Investigation and Forecast of the Baltic Dry Index
Financial Econometrics Project 2021
John Clair Øyaas
Copenhagen Business School
MSc Advanced Economics and Finance
August 2021
, John Clair Øyaas – Financial Econometrics final project (COECO1057D)
The Baltic Dry Index
Banks and Financial institutions have for a long time been interested in the ocean-going shipping
industry. In the United Nations review of maritime transport, it was estimate that 80% of the total
trade volume in 2015 was seaborn, in addition containerized trade was shown to have increased by a
15-fold since the beginning of the 21st century (UNCTAD, 2015). The Baltic Dry Index (BDI) is
consequently becoming more important with the increase in maritime transport. The BDI is defined as
the daily weighted average price paid for transporting dry bulk/raw materials overseas. Due to an
increasing amount of products being shipped globally, the BDI index has become viewed as a leading
economic indicator. The index reflects supply and demand for important raw materials used in
manufacturing. As a result, endogeneity complicates its dynamics, the index affects and is affected by
global activity, making it volatile (Papailias, F., Thomakos, D.D. & Liu, J. 2017). Though demanding,
the volatility makes the ability to forecast cycles and freight rates extremely advantageous when
making business decisions.
The dynamics of the dry bulk shipping sector has been investigated before, in literature from Katris &
Kavussanos (2021) and Batrinca and Cojanu (2013). In these papers, they investigate the BDI and
freight drivers, submarkets, pricing functions, and other variables that can be used to forecast the BDI
and freight rates. Investigation shows that following the Box-Jenkins methodology, building an
ARIMA-GARCH, yields good forecasts for indicators of the BDI, such as crude oil (Lam,2013),
however the best are VCEM models using freight rate futures (Chen, 2011).
Considering the findings of earlier research, this paper aims to investigate whether a good forecast of
the Baltic Dry Index can be modeled using an ARIMARCH model. Moreover, the model is limited to
time series data and does not consider the economic drivers behind the index.
Data and preparatory work
The data is retrieved from Bloomberg and consists of weekly closing prices of the Baltic Dry Index. A
total of 1579 weekly observations have been exported, starting from 1991.01.04 until 2021.06.18. The
data is called BDIY.
From the original dataset there is derived one additional data version, for conducting a more robust
analysis. This series consists of a shorter observation period of the BDI. Going from 2010.01.08 to
2021.06.18 with 592 weekly observations, called Short.BDIY. Considering that the aim of this paper is
to create an optimal forecasting model, different specifications to the same data can help to identify the
best forecasting model of the underlying.
Since the data has a change in variance over time, a log transformation is used to stabilize the variance
for both data sets. Furthermore, the data is originally closing prices of the BDI, by taking the log it is
1
Financial Econometrics Project 2021
John Clair Øyaas
Copenhagen Business School
MSc Advanced Economics and Finance
August 2021
, John Clair Øyaas – Financial Econometrics final project (COECO1057D)
The Baltic Dry Index
Banks and Financial institutions have for a long time been interested in the ocean-going shipping
industry. In the United Nations review of maritime transport, it was estimate that 80% of the total
trade volume in 2015 was seaborn, in addition containerized trade was shown to have increased by a
15-fold since the beginning of the 21st century (UNCTAD, 2015). The Baltic Dry Index (BDI) is
consequently becoming more important with the increase in maritime transport. The BDI is defined as
the daily weighted average price paid for transporting dry bulk/raw materials overseas. Due to an
increasing amount of products being shipped globally, the BDI index has become viewed as a leading
economic indicator. The index reflects supply and demand for important raw materials used in
manufacturing. As a result, endogeneity complicates its dynamics, the index affects and is affected by
global activity, making it volatile (Papailias, F., Thomakos, D.D. & Liu, J. 2017). Though demanding,
the volatility makes the ability to forecast cycles and freight rates extremely advantageous when
making business decisions.
The dynamics of the dry bulk shipping sector has been investigated before, in literature from Katris &
Kavussanos (2021) and Batrinca and Cojanu (2013). In these papers, they investigate the BDI and
freight drivers, submarkets, pricing functions, and other variables that can be used to forecast the BDI
and freight rates. Investigation shows that following the Box-Jenkins methodology, building an
ARIMA-GARCH, yields good forecasts for indicators of the BDI, such as crude oil (Lam,2013),
however the best are VCEM models using freight rate futures (Chen, 2011).
Considering the findings of earlier research, this paper aims to investigate whether a good forecast of
the Baltic Dry Index can be modeled using an ARIMARCH model. Moreover, the model is limited to
time series data and does not consider the economic drivers behind the index.
Data and preparatory work
The data is retrieved from Bloomberg and consists of weekly closing prices of the Baltic Dry Index. A
total of 1579 weekly observations have been exported, starting from 1991.01.04 until 2021.06.18. The
data is called BDIY.
From the original dataset there is derived one additional data version, for conducting a more robust
analysis. This series consists of a shorter observation period of the BDI. Going from 2010.01.08 to
2021.06.18 with 592 weekly observations, called Short.BDIY. Considering that the aim of this paper is
to create an optimal forecasting model, different specifications to the same data can help to identify the
best forecasting model of the underlying.
Since the data has a change in variance over time, a log transformation is used to stabilize the variance
for both data sets. Furthermore, the data is originally closing prices of the BDI, by taking the log it is
1
