is stabilized by log transforming.

​

​

​

• Box – Ljung test is performed.
• AIC and BIC values for the training test.

​

​

​

SARIMA & Holt-Winters’ Exponential Smoothing

Key Findings

• Best Model

​

​

​

​

​

​

Holt Winters Exponential smoothing

with α=0. 3466,

β=0.0004 and

γ=0.0001

Key Findings

• Forecasted Textile and Apparel Exports from August 2021 - July 2022

​

Key Findings

• COVID-19 effect on T&A Exports, Sri Lanka

​

Conclusion

• Generally, SARIMA models are more accurate than other time series forecasting methods.
• This study suggested that Holt Winters’ Exponential Smoothing model can be used to provide more accurate forecasts than the obtained ARIMA model.
• Holt Winters Exponential Smoothing models are more feasible and less time consuming than ARIMA models.
• Therefore, Holt Winters model were suggested as the adequate and efficient model to forecast apparel exports in Sri Lanka.

​

Contribution

T & A Exporters

Policy Makers

Further Research

• Latest techniques, machinery and chemicals
• Fair wage for the employees

• Facilitate the processing of raw materials in Sri Lanka

• Investigate the relationship among other factors.
• Provide alternative methods to assess the impact on the exports.

Thank You !