Data and measures

energy efficiency (H3(a)) and slack (H3(b)) is positive. The key operational lever of the analysis is a vessel’s lifetime. We propose that vessels’ lifetime is negatively associated with their energy efficiency (H2(a)) and slack (H2(b)). Further, drawing from our empirical context and theoreti-cal lens, we posit that this effect is increasing across the range of energy efficiency and slack (H2(c)).

Figure 2.3: Research model

2.5 Method

In this paper, we are interested in the energy efficiency quantified by the EEDI rating. In 2018, 2,106 distinct vessels with EEDI ratings between 1 and 100 were reported in the EU-MRV. We decided to exclude the 17 passenger ships and six ships with electric propulsion from the analysis, as the EEDI benchmark for passenger ships is determined using a different measure for capac-ity (namely gross tonnage), and the main engine’s power for electric ships is defined differently (namely, total electric propulsion). We matched the remaining vessels by their IMO number with the CWFR and TRES databases. Both databases contain detailed information about ship-specific design features for the global fleet. Therefore, we were able to match all but two vessels of interest in the EU-MRV with the data from CWFR and TRES. The quality of the resulting data set is overall very good, except for 206 and 263 missing values for the service speed and main engine’s fuel efficiency variables, respectively.

To address the issue of missing values for the main engine’s fuel efficiency, if available, either computerized engine application system (CEAS) tools or technical specifications provided by the engine manufacturers have been used to obtain the missing fuel efficiency data. We were able to find the required data for most of the main engines and were left with only 19 missing values after this exercise. To derive an approximation of the missing speed values, a different imputation strategy was utilized. The intuition of the approach is as follows: a vessel’s service speed, defined as the average speed under normal load and weather conditions, should be well approximated by the actual recorded average speed of the respective vessel. The recorded average speed is publicly available through the automatic identification system (AIS) data service, which tracks the current speed and position of the world fleet. Hence, as a first step, we collected the average recorded speed by AIS data for all ships in our sample. Then, the service speed was regressed on the col-lected average speed and ship type to allow for ship type-specific intercepts for the full sample in a multivariate regression. The derived regression coefficients were then used to predict the missing service speed variables of our data set. After these additional steps, the final data set consists of 2,058 complete observations.

By collecting and generating this novel data set, we are, to our knowledge, the first to study empirically the impact of technology and operational drivers on environmental performance in the

Table 2.1: Summary statistics

Variable Mean / Share St. Dev. Description

EEDI 7.02 5.69 EEDI value (in gCO2/ capacity-mile)

EERS 0.81 0.30 Slack with EEDI regulation

Age 5.39 3.93 Age of ship (in years)

Speed 16.38 3.58 Average speed under normal load and weather (in knots) Chemical tanker 0.16 1 if ship type is chemical tanker

Container ship 0.16 1 if ship type is container ship

Gas carrier 0.04 1 if ship type is gas carrier

General cargo 0.05 1 if ship type is general cargo ship

Oil tanker 0.21 1 if ship type is oil tanker

Other 0.05 1 if ship type is other ship type

VLS fuel 0.83 1 if ME fuel type is IFO-VLS or IFO-ULS

Alternative fuel 0.01 1 if dual fuel ME for alternative fuels Main efficiency 169.80 3.33 ME-specific fuel consumption (in g/kWh)

Main power 15.24 14.96 ME derived total mechanical propulsion (in thousand kW) Capacity 74.96 58.01 Capacity of ship (in thousand deadweight tonnes)

Draught 12.86 2.81 Draught of ship’s hull (in m)

LOA 220.23 62.29 Length overall of ship (in m)

Beam 35.26 9.55 Width overall of ship (in m)

maritime context. Prior to the EU-MRV data publication, studies examining the EEDI relied mainly on calculated estimates of the rating, as the EEDI for specific ships was not publicly available on a large scale. One notable exemption is the report by T&E (2017) due to their access to the IMO EEDI database. However, they do not include detailed ship-specific features in their empirical analysis. We now describe the specific variables of our econometric analysis in more detail.

The variables of interest are summarized in Table (2.1). Note that all the numerical variables are centered around their respective mean in the forthcoming empirical analysis. Hence, the intercept in our quantile regressions (QRs) has a direct interpretation as the estimated conditional quantile function of a bulk carrier with average dimensions and of average age, capacity, and speed; having

a main engine of average power and fuel consumption; and using heavy fuel oil.

2.5.1.1 Dependent variables

The dependent variable for testing the hypotheses H1(a), H2(a), H2(c), and H3(a) is the attained EEDI rating (EEDI) expressing a vessel’s energy efficiency. As stated previously, we are par-ticularly concerned with the high quantiles that capture vessels with low energy efficiency. For example, to give the reader a specific reference point, the 90% quantile of the unconditional EEDI distribution is around a value of 15.25 grams of CO₂ per capacity-mile. Note that due to the strong positive skew of the unconditional EEDI distribution, we decided to use the logarithm of the dependent variable in the econometric QR model.

To test our hypotheses H1(b), H2(b), H2(c), and H3(b), we formulate a dependent variable indicat-ing a vessel’s slack with the energy efficiency regulation (EERS). Such a measure can be directly derived from Equation (2.2) by replacing a vessel’s required EEDI with the attained EEDI and rearranging terms,

Attained EEDI= (1−Reduction factor

100 )×Reference line value EERS:= (1−Reduction factor

100 ) = Attained EEDI Reference line value.

The procedure is as follows: we first derive the reference line value for each ship in our data set according to ship type-specific formulas (ClassNK, 2015). Then, we divide the attained (absolute) EEDI value by the individual reference line value to attain theEERS variable. Intuitively, a value of 0.65 indicates that a vessel’s attained EEDI is 65% of its reference line value and, thus, has a slack of 35% with respect to the energy efficiency regulation. Note that due to formulating the EERS based on Equation (2.2), it is directly akin to the reduction factors of the EEDI regulation.

For instance, a EERS value of 0.65 also indicates that the ship is compliant with the Phase III reduction target of 30% because its slack is 35%. Another feasible feature of the measure is that it allows us to make more general statements about a vessel’s likelihood of compliance with the energy efficiency regulation. A lowerEERS implies a higher slack and increases a ship’s likelihood of compliance withanyset of uniform reduction factors in the EEDI regulation. Hence, the results

of our analysis are not sensitive to adjustments or changes in the reduction factors. Similar to before, the logarithm of the dependent variable is used in our econometric models.

2.5.1.2 Independent variables

The explanatory variables in our empirical analysis are concerned with the ship design features related to the technology and operational levers potentially impacting technical environmental performance. The measure for a vessel’s lifetime is the variableAge indicating the age of the ship, which is derived from the year the ship was built. The speed of the vessel (Speed) is measured by the vessel’s service speed in knots and is defined as the average speed of the ship under normal load and weather conditions. Note that due to data quality, we use the service speed instead of the design speed, which is the speed at which a ship was designed to operate. These two vari-ables, however, appear to nearly coincide in our data set, which makes this a worthwhile trade-off³.

The ship’s main engine features are measured by the following variables: the main engine’s de-rived total mechanical propulsion (Main power) in thousand kW, the main engine-specific fuel consumption ing/kW h(Main efficiency), and the main engine’s fuel type. Note thatMain power indicates the sum of power generated by the main engine(s) and transferred mechanically to the main propulsor(s), and it thus describes the actual power arriving at the propulsors instead of a nominal power output by the main engine (expressed by the maximum continuous rating), which makes our analysis more relevant to practice. The main engine’s fuel type is divided into three categories: very (ultra) low sulfur fuel (VLS fuel) and the option of using alternative fuels (e.g., LNG, methanol, or ethanol) (Alternative fuel). The omitted category is heavy fuel oil (IFO-380) as the main fuel type, and the fuel type categorical variables are interpreted relative to this category.

2.5.1.3 Control variables

We control for the ship’s dimensions in the empirical analysis. The ship’s dimensions are expressed by the variables length overall (LOA), Beam, and Draught, which are all measured in meters.

While not being directly part of the EEDI formula, the relation of these dimensional variables determines the so-called block coefficient, which is the ratio of a ship’s underwater volume to the

3To illustrate, the pairwise correlation coefficient for original service and design speed in our data set is 0.94, and a simple linear regression (without intercept) suggests a regression coefficient of 0.99.

volume of a rectangular block defined by these three variables. A higher block coefficient leads to a higher hull resistance, and in turn, more power is required for propulsion and lower fuel ef-ficiency. Therefore, controlling for these ship design features in the empirical analysis is imperative.

The cargo-carrying capacity of a ship is reflected by the variable Capacity, which is measured in thousand deadweight tonnes (DWT). DWT states the vessel’s maximum weight-carrying capacity, excluding its own weight, and is in practice indicated by the certified load line marking amidships.

Capacity is an important part of the EEDI formula, as it reflects, together with a ship’s speed, the benefit part of the equation. Further, we control for the type of vessel due to the structural differences in ship design. The ship type is reflected by a categorical variable divided into seven categories: chemical tanker, container ship, gas carrier, general cargo ship, oil tanker, and other ship type, and the omitted categorical level is bulk carrier. We have chosen to follow the classifi-cation in the EU-MRV data set to determine the ship type.