TOP MaNaGeMeNT cOMPeNsaTION aND PeRFORMaNce IN RussIaN cOMPaNIes

This paper contributes to the research on top management remuneration policy and the way in which it relates to performance in Russian corporations. Following an overview of the evidence presented in previous studies focusing on other markets, the paper presents a new empirical study of pay and performance using self-collected data on 93 Russian public companies for the five-year period between 2009 and 2013. The data on key management personnel compensation has been collected from the companies’ official reports, including annual or financial reports and other stock exchange reports. The data on other financial indicators has been procured from the Bloomberg Professional® database. Using fixed effect models for econometric analysis, we find evidence of a positive relationship between compensation and business performance in Russia, although this is not evident for every performance indicator. Indeed, both short-term and long-term forms of compensation as well as their sum show a positive relationship to return on assets (ROa), and the respective sensitivities of pay to performance are not negligible. however, institutional or state ownership can weaken this sensitivity. The relationship is only evident in companies with no state participation. whereas 33 state-owned companies in the sample exhibit no significant relationship of top management compensation to corporate performance, the accounting profits and ROA of the remaining 60 (private) companies serve as significant determinants of remuneration levels.


Introduction
Executive remuneration 1 policy is one of the key topics that attracts economists' attention when they study governance and economic incentives inside big corporations. Indeed, from year to year, top management pay is at the center of attention in numerous articles: as it was noted already in 1999 in the "handbook of labor economics", since the beginning of the 1990s dozens of papers had been devoted to CEO pay every year (Murphy, 1999). This discussion has only expanded today due to plenty of academic contributions from different countries. However, empirical research on executive compensation generally focuses on developed markets, first and foremost the United States (Oxelheim et al., 2008). Therefore, this paper is aimed to fill the gap in studying top management pay in Russian companies, which has gained almost no attention in previous studies.
The importance of labor remuneration for top management staff is proven by a constant interest in this topic that goes beyond academic studies. For corporates, executive compensation policy is expected mitigate the agency conflict by providing the managers with incentives that are aligned with the other stakeholders' interests. Therefore, it is no coincidence that annual reports increasingly disclose pay figures, even though some managers may want to keep them private. Moreover, these disclosures are nowadays often supported or even required of companies by respective regulations that governments introduce to protect investor interests. Indeed, according to G20 Financial stability Board, since 2010 more than 20 countries have experienced changes in regulations of corporate remuneration policy (Financial Stability Board, 2015). In particular, since 2012 Russian companies with a stock exchange listing are required to disclose compensation figures in their annual reporting, which eventually inspired and enabled this study. For the empirical research, a new dataset on executive compensation in Russia comprised of 93 companies has been manually collected. Including total compensation sums and their breakdown into basic components, it allowed for deriving executive compensation indicators that are comparable between different companies. The data were collected for the years between 2009 and 2013. Many companies started applying the new accounting standards in the reporting for 2011, which eventually also meant two years of retrospective compensation disclosure and helped to form a panel close to a balanced one in terms of compensation data completeness. using this dataset, we tested the relationship between key executive compensation and firm performance indicators. Such indicators included both absolute (monetary) and relative (ratios and returns) measures based on company's financial reporting and market performance, namely accounting returns and profits, Tobin's Q, economic value added, market capitalization and stock return.
Few of the performance proxy variables appeared to be significant in compensation equations for the whole sample. however, subgroup analysis of companies with and without state ownership in the sample has shown that the pay-performance relationship is significant only for the latter category. For private corporations, both short-and long-term compensation demonstrated positive pay-to-performance sensitivity with regard to return on assets. The same was evident for absolute values of the accounting profits and market capitalization, though with less statistical confidence for the latter.

The relation between pay and performance
Research in management compensation can be generally divided into two main categories: (1) relationships between pay and performance and (2) relationships among pay and behaviors (Devers at el., 2007). As follows from the title, this paper is contributes to the first research direction.
Performance serves as an obvious benchmark for every work remuneration -however, the way and degree it is considered in payment schemes can differ tremendously even among similar cases. This problem is often underestimated in the media and public discussions about top management compensation, which tend to focus on the amounts of payment with less attention to the factors it can correspond to. as Jensen and Murphy noted in their seminal study on the pay to performance relationship, "the relentless focus on how much ceOs are paid diverts public attention from the real problemhow ceOs are paid" (Jensen and Murphy, 1990). The question how top executives are paid means mainly whether and by what means the compensation is aligned with business performance.
Compensation for top managers can include different components, namely almost always short-term pay in the form of a salary and annual bonuses, occasionally long-term compensation based on share appreciation, as well as other, usually less substantial components, such as severance pay, pension benefits and perquisites (Goergen and Renneboog, 2011). Whereas these less substantial elements are often disregarded when just the relationship with business performance is studied, some authors manage to examine not only work remuneration but also the managers' income from shareholdings as a proxy for their private goals or incentives, which the other shareholders, in theory, aim to align with their own interests (Jensen & Murphy, 1990). however, the disclosure of managers' stock ownership as a regulation requirement is rather an exception than a rule, which is why most of the studies on countries other than the u.s. and the uK operate only with short-term and long-term compensation as proxies for executives' incentives ( Sarkar and Jafar, 2012).
similarly, performance is a broader term, which leads to different particular company-level indicators in top management compensation studies (see Table 1.1). Not only can different variables proxy for the notion of business performance, but various studies also found evidence of existing positive relationships between executive compensation and value-based (e.g., shareholder wealth or stock return) or accounting-based (accounting profits and return) performance indicators. The studies in the  The idea of a relation between business performance and the remuneration of managers who do not entirely own the company is elaborated in agency theory (the applicability of this theoretical framework to top executives is reviewed in (Murphy, 1999;Bebchuk and Fried, 2003). According to the principle-agent models, the manager's private interests in a firm do not fully coincide with those of the owners. Therefore, in many cases it is more efficient for the owner to provide additional incentives for the manager, so that their incentives are aligned in the end and the manager works to the benefit of the company, i.e. the owner. A common way to do this is to tie executive compensation with company performance. If this is the case, one can expect positive relationship between the two.
In reality, money incentives of shareholders and top managers in corporations cannot be perfectly aligned. Dealing with the agency conflict between those groups by means of corporate governance is associated with unavoidable costs for shareholders. Therefore, studying the relationship between company performance and executive compensation empirically can provide insights into the quality of corporate governance and the severity of agency costs in the sample that is examined. A stronger positive pay-to-performance sensitivity can serve as a signal of better remuneration policy and thus a less severe agency problem (Murphy, 1999).

The notion of pay-to-performance sensitivity
The key notion that is used in management compensation studies of firm performance is pay-(to)-performance sensitivity, or shortly PPs. what is generally understood under PPs is a numerical measure that explains the direction and degree of the correlation or explicit dependence of management compensation with regard to firm performance 1 . however, both the related terms and the relationship itself can be measured differently, which is why particular empirical studies often differ in the way they obtain PPs.
Before reviewing these differences to define the framework of the study below, it is worth underlining one more assumption that stands behind the whole methodology. when speaking about pay-(to)performance sensitivity, we already focus on effects of performance on remuneration and formulate the model accordingly. performance might be a question for further research based on other models, needing more precise data on contracting and payment time periods than researchers have for Russian companies so far 1 .
The first and most important methodical difference with regard to PPS is calculating it versus inferring it statistically. Often compensation contracts and financial instruments used for remuneration (e.g., stock options) allow for computing the "mechanical" built-in sensitivity to performance indicators that they are explicitly tied to (sometimes also called "effective ownership percentage"). This is possible if the researchers have such data available (Conyon andMurphy, 2000;Aggarwal and samwick, 2003). however, this is often not the case in studies on markets other than the usa and the UK -in particular, not in this paper either. Furthermore, a direct calculation of the sensitivity is not applicable to compensation forms which are not explicitly (but still can be implicitly) linked with firm performance. It is reasonable to take these pay forms into account if we consider the relationship between pay and performance as a whole.
In case of statistical inference, the methods used in the previous studies had a consensus about neither the estimator nor using logarithms in the model. concerning the estimator, most of the authors apply the "standard" panel data analysis based, mainly the fixed effects model (the so-called "within" estimator). However, some studies argue that other estimators find their application in compensationperformance regressions, e.g. median regressions or dynamic panel models. a review of estimators used to infer pay-to-performance sensitivity is presented in the  Many authors distinguish between pay-to performance sensitivity (PPs) and elasticity (PPe) as two ways to measure the examined relationship. Under this classification, PPS is an absolute, numerical measure which is derived from a model where pay is measured in money units: Under the assumption that the relationship between pay and performance is presented by (1), β is interpreted as the amount of money that an average executive receives for an additional unit of company performance, i.e. the PPs. as opposed to this, the other approach is given by: Under (2), β is interpreted as the percentage change in compensation related to a 1% change in company performance (such a change is expressed, for instance, in return rates), i.e. the pay-performance elasticity. Murphy (1999) discussed the advantages and drawbacks of PPs and PPe, eventually giving no preference to any of them and using both in the study. Pay and performance in prior studies on developed and emerging markets although studies devoted to the question of compensation and performance already number in the hundreds, so far there is almost nothing said about companies in Russia about this point. The only study on PPS in Russia was conducted on a sample of 26 companies and 3 years, i.e. 78 observations as a whole (Baiburina, Shustrova, 2008, in Russian). Moreover, the only compensation measure used by the authors was scarcely comparable between different firms because this indicator, namely the total key management compensation, was actually defined for a different number of top managers by companies themselves, which eventually was likely to yield very biased estimates.
Nonetheless, the existing body of literature on pay and performance in different countries serves as a methodological basis for this new study. The fact that many other authors found evidence of this relationship in different countries underlies the hypothesis that executive compensation is connected to firm performance in Russia as well. Figure 1 illustrates this evidence by answering the question in which countries there was found a profound evidence of positive pay-to-performance relationship in previous studies (colored blue on the map). although this overview might be still incomplete due to a big number of papers that address the payto-performance relationship, Figure 1.1 shows that evidence for it was found in different parts of the world in countries with obviously different economic systems and standards of living. This is one more reason not to underestimate the likelihood of this phenomena in Russian firms.
at the same time, Russia is conventionally considered an emerging market, which is why the features of such markets (observed in studies on china, India, the Philippines and south africa) as opposed to more developed ones are also of interest with regard to pay and performance. a brief overview of existing research shows that performance tends to be more often significant and significant in more proxy variables in compensation equations for companies from developed markets. Differences become particularly evident in comparative studies: for example, Conyon and He (2011) showed that American executives received substantially higher compensation in 2001-2005 than the Chinese, even after controlling for economic and governance differences between the two countries. It is worth mentioning, however, that chinese manager perquisites (usually not considered in studies) were estimated as 15% to 32% of the whole compensation (Kato and Long, 2006). Meanwhile, such incentive instruments as stock options, already common for top executives in the U.S. market, were virtually not used in china (conyon and he, 2011).  Kato, Long (2006) 937 listed companies in china (1998)(1999)(2000)(2001)(2002) There is a positive relationship between top management compensation and shareholder wealth as well as stock return, but for companies with state ownership it is weak. ROA is insignificant. CEO remuneration is positively related to shareholder wealth in firms without state ownership or with foreign ownership. No relationship to return on sales or stock return revealed. short-term compensations are positively related to total assets, turnover and share prices. The latter are not significant if only the crisis years 2008-2010 are analyzed. In nondynamic panel models ceO pay is positively related to stock return and return on assets for the current or the last year. however, in dynamic models stock returns are no more significant.

Rashid (2013)
94 listed nonfinancial firms from Bangladesh There is a significant positive relationship between return on assets or Tobin's Q and total top management compensation.
Such striking differences in compensation structure may mean that performance can also be linked to pay through different channels. In fact, those studies which were focused on emerging markets found evidence of a positive relationship between executive pay and both accounting and market-based performance measures, such as return on assets (

Data and methods
Sample as many other studies on Russian companies, this one faced a challenge due to low data availability. when studying companies in Russia, one has to deal with at least three big limitations: second, a short history of market economy in Russia generally does not allow to collect any longitudinal compensation data series so far. as in any other country, top management compensation in Russian listed companies is disclosed on yearly basis. Moreover, the most important limitation in this sense is that disclosure of executive compensation is required only since 2011 reporting (due to the introduction of International Financial Reporting standards, shortly IFRs, for listed companies in Russia 1 ). Therefore, inter alia, testing compensation equations in dynamic econometric models is questionable so far because this would mean giving up a substantial part of data for the sake of building lags into the equation.
Finally, collecting relevant data eventually complicates the study because not all indicators are aggregated by research and business databases and not all the companies that theoretically have to disclose information actually do it in practice. Indeed, among approximately 200 companies, only about a half actually followed the standards appropriately so that top management compensation was eventually disclosed, including the retrospective comparable figures.
as a result, collecting all the necessary data made possible to work with a sample of 93 companies that provided information about their top management compensation in the years between 2009 and 2013, with unsubstantial missing parts. all of these were companies in Russian domain, i.e. with their main assets and operations located in Russia, although such companies often register legal entities abroad (e.g., to contract under the common law). In order to study pay-to-performance sensitivity, we used a set of variables that includes compensation sums, different performance indicators (most of which were already mentioned in chapter 1) and control variables.
The most important and unique part of this data is a panel of compensation indicators collected manually especially for this sample, using primarily companies' financial reporting, as well as annual reports, reports to market regulators, IPO prospectuses, corporate websites and business media references to enhance the data completeness and quality. In some rather rare cases, Russian companies described their remuneration policies in detail, distinguishing between all the different pay forms for every top manager in particular. however, in most cases companies only disclosed what the IFRs standards required, namely the total sum of pay for the so-called key management personnel and its breakdown into short-term, share-based, termination and post-employment benefits 2 . The size and composition of the key management personnel are not strictly defined by the reporting standard, however, almost all companies include their executive body (sometimes called management board etc.) either alone or together with the board of directors in this notion. Their size is usually disclosed in reporting forms mentioned above.
however, for purposes of statistic inference, the study needed compensation variables that are comparable between different companies in the sample. To transform the total sums of top management compensation in different forms into such variables, two steps were undertaken. First, respectable sums were refined (via subtraction) from the non-executive directors' compensation because nonexecutive board members are usually paid incomparably less so that no averaging would make sense. Second, a comparable indicator, namely annual compensation per one executive was yielded through division of the total executive compensation sum by the number of executives it was attributed to, on average during the year (to take executive turnover into account): Although averaging the compensation between the CEO and other executive team members is also not perfect as it could be that not all of the compensation forms are attributed to all of the team members (therefore, many previous studies only focused on ceO pay), the data sources leave no other choice. Moreover, taking the top management team into account can also benefit the study and even help find "the missing link between CEO pay and firm performance" (Carpenter & Sanders, 2002).
1. although before this, such disclosures were already a part of public issuer reports required by the regulator (Federal Financial Market service), the common practice of underreporting or non-disclosure of compensation sums in these reports was repeatedly admitted (e.g., in the article Olenkov/оленьков (2006, in Russian). "Раскрытие информации: работа над ошибками" published in акционерное общество: вопросы корпоративного управления Among 93 companies in the sample, the executive team for which the total (net of board) executive compensation sum was given averaged 9.9 members.

Sample description and preliminary analysis
The sample of companies in Russia in this study is fairly heterogeneous. The descriptive statistics of executive and director compensation are presented in Table 2.2. In tables hereinafter, all the monetary indicators are measured in Russian rubles (RUB) for the sake of conformity 1 . The prices are adjusted to the 2009 level by using the Consumer Price Index given by the Federal State Statistics Service, so that simultaneous changes in firm performance and incomes did not affect the inference of pay-toperformance sensitivity (market return rates are net of inflation as well). The sample demonstrated a great variation in compensation sums both for executives and directors (non-executive compensation values are collected for testing hypothesis). Table 2.2 also shows that pay for the non-executive board members is actually much less than the executive pay, the former in total (BCOMtot) is on average less than the latter per person (SLCOM). That is, distinguishing between these groups was actually reasonable. Moreover, though there can obviously be common trends in executive and non-executive compensation, the correlation between BCOMtot and SLCOM is still not absolute (0.467) and allows for further analysis.  share in their ownership structure (mostly a majority share). More than 83% of companies are linked to institutional ownership: however, in the database the state is also understood as "institution" so that all state-owned companies eventually also belong here.
The role of all the variables in the empirical analysis in explained the part 2.3.

Hypotheses and methods
The basic question of the empirical study is whether there is any pay-to-performance sensitivity in Russian companies, and whether it is positive and substantial. To test the relationship between executive compensation and firm performance, the two are placed on the two different sides of the regression equation. In other words, generally the model looks like (4) In this equation, the dollar sign stands for a dependent compensation variable (slcOM, scOM or lcOM), it Perf is a vector of performance variables (possibly with several of them if there might be different relationships at the same time), and it Ω is a vector of other independent variables. Thus, 1 β is a pay-to-performance sensitivity vector.
Obviously not many performance variables may form be a part of the equation at the same time due to economic interdependencies and following multicollinearity problem. This is easy to see on the correlation matrix (Table 2.5). For instance, both ROA with ROE or different profits together may not form the same performance vector because of their close interdependence and basically coinciding economic interpretation. however, return on assets (or return on equity) can be a part of the same equation with any of the last four, value-based performance indicators. It is worth mentioning that many authors even do not interpret Tobin's Q as a performance measure, considering it a proxy for future growth opportunities (Ozhkan, 2011).
Moreover, sales and total assets are not perfect control variables for size since they are also correlated with performance. However, size cannot be ignored as it has almost always proven to be significant in previous research, and even explained much more variance than performance (according to metastudy by Tosi et al., 2000). As sales volume can also be interpreted as a kind of firm performance, total assets were preferred in the following analysis as a size variable.
The key hypothesis tested in chapter 3 by models of the type (4) is the existence of positive and significant pay-to-performance sensitivity. In addition, following the hypotheses with regard to PPS and ownership structure pointed out in the literature review (see 1. 3), we also tested if firms with state or institutional ownership have weaker pay-to-performance sensitivity than the others. The models are estimated on panel data by regressions with fixed effects. This method was preferred to random effects in every case according to the Hausman specification test.

Sensitivity of pay to accounting and value performance
To test pay to performance sensitivity, models of type (4) were used separately for short-term and long-term executive compensation and their sum. As previous studies found diverse evidence on the relationship of these indicators to performance, they are all worth testing statistically. For each of the three executive pay indicators as dependent variables, a number of regressions that differed in independent variables (proxies for performance) were estimated.
For the entire direct compensation, the pay to performance sensitivity has been tested by the models given by the equation (5): where Ta is stands for total assets. The estimation results are summarized in Table 3.1. Most of the performance variables showed no significant relationship to SLCOM. However, return on equity appeared to be very promising in this sense, even under the presence of market capitalization in the same equation. One average, when ROA increases by 10 p.p., a top executive receives at least 6.35 million RUB more (which could be a substantial incentive if the yearly compensation makes about 39 million RUB a whole). Net income and market capitalization also were positively related to SLCOM at 10% confidence level, but the PPS was also less. In particular, an additional billion of profit (while net income average in the sample is about 20 billion RUB) corresponds only to 157.7 thousands RUB of additional executive compensation.
For the long-term compensation, the pay to performance sensitivity has been tested by the models given by the equation (6): The estimation results are summarized in Table 3.2. For the short-term compensation, the pay to performance sensitivity has been tested by the models given by the equation (7): The estimation results are summarized in Table 3.3.
This time, only accounting performance indicators, namely return on assets, return on equity and net income demonstrated a significant positive relationship to the short-term compensation. This stand in line with previous studies, which more often found evidence of accounting than market performance as determinant of short-term compensation, i.e. salaries and bonuses. The PPS figures are comparable with those for SLCOM and LCOM. For example, an increase of 10 p.p. in ROA corresponds to a 2.48 million RUB of additional short-term compensation. For the whole sample, the hypothesis of a positive relationship between executive pay and performance is mainly supported by such performance indicators as ROa and net income. however, the absolute values of pay-to-performance sensitivity are rather modest and not necessarily provide managers with strong incentives. a good criticism of low PPs can be found in the seminar pay and performance study by Jensen and Murphy (1990).

Ownership and PPS
according to the hypotheses stated in chapter 2, it is to be tested if pay-to-performance sensitivity differs for companies with different ownership structure. Using STATE as a dummy and ROA as performance yielded pay-to-performance coefficients that are not significant at 10%. However, INST seemed to actually affect the PPS. Figure 3.1 shows clearly that firms with (broadly defined) institutional ownership have lower pay-to-performance sensitivity. While managers in other firms receive 2.67 million RUB with every additional ROA percentage point, the PPS for firms with "institutional" shareholders is, on average, 2.33 lower, that is, only 0.33 million for the same firms performance change.  For companies with no state ownership, there is more statistical evidence of a positive relationship between pay and performance. It can be observed on more variables, such as all accounting profits.
The values of the PPs are also greater than in the sample as a whole. Indeed, while in companies with no state ownership one received about 72,500 RUB more with every additional billion of market capitalization, the same sensitivity for the whole sample (including also the companies with state ownership) equaled about a half of this, i.e. 37,800 (see Table 3.1).
Testing the basic pay-to-performance hypothesis on the subsample of companies with state ownership yielded absolutely no evidence of the studied relationship. even where the performance variables were significant for the entire sample, in the state-owned subsample the same variables appeared to be far from significance. This corresponds to the findings of Kato and Long (2006) and Firth et al. (2006) for Chinese companies.
Overall, although data provide some evidence for pay-to-performance sensitivity in Russia, this is true only for specific variables (return on assets, net income, market capitalization) and on average only for companies without state ownership. such a picture is similar to those observed in other emerging market where the examined relationship is also sometimes elusive or attributed only specific categories of firms (see 1. 3).

Discussion
The purpose of the paper is to investigate the relationship between remuneration of executive management and business performance of a company. This study is supposedly one of the first ones that consider executive compensation totals in Russian companies (the only previous study we have found was in Russian and is briefly reviewed in Part 1). Therefore, the focus was placed on designing a proper research framework, accurate use of new data and testing the basic hypotheses in this research direction if the data allows for testing.
The data collection process inevitably faced a few limitations that eventually constrained the scale of this research project. whereas the studies designed for the most advanced capital markets could exploit data that has been cumulated for hundreds of companies with decades of observations, examining public companies from Russia imposed constraints due to the relatively short history of Russian companies' listings and, more importantly, the even shorter history of disclosure of management remuneration. certainly, the obligatory application of the IFRs for companies listed in Russia since 2011 enabled the collection of research data to the full extent available, with a sample of 93 companies for the five-year period between 2009 and 2013. This sample allowed for the testing of pay-to-performance sensitivity hypotheses within this scale by means of fixed effect models, the most frequently used instrument in this research direction to date. however, the small number of periods for which data is available so far hindered the use of dynamic econometric models because the testing lag would have reduced the original sample by 20% for every lagged period in the equation, and could thus substantially reduce the explanatory power. Testing the relationship between key executive compensation and company performance indicators yielded the conclusion that this relationship was significant only for companies which were not owned by the state. The weakening effect of state ownership on pay-to-performance sensitivity was previously observed by Kato and Long (2006) and Firth et al. (2006) in the case of Chinese companies. However, the reasons behind this effect have not been diagnosed in remuneration studies so far. With regard to China's public listed companies, Quiang (2003) attributed this to ambiguous principal-agent relationships and weak market discipline, e.g. no takeover threat. Ncube and Maunganidze (2014) argued that corporate governance structures in Zimbabwean parastatal enterprises appeared too fragile to restrain over-compensation of executives. The link between pay and performance in Russia parastatal companies might be weak or absent because of state ownership's hindrance of good governance procedures.
Concerning the private corporations examined in the study, the relationship of executive compensation to business performance appeared to be positive, as one would expect from the economic sense of pay-for-performance as well as the evidence from previous studies. however, corresponding to the major existing evidence from emerging markets, the pay executives receive was associated first and foremost with companies' accounting performance, i.e. indicators based on reporting such as profits or ROa. we attribute this mainly to the moderate use of long-term compensation instruments in Russian public companies. Indeed, even for the sub-sample of 60 private corporations that appeared to pay their managers for performance, 25 of them, in fact, paid literally zero long-term compensation during the five-year period observed (another five companies did not disclose the pay structure at all). among their state-owned peers in the sample, only 11 of 33 companies ever paid a long-term compensation package during this period. This corresponds to the findings of Petrov and Chirkova (2012), who identified only 27 public companies in Russia that had any long-term management incentive scheme by 2009. Paying managers for the company's value performance using long-term incentive instruments is still not common in Russia.