4.1. Descriptive Analysis of the Italian Plastic Production Sector
As for the economic magnitude of this industrial branch, according to the Italian National Statistic Institution [
31], in Italy in 2019, the entire plastic processing sector involved almost 10,000 enterprises with more than 175,000 employees. The companies also show increasing levels of profitability with a dependence on imports higher than their contribution on exports. There is a clear concentration of production activities in the northern regions of the peninsula, accounting for more than half of the total. The sampled businesses account for about 5 percent of the total number of production sites belonging to NACE code 22.
For the 453 companies included in the final sample obtained after the merge, AIDA data highlighted that in 2018, they employed 35,527 individuals for an average revenue per employee that reached EUR 370.664. Total turnover exceeded EUR 11 million. The companies also displayed a turnover rate of invested capital of 1.14, interpretable as an efficient ability to achieve high levels of revenue by rationally deploying available resources, confirming the dynamism of the sector itself.
The energy consumption levels of these enterprises reiterate the importance of the sector in terms of energy efficiency potential. Electricity constitutes most of the energy uses of the Italian plastic manufacturing sector, followed by thermal and gas consumption, indicating a strong electrification of the entire production system.
The energy savings flows are equally relevant: 70,231.97 toe are the primary savings that can be achieved through EEMs, which are inherent in the implementation of energy self-production interventions (renewable energy production and cogeneration/trigeneration plants) and 13,003.85 toe of final energy savings achievable by deploying all the other EEMs. These stylised facts relate to a period prior to the economic contractions caused by the COVID-19 pandemic, which, together with the current energy crisis hampering production, increased the uncertainty associated with production activities. Uncertainty pertains also to the potential future progression of taxation on plastics (currently applied only to single-use products). More information is included in
Table 7 and
Table 8. In the first table, potential savings are calculated in final energy for all areas except for two of them; for these two areas, involving self-production of energy, potential savings are shown in the second table in primary energy.
Focusing on the different areas of energy efficiency interventions, the values reported in
Table 7 show the high numerosity of interventions concerning the lighting system, which constitute 11.4% of the total, followed by those aimed at improving compressed air (9.5% of the total) and those concerning the general/managerial area (6.4% of the total). Slightly less numerous are instead the EEMs associated with production lines (4.5%), signalling opportunity for companies to review production processes to improve the efficiency in the use of resources, including energy vectors. The terminology of the above areas should not mislead readers, as EEMs could refer to extensive and complex operations as well as to simpler ones. For example, energy efficiency in production lines could be attained through the replacement of a melting furnace with another one with a higher energy efficiency. In contrast, energy efficiency in the compressed air area is often associated with leak detection, which is a relatively simpler form of intervention. The production of energy from renewable sources, on the other hand, includes 171 interventions (6.6% of the total), many of which refer to installing photovoltaic panels, which increases the company’s independence from the purchase of energy flows. Many efficiency measures have adaptable characteristics. For example, heat recovery can also be implemented by referring to compressed air or production lines equipment, making it relevant not only to the thermal power plant/heat recovery area, but also for the others. Therefore, the absence of such measures does not necessarily imply that there are no identified interventions that could affect the above areas.
The replacement of lighting equipment with LED is a widely identifiable and feasible intervention that would lead to the saving of 16,533.7 toe. However, it is necessary to underline how an excessive emphasis on such interventions would prevent the delineation of more efficient technological trajectories pertaining to the production processes; such emphasis is already supported by the relatively low cost of such interventions.
Concerning the extent to which the concept of innovation is extended to the introduction of new organisational forms, the general/managerial area interventions provide a better understanding of the current management of energy flows; by broadening the knowledge base, the search for new rationalisation mechanisms is stimulated. The 166 interventions identified in relation to the introduction of new organisational practices, energy management tools, and the adoption of ISO 50001 certification (general/managerial area) would allow for a reduction in consumption by 16,086.0 toe, confirming that there is large room for improvement when a more organised and careful management of energy sources is implemented within the company.
The greatest source of savings is identified instead in the area pertaining to the installation of cogenerators and trigenerators. In recent years, these systems have been the subject of in-depth studies within the European Union with the objective to identify new ways of achieving energy efficiency in buildings and industry, as they allow for the combined and simultaneous production of electrical and thermal energy, thus increasing the efficiency of installations and optimising energy self-sufficiency. This type of intervention is characterised by its criticality, relatively large scale except for micro-cogenerators, and long technical life and payback period. As a result, such measures may be implemented with resistance, which is exacerbated by the current economic conditions.
However, the implementation of the above measures would lead to a reduction in consumption of 3465.3 tons of oil equivalent for an equally important investment of EUR 12,219,494.44. Similarly, production from renewable energy sources involves a high level of initial investment, equivalent to EUR 25,159,500.59, and relatively long payback periods, which could be shortened using existing incentives. Accordingly, the associated energy savings reach 10,215.2 toe. These EEMs offer significant benefits, not only in terms of reduced consumption, cost reductions, and productivity improvements, but also in terms of competitiveness and the ability of the companies to meet their social and environmental commitments. As highlighted in
Figure 1, electricity flows not only account for most of the consumption in the three NACE sectors considered, but also have the most significant savings potential; in fact, the total energy savings is given by the sum of electrical savings, which accounts for more than half of the total (75%), thermal savings, fuel savings, and other savings.
The observations obtained from the audits also display a concentration of identified interventions related to low levels of investment and savings; overall, the companies subjected to energy audits with an investment of EUR 193,525,069.46 would be able to obtain 13,680.5 toe of primary energy savings, associated with the areas cogeneration/trigeneration and production from renewables, and 70,361.2 toe of final energy savings, relative to all other intervention areas. These savings represent an upper threshold, since not all the EEMs identified in the energy audits will be chosen for implementation.
Furthermore, given the relevance of the cost-effectiveness to business decisions regarding the adoption of EEMs, it is possible to see, as shown in
Figure 2, how most interventions display levels of such indicators around EUR 5001 and 15,000, signalling a positive economic performance for many of the interventions (this comparison excludes those sectors where the financial characteristics of the investments are not comparable on a large scale, such as production of renewable energy and CHP/trigeneration.
Turning our attention to the relationship between energy savings and the payback period of the investments necessary for their realisation, the weight assumed for the total savings by cogeneration/trigeneration measures and those related to the self-production of energy through production from renewables prompted us to exclude them from
Figure 3. In
Figure 3, we can observe a clear concentration of the achievable savings on investments characterised by medium–long payback periods (up to 5 years), which may constitute a potential constraint for the propensity to adopt the efficiency measure and for the realisation of the savings themselves. At the same time, about 26.84% of the potential savings can be achieved through less risky investment options from the point of view of business decision makers and thus characterised by shorter payback periods (up to 3 years).
The regional distribution of the identified EEMs (the regional distribution refers to all energy efficiency interventions identified by the energy audits received by the plastics sector in 2019) (
Figure 4) reflects the location of examined companies in the most industrialised regions of the Peninsula, namely in Northern Italy.
Indeed, the analysis shows a clear concentration of identified interventions in Lombardy, Piedmont, and Veneto. Such EEMs could be implemented or not depending also on the commitment of regional governments to promote the ecological and energy transition and the consequent reduction of environmental impacts. No interventions and therefore savings are found instead for the regions of Calabria and Val D’Aosta, from which no energy audits belonging to the examined sectors were received.
4.2. Econometric Results
This section is dedicated to the discussion of empirical findings obtained through the application of the methodologies described in
Section 2. The approaches employed have confirmed the great potential of the analysis of EA information, combining elements associated with the intrinsic characteristics of companies and information on measures to achieve energy savings, capable of driving the energy performance of the plastics manufacturing sector. Results are shown in
Table 9. In the field of econometrics, the Breusch–Pagan (BP) test is utilised to examine the presence of heteroskedasticity, which refers to the situation when the variance of the residuals is not constant, within a linear regression model. The test is conducted in order to evaluate the null hypothesis of homoskedasticity. In this specific case, the BP test is applied to the pooled OLS model, resulting in a BP value of 42.532, accompanied by a
p-value of 0.01124. Consequently, we reject the null hypothesis of homoskedasticity and reach the conclusion that heteroskedasticity does exist in our pooled OLS model. Considering the potential limitations of the OLS model identified through the BP test, we opt for the adoption of a nonlinear logistic regression model, incorporating a binary response variable. The results also report the mean values of the variance inflation factor (VIF) for the ordinary least squares (OLS) model. These metrics are used to gauge the degree of multicollinearity among the predictor variables. In this particular case, the low VIF values observed (1.460, 1.398, and 1.390) suggest that the variables included in the model exhibit a significant degree of independence from one another. In all models, the potential savings are calculated as primary energy savings for all interventions areas.
First, the coefficients related to the economic dimension of the energy efficiency measures show very encouraging results. All the specifications used led us to observe a significant and positive role of the investment level in influencing the future level of energy performances in the sector of plastic production: the greater the initial expenditure made for the implementation of the measures, the higher the likelihood of businesses experiencing a higher energy performance. In the pooled OLS model without fixed effects, the impact of the investment level on the dependent variable can be interpreted in terms of elasticity, as the logarithm is applied to this predictor (to reduce the impact of observations particularly distant from the mean). Therefore, a one-percent variation in investments is associated with an increase in savings performance equal to 0.63%. As for the OLS model including fixed effects, the elasticity among the two variables concerned reaches 0.87%, while in the logistic regression, if the investment carried out by the production sites increases, the expected percentage change in the probability that the energy-saving performance is above its average is about 187.77%.
Moreover, if the payback period (PBP) increases, the financial concern for the company decision makers increases as well, discouraging the implementation of EEMs. In the OLS without fixed effects, we observe a 0.08% reduction in the dependent variable related to a unit increase in the predictor, while once we control for firms’ characteristics, this percentage reduction is equal to 0.10%. Considering the logistic formulation, the same alteration of PBP leads to a higher expected percentual change of the odds ratio, consisting of 10.62%. Overall, the outcomes relating to PBP are in line with the empirical findings of [
10], confirming a role of the risk level of EEMs, therefore inherent to their different technical nature and in some cases to their economic dimension. Thus, the control effectively acts as a barrier to the adoption of EEMs.
The existing relationship between costs faced by companies and the attainable benefits proves to be a widely explored relation in the literature on energy efficiency, with the objective of reducing the resistance to the adoption of EEMs. This study, using the information extracted from the audits, contributes to empirically demonstrate the beneficial impact, albeit small, that increases in cost-effectiveness produce on energy-saving performances. All the econometric models considered confirm the significance of this parameter.
Moving on to the variables inherent to the firm’s traits, in the OLS without fixed effects and in the logistic regression, it is possible to recognise the role played by the parameter monitoring, which brings a reduction in energy-saving performance (0.44% in the OLS, 191.53% in the odds ratio), although minor with respect to the other company attributes examined. Nevertheless, this does not directly imply that the monitoring of production activities is detrimental to the development of energy efficiency pathways. On the contrary, this could be related to the fact that companies already having a monitoring system have more accurate estimations of the potential savings associated with identified EEMs. For this reason, the saving potential and the performance could be adversely impacted, resulting in a means that is more “realistic”.
The dummy variables representing the geographical location of the firms from which the audits, and so the relative interventions, have been received, assume distinct roles in the models considered. In the linear regression formulation with fixed effects, we clearly see the significance of all the three geographical macro areas. In the northeast, compared to all other locations, including those referring to the central part of Italy used as the reference area, a 4.83% decrease in energy performance is observed. The drop is even higher when considering the southern regions of the peninsula instead. In fact, the south parameter resulted in a 7.7% reduction in the level of energy-saving performance, compared to the other macro regions. Finally, for the north–west variable, the highest relative decrease in the dependent variable was found to be around 9.8%. Alternatively, when examining the model without fixed effects, only one macro area assumes significance, namely the northwest area, while no significance is observed for the geographic areas in the logistic modelling choice.
It is interesting to note that all these variables become significant only when we include a component that allows us to account the fact that the production sites belong to the same company, and thus the existence of common strategies and elements. Nonetheless, the contradictory results obtained by the various econometric models suggest the possibility of refining the approach to better reflect the role of the geographical location of firms and the consequences of their geographical proximity, by relying on spatial econometric methods.
Moving to the variables obtained from the AIDA database, the related parameters assume relevance exclusively for the linear regression models. Regarding the outcomes achieved, including the firm’s fixed effects, we observe that the coefficients associated with the variables debt ratio and capital turnover are statistically significant; in particular, this result indicates that the greater the number of times in which the company can recover the capital invested in management through sales and revenues, the lower the rate of achievable energy-saving performance. Despite the general expectation that firms with higher profitability—and in this case, a higher aptitude for dynamic cash flows—face fewer financial constraints and thus are more inclined to adopt EEMs, the empirical evidence in several contributions, especially for manufacturing SMEs, did not confirm this hypothesis. The analysis detects a clear negative role of capital turnover, indicating how an efficient management is not necessarily linked to increases in energy efficiency: in fact, a percentage reduction of the performance equal to 4.78% has been found for the capital turnover unitarian variation.
Shifting our attention to the impact of the financial liabilities of firms (debt ratio) in the literature, this impact is generally assessed in terms of whether the availability of financing affects energy efficiency pathways, alternatively measured by considering the company’s profitability. Companies encounter numerous challenges related to debt servicing due to a relatively higher share of liabilities, which discourages the adoption of the measures. Regardless, the contributions on the issue once again offer opposing propositions; for instance, the authors of [
32] found that profitability has a deleterious influence on energy efficiency investments, while the authors of [
33] did not find that the debt-to-equity ratio behaves as a real barrier. Concerning the finding obtained here for indebtedness, the variable debt ratio assumes relevance in both OLS approaches, but once again, conflicting results are found. In the modelling choice without fixed effects, a unit increase in liability leads to a slight percentage increment in the dependent variable (0.008%). In the alternative modelling choice, controlling for factors common to the production sites examined, the coefficient of this variable becomes negative, although it remains particularly low (−0.086%). These results suggest that the influence of an increase in the amount of debt-financed assets has a different impact on energy-saving performance depending on the characteristics of the firm.
As for the results inherent to fixed effects themselves (OLS model), since these dummies are constructed by exploiting the site ID code, they cannot be explicitly listed in this analysis as particularly sensitive information would be displayed; therefore, here, we simply confirm that they are significant for the majority of the sites considered, validating the existence of specific features, not controlled by the other predictors, common to sites belonging to the same VAT. These findings open the way for further investigation of the characteristics capable of generating better energy performances.
The intervention areas show different behavioural patterns, although most areas are significant in all the models exhibited. Beginning with the OLS model without fixed effects, we observe how, relative to all the other intervention areas and to the baseline area (represented by lighting, not incorporated in the modelling), cogeneration/trigeneration, thermal power plant/heat recovery and general/managerial assume the highest positive impact on the response parameter. Regarding the first area, such a result is not surprising given the magnitude of the measures themselves. However, it is interesting to observe the role played by the thermal power plant/heat recovery area; in fact, despite gathering 1.4% of the total identified EEMs and corresponding to 5.78% of the potential energy savings, the performance variation associated with its own unit variation is relatively high and equal to 1.11%. With regard to the general/managerial area, to quantify the associated 0.71% increase in energy performance, it is necessary to consider that the total number of interventions in this area would allow us to achieve 22.86% of potential energy savings. Production from renewable sources, similarly to the other areas, also positively affects the response variable. In the case of the fixed-effects OLS model, the findings are reasonably close to the previous ones, while in the logit model, the intervention areas where very interesting behaviours can be observed are production lines, transport, and cold production units. The last two areas account only for 1.15% and 1.69% of the total interventions individuated and are associated with an expected change in odds ratio of about 1173.05% and 344.59%, respectively.
No significance was found, in any of the model specifications considered, for the NPV associated with the energy efficiency interventions and for the per capita revenue of the production sites.
Finally, regarding the incentive policy tool included (White Certificate), its significance is detected only in the OLS formulation (without fixed effects); in particular, the coefficient suggests a performance change of 0.16%. It is, however, necessary to say that these outcomes are influenced by the assumptions underlying the construction of the White Certificate variable, which leads to an underestimation of the interventions incentivised. Moreover, the information content of the variable itself is relatively poor: more information on the economic dimension of these incentives would enrich the analysis. In addition, the literature on the subject still seems far from including instruments reflecting national and/or regional policies, but the results obtained confirm the need to broaden the prospects for the adoption of EEMs.