Theoretical and empirical study of the efficiency of cities as scientific production units

The third source of data for this study was the official statistics collected by the Federal State Statistics Service (Rosstat) on a regular basis. From this resource base has been drawn an overview of factors in the development of urban environment, namely demographic situation: population, natural growth and loss rates, fertility and mortality; economic factors (average monthly wages); infrastructure development (amount of work performed by activity "Construction").

The data from all three sources has been pivoted into a single table of cities and their characteristics. It has been cleaned from noise, i.e. errors, missing values and irrelevant information. They were then aggregated at the level of integral and autonomous municipal entities. That is not necessarily city, for example, administratively Troitsk is a part of Moscow metropolitan area, however, it has a status of science city and occupies its own position in the database.

The initial table can be found in the Appendix, it consists of 46 rows representing the number of considered municipal units, and 104 columns -- indices of science and technology development, demographic and economic situation in a city. Data is presented both in absolute values and weighted to a single organization. The posted figures include:

· human resources: number of researchers, their age and scientific degree;

· financial indicators expenditure on research equipment, including detailing the cost and time of operation (age);

· the number and costs of pilot production and experimental work;

· intellectual property: publication and patent activity, including the density of cooperation with external organizations;

· commercialization of science: revenues from research and development, including income documents for the protection of intellectual property, receipts (broken down by sectors), payments (with details of recipients);

· the demographics of the city: population, natural increase and wane;

· the city's economy: the average wage and the volume of construction work (detail by year of account).

3.3 Data Envelopment Analysis - method description

The idea of the method.

Data Envelopment Analysis (DEA) is a methodology of comparative analysis designed to measure activity of complex technical, economic and social systems. In general, the method aims at evaluating the relative efficiency of autonomous units of any organization. In context of this study an organization would be Russian national system of R&D, where units are cities. The theoretical basis of DEA lies in the field of econometrics, systems analysis, multi-channel optimization. The choice of the DEA is justified with its advantages over other (e.g., regression), methods of efficiency evaluation: feasibility of DEA is most recognized in the context of multiple parameters, as with its increasing number dramatically increases computational complexity in other methods. As it was pointed out in the literature review, this method is frequently applied at city level in China, where researchers aim in identifying weak and strong points of urban development to justify distribution of resources by public administration.

The idea of Data Envelopment Analysis consists in defining best practices and then evaluate all samples relatively to it. This idea could be illustrated by drawing production possibilities curve, enveloping the set of results according to available statistics on the performance of constituent entities of the production organization, so-called decision making units (DMUs). The linear programming algorithm finds best practices among the existing DMU and pull it in the evaluation of the performance of other entities. An important feature of DEA is the ability to process a variety of indicators of resources expended versus the results obtained, pivoting them to a single norm of efficiency. The method was developed by Charnes, Cooper and Rhodes in 1987 for evaluating the performance of non-profit foundations (NPO). The challenge thy faced back then was that those foundations had no profit, therefore this ultimate performance indicator was simply irrelevant in case of NPOs. Our case of scientific production encompasses this particularity as well - in contrast to commercial organizations, it is insufficient to measure direct profit from R&D .

An important characteristic of the DEA method is absence of requirements to formalize explicitly production function and interconnection between the performance indicators. The problem is that these patterns are complex and often unknown. Without such limitation, DEA covers a wider range of aspects of organizational activities, than traditional microeconomic models that require production function. DEA also calculates efficiency of each unit, making successful in comparison with regression methods that show average results and overall trend.

Today, the practical application of the method went far beyond the original idea due to the universality of the algorithms used. DEA is applied to analyze the performance of various actors in the organization of production, involved in various activities, measured by different indicators, in different geopolitical and other contexts. For example, there are studies of ecological efficiency of industrial production, there is a predictive analysis of the bank's branches measuring probability of its bankruptcy, and other studies are dedicated to assessment of quality of living conditions in cities, regions and countries.

In addition, DEA allows to determine the sources and level of inefficiency for each indicator of input expended and the output obtained in each DMU (industrial plant, office branch, city, country, etc.). Applicability of this method is ensured by the way how it can be used to enrich the benchmarking, namely by providing additional knowledge about the origins of potential inefficiency of entities, considered as a benchmark according to traditional indicators (profitability). Besides, the practical significance of the approach is that it creates the possibility of productive interaction between analyst and decision maker. Such dialog could be opened on the basis of joint selection of input/output data, variations in resource allocation, etc.

Major concepts of the method.

This section consists of setting out key DEA concepts used in further analysis, and its explanation mainly refers to most influential textbooks on this matter.

a) Productive efficiency, or productivity

The most common method to measure efficiency is the ratio of the result to spent resources. In general, it is calculated by the formula:

Efficiency = Output / Input

In relation to the topic of this work, this formula can be specified as follows:

The efficiency of a city =

In later variations of the DEA expanded to differentiation diverse types of efficiency: technical, price, efficiency of scale and other. This work uses the classical approach based on the principle of Pareto-Koopmans:

- production (the city) is efficient if/when and only if/when none of the input or output parameters cannot be improved without worsening any other input or output parameter;

- production (the city) is inefficient if/when and only if/when it is possible to improve the performance of one or more input or output parameter without worsening others.

b) Input and output data.

Inputs and outputs are the main categories of parameters in the model. As indicated above, the ratio of outputs to inputs constitute the basic understanding of productivity.

For calculation of the total performance score, parameters are summed and weighted so that the final value will fit within the one hundred per cent, i.e., 0 < efficiency score < 1.

An important feature of DEA is ability to process an arbitrary number of inputs and outputs. Still there is an empirical recommendation not to exceed one-third of the sample number, or to follow the rule:

(inputs + outputs) * 3 < N cities.

c) Decision Making Unit (DMU).

In this study DMU is a town. It is an autonomous unit of scientific production and has the capacity to take decisions related to the allocation of resources and prioritization of desired results. An example of such decision making could be a development strategy, now being developed in many cities-candidates for the official status of science city. The performance estimation algorithm is based on the important assumption - homogeneity of decision making units, meaning that comparable use of resource base will certainly lead to comparable achievements. For example, if the Obninsk reaches 40 published articles per year having 460 people hired as research staff, it is assumed to be sufficient for other city to accumulate same quantity of people to get same number of publications.

d) Production Possibility Frontier (PPF).

This approach first came to prominence after the work of John. Farrell, where the effectiveness of the organization was estimated on the basis of existing production capacity. The method is to first determine the production possibility frontier (PPF) according to the available indicators, and then assess the performance of each entity of the organization, by pair-wise comparing with the most productive unit. In the case of science cities this means that each city will be assigned a set of weights that "expose it in the most favorable light" and then the city with weighted scores will be correlated with each other to assess the results of scientific and economic advances relative to invested capital and human resources. This will allow to quantify the notion of "best practices" the impact of cities as the unit of organization of science.

Graphically, the production possibility frontier will be a piecewise linear function, the envelope of the highest coefficients of performance in the space of input and output data. Thus, for a DMU located inside the set of production possibilities, distance to the frontier indicates the degree of their effectiveness: the closer to the border, the closer to the standard. This is due to the fact that all indicators are calculated not in absolute terms, and relative to each other.

For example, in the case of a single resource "domestic expenditures on research and development in the reporting year," and the only result of "the number of applications for protection documents filed in Russia," the production possibility frontier will look like the following:

Fig. 1. The production possibility Frontier

where the abscissa Y -- the output (the number of applications for registration of IP objects, units), the ordinate of C -- input (R&D funds, thousand RUR.), each point -- city, each line - specific function for building a product possibility frontier (PPF):

· DEA-crs: constant return to scale, CCR is the basic model.

· DEA-vrs: variable return to scale BCC -- subject to economies of scale.

· FDH: free disposal hull -- with free accommodation (a minimal set that includes all the productive possibilities that can be obtained from observations.)

e) Slack.

Slack is the situation of inefficiency of a DMU, when either the amount of inputs exceeds the minimum possible amount it takes to achieve current rate of output, or the result data is below the maximum achievable with the resources expended. Mathematically slack is the distance between the PPF and the vector of input and output data of the DMU, i.e. it is always non-negative for any solution subject to the set of valid values.

f) Weight (weight).

The weight factor is assigned to each input and output, it determines the contribution of this parameter to the overall performance of the city. This element of the model deserves further attention, since it is calculated automatically, and the validity of the obtained values is widely discussed in the literature.

With the assumption that DMUs are homogeneous, the same set of weights could be awarded to all units in sample. However, such approach has significant limitations: first, the universal set of weights is difficult to find, because the contribution of each indicator is difficult to assess. Monetary value of produced goods is not always relevant even for commercial enterprises, not to mention other organizational structures. For example, in the situation of evaluation of the impact of science cities, the weight of the result obtained in the form of patent activity, undoubtedly depends on the possibilities of further use of a registered intellectual property right in the value chain. However, such possibility is very difficult to measure. Second, the development strategy and, accordingly, the organization of business processes varies from one DMU to another, for example, a science city can be focused on the development of fundamental science or otherwise applied research. In this case, the priority indicator for the first one would be the statistics of publications, and for the second one - number of pilot launches. This means that the weight ratio of publications in the first case should be much higher than in the second.

Thus, it is necessary to vary the weights of parameters from one DMU to another. The possibility to do so is a strong point of DEA. By solving equations of linear programming, the algorithm assorts values of for each unit to get a set of weights that maximizes its efficiency. There are restrictions: all weights must be greater than some Epsilon (?), which ensures that the index will not be ignored at all, and the total productivity of the subject must belong to the segment [0;1], where 1 means productive work, lying on the production possibility frontier, and 0 signifies zero results.

The flexibility of DEA relative to the assigned weights is not the only advantage but also disadvantage. The downside is the lack of connection between the weight ratio and the intrinsic value of the indicator to which it is attributed. When solving the algebraic problem of maximization, the importance of parameters is not assessed qualitatively. No strategic evaluation of productivity factors is available beforehand. Consequently, the rate of productivity could be maximized due to subjectively insignificant indicators, resulting in higher efficiency rate than expected. On the other hand, the selection of significant variables is the task of the researcher, and if a city can't reach efficiency even at maximizing values of the weights, this suggests either a missing factor in its development model, or unfortunate situation. In other words, DEA works best in cases where there is proper reliable statistics on resources spent and results achieved in each DMU, however, there is uncertainty about the contribution of each indicator to the overall level of performance.

Conclusion

science city technology cluster

To conclude, this study has been devoted to theoretical and empirical study of the efficiency of cities as scientific production units. Theoretical part of the thesis has been based on the literature review which covered the following aspects of the topic: the definitions and context of its appearance, uniqueness of science city as a form of scientifically intense territorial entity; legitimacy of taking city to be the basis of analysis; and the up-to-date research questions and approaches on the studied matter. Empirical part is based on the statistical data on major performance indicators of R&D production. The non-mainstream, yet promising method of Data Envelopment Analysis (DEA) has been used to analyze this data and to test two hypotheses, formulated the following way:

Hypothesis #1. Science cities are more productive than other cities, that don't have such legal status. As a result of the analysis, this hypothesis has been rejected. It was discovered that science cities are distributed evenly within the efficiency ranking of all the samples and form no cluster with specific efficiency rate that would characterize them.

Hypothesis #2. the sources of inefficiency aren't homogeneous among cities. As a result of the analysis, this hypothesis has been validated. The sources of inefficiency among cities were analyzed and no common pattern has been found. Each city has unique configuration of inefficiency sources, however some of them appear more frequently than other. The dissemination of different types of inefficiencies has been analyzed and the probability of its occurring has been calculated. Thus, the ranking of inefficiencies characterizing scientific production in Russia has been obtained.

Besides providing the answers to the formulated hypotheses, Data Envelopment Analysis results in detailed information about resource configuration and its significance to overall efficiency score for each decision-making unit. This data is out of scope of the thesis; however, it has a potential for further case by case analysis, that could become a basis for better informed policymaking, governance and management. Any stakeholder of research and development in Russian cities could profit from this information.

References

1. Lukjanchikov G, Shukin A. From Technoparks to Science Cities // Expert: journal. -- М., 6 December, 2010. -- № 48.

2. Kostunina G, Baronov V. Technoparks in Russia and abroad //Vestnik MGIMO. - 2012. - №. 3.

3. Agirrechu A. Science Cities in Russia: History of Establishment and Development // Moscow, MSU Publishing house - 2009.

4. On the Development Stage of Governmental Policy on Science Cities and directions of its further advancements. // Federal Assembly of Russian Federation // Local Government Policy Committee. Report, Moscow, 2006.

Размещено на Allbest.ru