Effects of Work Choices Act on Rio Tinto Case Study

Effects of Work Choices Act on Rio Tinto - Case Study ExampleTrade unions will find it more(prenominal) difficult to service their existing members, recruit new members and organize industrial action. This legislation also clearly shifts the balance of the power in the employment family more towards the employer. This will be most obvious in businesses with 100 employees or less, where employers will have the right to dismiss workers almost at will. nigh importantly, if offered an AWA, the only real alternative for many employees will be to quit.This entails the creation of the Australian Fair Pay Commission which will be responsible for establishing and vary the federal minimum wage, all other wages specified in awards, and casual pay loadings. The abolition of the no-disadvantage test will be implemented. Its replacement by the Australian Fair Pay and Conditions Standard covering just five basic conditions award rates of pay, maximum ordinary hours of work, annual leave, person al and carers leave, and agnatic leave. These five conditions will be the only award conditions which cannot be varied or excluded in a employment discernment. Further, in the event of the termination of an agreement it is these five minimum conditions that the parties fall back to. There are major changes in the nature and operation of federal awards. With the exception of the five minimum conditions, workplace agreements will replace all award conditions that otherwise would apply to the workers covered by these agreements. Australian Workplace Agreements, or AWAs will also take precession over collective agreements, even if those collective agreements are currently in operation. A much simpler approval process for agreements, with all agreements, both collective and individual, call for only to be lodged with the Office of the Employment Advocate (OEA) to take effect. There are tighter rules governing rights of entry by trade unions to workplaces. POSITIVE OPTIONS BY RIO TIN TO focussing The management can fire employees who are members of the Construction, Forestry Mining and Energy Union (CFMEU) and replace them with non-union workers and immigrant workers. The Management can establish a black list of critical, infertile or non-performing workers and offer them AWAs. If they refuse it, management can ask them to resign immediately.The Management can enforce a strike ban on the campaign union hence guaranteeing industrial peace. If the workers resort to strike, the company can lay off the workers with the reason that they have affected the economy by crippling the company. The fraternity can request the Labor Minister to declare the strike as illegal. The company can check the collective agreements of the different subsidiaries and decide not to re-negotiate a new agreement with these unions. The company can just offer individual AWAs to the workers. The Management can terminate personnel with medical problems, numerous absences and recurrent poor work performance. AREAS OF OPPORTUNITIES IN LABOR MANAGEMENT The strong anti-union stance of Work Choices Act will enable Rio Tinto Mining to derive financial benefits from the utilization of apprehend resources.(Nigel Munro) Corporate Affairs

Nursing theory Essay Example | Topics and Well Written Essays - 250 words

Nursing theory - Essay ExampleJBI is a research institute located within the University of Adelaide in the region of Southern Australia it is a non put on research area and is a member based organization (JBI, 2012). The purpose of the organization is to work in collaboration with 70 international entities for the expansion and check of health car system and health care services generated throughout the world. These three organizations are very essential for the development and the progress of the profession of nursing. These facilities provide services of research and development to the profession of nursing and these institutes promote the notion of high quality health care services through their research activities. These organizations corroborate even taken the responsibility of bring technological advancements to the health care systems throughout the world in order to provide effective and efficient doer of treatment to patients. All these organizations are very important for the future of the profession of health care and the well being of individuals throughout the

The Ruthless 2010 Winter in America Essay Example | Topics and Well Written Essays - 750 words

The Ruthless 2010 Winter in America - Essay archetypeBeing a resident of New York, I personally hold the opinion that this storm brought with itself the most horrific casualties compared to the many other previous hurricanes met by America.Though every winter, snowstorms manage in creating scenes that require endurance but this year, the drastic havoc created by the cold and bitter winter was actually higher up the threshold of suffering. The storm caused many deaths and historic snowfall totals were created by the extensive landslides. It competes with the other epic storms that previously slammed the States. In snowfall, the Blizzard of 2010 now rivals the Knickerbocker Storm of 1922 for its regard on the Eastern Seaboard of the United States. (Anon. 2010). Thousands of flights were canceled owing to the strong winds. Rail and road service was suspended bringing the public to a halt at many stations in severe cold and deaths were caused by the bitter winter. The storm also dampe ned enthusiasm on Wall Street where trading on Friday was light. U.S. Treasury and bond trading was also dull. (Nichols, 2010). I myself was tired of shoveling for many hours when I was supposed to hurry for the airport because I had to see my father who was in the hospital, but I was poorly exhausted by the snowstorm just like many other helpless people. The bad effects of the storm did not leave the economy safe since it had to struggle a deal once the winter got over, trying to emerge from the collapse. Over a million businesses suffered blackouts. Relentless winter rains following the storms across the region sent the rivers flooding over their banks and left many people dead.

A Monologue for Love Essay Example for Free

A Monologue for Love EssayWhat becomes of the broken hearted? Nothing. You think you know savor then that love turns out to be an egotistical self-involved bastard whos no braver than the house mouse living in your walls waiting til you least await it. Ive made mistakes before. only nothing compares to the ones made with him. The ones made in his arms, his unloving false arms embrace that somehow once made me feel straightaway and made heart be tender. And in that false embrace I made mistakes under false illusions. Illusions that this.. son, this child, could love me and make me feel like a wo reality A woman I am not, I am but a child, a girl who has been broken by the same boy too many times and more. I tried to end it once, twice, and ended it be on the fifth. just now this was all too little too late, as he had touched me and I him.And is it so much to ask that I be loved again? It must have been for once another man loved me A man this time. Not a child, a boy. A ma n. And this man I turned down. I gave him false reason. For the real reason was that I did wish to be with this man but I was too afraid. The fear growing inside of me, a seed planted by a boy. And in the heat of the moment I fell for that man only to wake the morrow to tell him it cannot be. Months later now, I thought I had convinced myself of this false reason. But in a moment of peace, no distractions to cloud my judgment, my heart caught up to my head and told me of the real reason. And now I am dead. The seed has stopped growing. But its venomous flower has already bloomed and done its damage. I am but one of many broken hearted souls with many mistakes done and many more to proceed, to follow. And I ask of you now What should a girl do? When she no longer welcomes love But instead wishes it she had never known?

A Brief History of Economics Essay Example for Free

A Brief History of Economics EssayThrough his use of dialectic materialism, Marx not only changed the history of economic thought, only found great illumination for himself regarding the bonds of hu valet de chambre society. The concept that seems relatively simple in todays complex economic world was utterly r exploitationary in the time of Marx people develop their ideas about the world (and thereby, their ideas for how society should be organized and stratified) establish on the material aspects of their lives.It was an elegant extension of basic Marxist theory capism, according to Marx, is interested in offering naked materialism as a kind of booby prize to make up for the means of convergenceion being taken away from the people. If the people can no longer die for themselves and no longer work at perfecting their birth craft as individuals, then, as Marx correctly deduced, people would require something to validate their work.This became the center of capitalism, as M arx silent it materialism exists as a kind of justification for capitalism, and since materialism has permeated our culture to such an extreme degree, eventually social orders began to revolve around it. Hence, what capitalism serves as the cause of what Marx saw as nothing less than ongoing class warf are. It is arouse to note that Marx believed a violent revolution on the part of the labor was not simply a possible method of changing things, but in truth served as the only method of changing things.This is because materialism was so enter into class structures (which, in turn, was so embedded in power structures), and the only way for society to functionally survive was for it to change state overturned completely. It is also interesting to note the ongoing effect that Marx has had on philosophic thought materialism had previously been dislodged by Descartes and the famous declaration I think, therefore I am as far as most people who pondered these things could conclude, thou ght preceded matter.However, Marx not only brought materialism to the philosophic forefrontthe then-controversial idea that material preceded thoughtbut illustrated the notion that the abstractions of materialism had been concretized by capitalism into purchasable goods. 2 Marx and the Secret Source of Profit Perhaps the most enduring notion of Marxs is the so-called secret of the source of profit under a capitalistic societythough such is Marxs legacy of intellectual thought, a great umpteen people simply accept this notion as reality the source of profit is the surplus value that an employer gains from laborers.After all, the means of production have been taken from the people skilled artisan cobblers have been replaced by factory line workers churning out shoe after shoe after shoe. The m atomic number 53y rescue by employing this assembly-line method of industrial production becomes pure profit for the employer. Interestingly, Marx tethered this to his own theories on circul ating capitalthat is, something that does not last, and is used up in the production of other goods and services, in direct opposition to fixed capital, which is traditionally held for over a year by a business or institution.Marx sapiently deduced that the distinction between these concepts is not only relative, but arbitrary the idea that capital held for 365 days is circulating and that capital held for 366 days is fixed is absurd. However, it allows the proletariat to essentially gloss over their own necessity to the entire institution of capitalism they are led to believe that society is held up by the fixed capital of major investors and their long-term investments.In reality, society is held aloft by the ongoing purchases of the common man (and woman) their disposable income is burned off to provide them a hotshot that capitalism is worth itit turn, their disposable income is used to make the rich richer, as the cliche goes, all the way up the capitalist pyramid. In Marxs view, this is one of many ways that those in power forestall the seemingly unavoidable class war that he advocates those with powerthe purchasing power of the common manare convinced they have none, and are bought off with trinkets.It is interesting to note that this echoes the master/slave morality inversion of Nietzsche. 3 Marginalists and the Economy In the evolution of economic theory, the impact of marginalism cannot be overstated. Once one had accepted the blunt realities of Marxspecifically, that society was organized based on the perceived value of itemsthe logical dubiousness remained how does one quantify the value of an item?Marginalism illustrated the diminishing returns on the marginal utility of resold products, which dramatically impacted analysis of capitalist economy, the focus of economic analysis, and theories of value and distribution. Regarding the analysis of capitalist economy, marginalism helped solidify the supply and demand notion of economics as that of m ainstream economic thought, as opposed to the labor theory of value espoused by Karl Marx.To put it mildly, this has had ongoing economic effects for the last century and a half. The focus of economic analysis shifted accordingly, as marginalism interacted with price theory this allowed economics to view demand curves utilizing marginal rates of substitution as a means of determining not only when, but why a seller is unstrained to relinquish a product for a particular price. Perhaps one of the most lasting effects of marginalism on economic thought has been its effect on theories of value and distribution.With the previous measures in place, economists (as well as common people) now had a more-or-less accurate measure of determining the value of any given product in any given social context. Of course, this also effects the ongoing distribution of certain products to certain demographics in order to maximize profitability. Interestingly, this is one of the mind criticisms of mar ginalism as well that it is a vague pseudoscience whose intent is to maintain not only the economic status quo, but the appearance of the status quo.It is no surprise, then, that Marxists hold out to grapple with marginalist theory they see it as a means of keeping the proletariat in check, and preventing any uprisings from them. 4 Thorstein Veblen Survival of the Fattest Thorstein Veblen, for better or worse, specialized in bringing the human element into economic theory. Specifically, he theorized that the institution of the leisure class was a parasite feeding upon America upon the backs of workers who are actually productive are a class who seek only profit and produce nothing but waste.He essentially created and popularized the notion of conspicuous use as the epitome of this theory that wealthy individuals spend large amounts of money on ostentatious goods whose sole benefit is to publicly display their wealth to the world. Unfortunately, this has become an integral part of the economy although an engineer might weep at the man spending over a hundred thousand dollars for a car, the American economy would be crippled if conspicuous consumption vanished overnight.Hence, the parasite metaphor conspicuous consumption and naked profit helps the wealthy to bloat themselves on the backs of the scummy until the entire enterprise inevitably crumbles under its own weight. Veblen perceived quite clearly that human notions about the world are social constructs created by individuals, and as time went on, those notions would change, necessitating a change in economic thought as well. He makes frequent comparisons to evolution to further this end idle curiosity spurs revolution, innovation spurs conflict between the old guard and the innovators, and an economic Darwinism is born.The advent of industry and technological revolutions merely expanded his original point the parasitic relationship go along unabated, as the innovations of the productive engineer class were inevitably utilized as means of conspicuous consumption and waste on the part of the leisure class. Planes are developed as a result of idle curiosity, for instance, but it is the leisure class that necessitates the invention of first class as a way of displaying their own status in the social hierarchy.

Student Resources Worksheet Essay Example for Free

Student Resources Worksheet EssayStudent resources include a renewing of encouraging sites and tools that ordure be of assistance when completing assignments, connecting to other students, and searching for careers. Complete this table regarding student resources provided by the university. In the first column, identify where the resource can be found. In the second column, summarize each resource in at least one sentence. When you are finished with the matrix, answer the follow-up marvel in part B.Part A Resources Scavenger HuntStudent resourceWhere foundSummary of the resourceSyllabusHome sectionLets you love all of your tasks and assignments and what is required of you to do throughout your whole course Class PoliciesClass HomeTells you the academic and instructor policiesUniversity LibraryLibraryHas a variety of resources and links to help us as students. We can also search for a variety of things. University Academic CatalogProgramContains academis course and policies effective to your program University Learning Goals

Exploring Optimal Levels of Data Filtering

Exploring Optimal Levels of Data FilteringIt is customary to drool raw fiscal information by removing err integrityous observations or outliers before conducting whatever analysis on it. In fact, it is often whizz of the first steps underinterpreted in empirical financial seek to improve the quality of raw information to avoid incorrect conclusions. However, permeateing of financial data screwing be quite complicated not just because of the dependability of the plethora of data sources, complexity of the quoted information and the many contrasting statistical properties of the variables but nigh cruci bothy because of the reason behind the existence of each distinguish outlier in the data. Some outliers may be driven by perfect burdens which bind an economic reason like a merger, take everyplace bid, global financial crises etc. rather than a data error. Under filtering can lead to inclusion of erroneous observations (data error) ca utilise by technical (e.g. compute r system failure) or adult male error (e.g. unintentional human error like typing mistake or intentional human error like producing dummy quotes for testing).1 Likewise, over filtering can also lead to wrong conclusions by deleting outliers motivated by entire compensatets which are important to the analysis. Thus, the question of the right come up of filtering of financial data, albeit subjective, is quite important to improve the conclusions from empirical research. In an attempt to somewhat answer this question, this seminar paper aims to explore the optimum level of data filtering.2The analysis conducted in this paper was on the Xetra Intraday data provided by the University of Mannheim. This time-sorted data for the entire Xetra universe had been extracted from the Deutsche Brse Group. The data consisted of the historical CDAX components that had been collected from Data stream, Bloomberg and CDAX. Bloombergs corporate actions calendar had been used to track dates of initi al offering listing, delisting and ISIN changes of companies. Corporations not covered by Bloomberg had been tracked manually. Even though few basic filters had been applied (for e.g. drop negative observations for circle/depth/volume), some of which were replicated from grocery Microstructure Database File, the data remained largely raw. The variables in the data had been calculated for each day and the data aggregated to daily data points.3The full-page analysis was conducted development the statistical software STATA. The following variables were taken into consideration for the purpose of identifying outliers, as comm precisely done in empirical researchDepth = depth_trade_valueTrading volume = trade_vol_sumQuoted bid-ask spread = quoted_trade_valueEffective bid-ask spread = effective_trade_valueClosing quote midpoint returns, which were calculated by applying Hussain (2011) approachrt = 100*( log (Pt) log (Pt1))Hence, closing_quote_midpoint_rlg = 100*log(closing_quote_mi dpoint(n)) log(closing_quote_midpoint(n-1)). Where closing_quote_midpoint = (closing_ask_price+ closing_bid_price)/2Our sample consisted of the first fifteen hundred and ninety five observations, out of which two hundred observations were outliers. Only the first two hundred outliers were analyzed (on a stress tail chronologically) and categorize as either data errors or extreme point events. These outliers were associated with two companies 313 Music JWP AG and 3U Holding AG. Alternatively, a assorted approach could have been used to consume the sample to include more than companies but the basics of how filters work should be independent of the sample selected for the filter to be free of any biases so for case if a filter is stalwart, it should perform relatively well on any stock or sample. It should be noted that we did not include any nail companies in our sample as those stocks are beyond the scope of this paper. Moreover, since we selected the sample chronological ly on a stock founding, we were able to analyze the impact of these filters more thoroughly on even the non-outlier observations in the sample, which we reckon is an important point to consider when deciding the optimal level of filtering. Our inevitably somewhat subjective commentary of an outlier wasAny observation lying outback(a) the initiative and the 99th percentile of each variable on a stock basisThe idea behind this was to pass on precisely the most extreme values for each variable of engross as an outlier. The reason why the outliers were identified on a per stock basis rather than the whole data was because the data consisted of many different stocks with greatly metamorphoseing levels of each variable of bet for e.g. the 99% percentile of volume for one stock might be seventy thousand trades, while that of another might be three fifty thousand trades and so any observations with lxxx thousand trades in both stocks might be too extreme for the first stock but completely normal for the second one. Hence, if we identified outliers (outside the 1st and the 99th percentile) for each variable of invade on the whole data, we would be ignoring the unique properties of each stock which might result in under or over filtering depending on the properties of the stock in question. An outlier could either be the result of a data error or an extreme event. A data error was delineate using Dacorogna (2008) definitionAn outlier that does not conform to the actual condition of the marketThe ninety four observations in the selected sample with scatty values for any of the variables of interest were also classified as data errors.4 Alternatively, we could have ignored the missing values completely by displace them from the analysis but the reason why they were included in this paper was because if they exist in the data sample, the researcher has to deal with them by deciding whether to consider them as data errors, which are to be removed through fil ters or change them for e.g. to the preceding value and hence it might be of value to see how various filters interact with them. An extreme event was defined asAn outlier backed by economic, social or legal reasons such as a merger, global financial crises, share buyback, major police suit etc.The outliers were identified, classified and analyzed in this paper using the following procedure Firstly, the intraday data was sorted on a stock-date basis. Observations without an instrument lean were dropped. This was followed by creating variables for the 1st and 99th percentile value for each stocks closing quote midpoint returns, depth, trading volume, quoted and effective bid-ask spread and subsequently dummy variables for outliers. Secondly, after pickings the company name and month of the first two hundred outliers, while keeping in consideration a filtering window of about one week, it was check over on Google if these outliers were probably caused by extreme events or the resul t of data errors and classified accordingly using a dummy variable. Thirdly, different filters which are used in financial literature for cleaning data before analysis were applied one by one in the next section and a comparison was made on how well each filter performed i.e. how many verisimilar data errors were filtered out as opposed to outliers probably caused by extreme events. These filters were chosen on the basis of how commonly they are used for cleaning financial data and some of the popular ones were selected.4.1. Rule of ThumbOne of the most widely used systems of filtering is to use some rule of thumb to remove observations that are too extreme to possibly be accurate. Many studies use different rules of thumb, some more arbitrary than others.5 Few of these rules were taken from famous papers on market microstructure and their impact on outliers was analyzed. For e.g.4.1.1. Quoted and Effective outspread FilterIn the paper Market Liquidity and Trading Activity, Ch ordia et al (2000) filter out data by looking at effective and quoted spread to remove observations that they believe are caused by key-punching errors. This manner involved dropping observations withQuoted Spread 5Effective Spread/Quoted spread 4.0% Effective Spread/%Quoted Spread 4.0Quoted Spread/ accomplishment Price 0.4Using the above filters resulted in the identification and consequent dropping of 61.5% of observations classified as equiprobable data errors, whereas no(prenominal) of the observations classified as probable extreme events were filtered out. Thus, these spread filter looks very promising as a reasonably large helping of probable data errors was removed while none of the probable extreme events were dropped. The reason why these filters produced good results was because it looked at the individual values of quoted and effective spread and removed the ones that did not make sense logically rather than just removing values from the tails of the distribution for each variable. It should be noted that these filters removed all the ninety four missing values, which conveys that only five data errors were detected in addition to the detection of all the missing values. If we were to drop all the missing value observations before applying this method, it would have helped filter out only 7.5%6 of probable data errors while not dropping any probable extreme values. Thus, this method stomachs good results and should be included in the data cleaning process. Perhaps, using this filter in conjunction with a logical threshold filter for depth, trading volume and returns might yield optimal results.4.1.2. Absolute Returns FilterResearchers are also known to drop absolute returns if they are above a certain threshold/ return window in the process of data cleaning. This threshold is subjective depending on the distribution of returns, varying from one study to another for e.g. HS use 10% threshold, Chung et al. 25% and Bessembinder 50%.7 In cas e of this paper, we decided to drop (absolute) closing quote midpoint returns 20%. Perhaps, a graphical representation of time series returns of 313music JWP 3U Holding can be used to explain why this particular threshold was chosen.Figure 1. Scatter plot of closing quote midpoint return and dateAs seen in the graph, most of the observations for returns lie between -20% and 20%. However, applying this filter did not yield the best results as only 2.5% of probable data errors were filtered out as opposed to 10.3% probable extreme events from our sample. Therefore, this filter applied in isolation doesnt really seem to hold much value. Perhaps, an improvement to this filter could be achieved by only dropping returns which are extreme but reversed8 within the next few days as this is indicative of data error. For e.g. if T1 return= 5%, T2 return= 21% and T3 return=7%, we can tell that in T3 returns were reversed, indicating that T2 returns might have been the result of a data error. This filter was implemented by only dropping return values 20% which in the next day or two, reverted back to the value of return, +/- 3%9of the day before the outlier occurred as shown belowr(_n) 20%r(n-1) -r(n+1)r(n-1) -r(n+2)Where r(_n) is closing quote midpoint return on any given day. This additional filter seemed to work as it prevented the filtering out of any probable extreme events. However, the percentage of filtered data errors from our sample fell from 2.5% to 1.9%. In conclusion, it makes sense to use this second return filter which accounts for reversals in conjunction with other filters for e.g. spread filter. Perhaps, this method can be further change by using a somewhat more objective range for determining price reversals or an improved algorithm for identifying return reversals.4.1.3. Price FilterWe constructed a price filter inspired by the Brownlees Gallo (2006) approach. The notion behind this filter is to gauge the validity of any transaction price based on its comparative distance to the neighboring prices. An outlier was identified using the following algorithm pi - 3*Where pi is the log of daily transaction price, the reason why logarithmic faulting was used is because the standard deviation method assumes a normal distribution.10 is the stock sorted mean and is the stock sorted standard deviation of log daily prices. The reason why we chose the stock sorted mean and standard deviation was that the range of prices vary greatly in our data set from one stock to another, hence, it made sense to look at each stocks individual price mean as an estimate of neighboring prices. This resulted in filtering 56.5% of probable data errors which were all missing values. Thus, this filter doesnt seem to hold any real value when used in conjunction with a missing value filter. Perhaps, using a better algorithm for identifying the mean price of the nextst neighbors might yield optimal results.4.2. Winsorization and newspaper clippingA very popular filtering method used in financial literature is trimming or winsorization. According to Green Martin (2015a), p. 8, if we want to winsorize the variables of interest at %, we must replace the n largest values by the n upper quantile of the data, and the n smallest values by the n lower quantile of the data. Whereas, if we want to trim the variables of interest by %, we should simply drop observations outside the range of % to 1- %. Thus, winsorization only reduces extreme observations rather than dropping them completely like trimming. For the purpose this paper, both methods will have similar impacts on dropping outliers outside certain %, hence, we will only analyze winsorization in detail. However, winsorization introduces an artificial structure11 to the dataset because instead of dropping outliers it changes them, therefore, if this research was to be taken a step further for e.g. to conduct robust regressions, choosing one method over the other would depend entirely on the kind of research creation conducted.The matter of how much to winsorize the variables, is completely arbitrary,10 however, it is a common practice in empirical finance to winsorize each tail of the distribution at 1% or 0.5%.5 We first winsorized the variables of interest at the 1% level, on a stock basis, which led to limiting 100% of probable extreme events and only 42.9% of probable data errors. Even though intuitively it would make sense for all the identified outliers to be curb because the method used for identifying outliers for each variable considered observations which were either great than the 99th percentile or less(prenominal) than the 1st percentile, and winsorizing the data at the same level should mean that all the outliers would be limited. However, this inconsistency in expectation and end point results from the existence of missing values winsorization only limits the extreme values in the data, overlooking the missing observations which have been inc luded in data errors. We then winsorized the variables of interest at a more stringent level i.e. 0.5%, on a stock basis, which led to 51.3% of the identified data errors and 18.6% of probable extreme events to be limited which doesnt exactly seem ideal as in addition to data errors, quite a large portion of extreme events identified was also filtered out. pickings this analysis a step further, the variables of interest were also winsorized on the whole data (which is also commonly done) as opposed to on a per stock basis, at the 0.5% and 1% level. Winsorizing at the 1% level led to limiting 51% extreme events, 24.2% data errors and an additional one thirty four observations in the sample not identified as outliers. This points toward over filtering. Doing it at the 0.5% level led to limiting 28% extreme events, 12.4% data errors and an additional seven observations in the sample not identified as outliers. Thus, it seems that no matter which level (1% or 0.5%) we winsorize on or wh ether we do it on a per stock basis or on the whole data, a considerable percentage of probable extreme events is filtered out. Of course, our definition of an outlier should also be taken into consideration when analyzing this filter. Winsorizing on a per stock basis does not yield very meaningful results as it clashes with our outlier definition. However, doing it on the whole data should not clash with this definition as we identify outliers outside the 1st and the 99th percentile of each variable on the data as a whole. Regardless, this filter doesnt yield optimal results as a square portion of probable extreme events get filtered out. This is because this technique doesnt define boundaries for the variables logically like the rule of thumb method, rather it inherently assumes that all outliers outside a pre-defined percentile must be evened out and outliers caused by extreme events dont of necessity lie within the defined boundary. It must also be noted that the winsorization filter does not limit missing values which are also classified as data errors in this paper. Thus, our analysis indicates that this filter might be wispy if we are interested in retaining the maximum amount of probable extreme events. Perhaps, using it with an additional filter for limiting missing values might yield a better solution if the researcher is willing to drop probable extreme events for the sake of dropping probable data errors.4.3. Standard Deviations Logarithmic transformationMany financial papers also use a filter based on x times the standard deviationxi + x*xi x* Where xi is any given observation of the variable of interest, is the variable mean and is variable standard deviation.12 An example would be Goodhart and Figliuoli (1991) who use a filter based on four times the standard deviation.13 However, this method assumes a normal distribution, 9 so problems might arise with distributions that are not normal and in our data set, except for returns (because we calculated them using log), the rest of the distributions for depth, trading volume, effective and quoted bid-ask spread are not normally distributed. Therefore, we first log transformed the latter four distributions usingy = log (x)14Where y is the log transformed function and x is the original function. The before and after graphs, using log transformation are shown in Exhibit 4. We then dropped observations for all the log transformed variables that were greater than Mean + x*Standard Deviation or less than Mean x*Standard Deviation, first on a stock basis and then on the whole data for values of x=4 and x=6. Applying this filter at the x=6 level on a stock basis seemed to yield better results than applying it at the x=4 level. This is because x=6 led to dropping 25.6% less probable extreme events for a negligible 3.1% fall in dropping probable data errors. The outcomes are shown in Exhibit 3. However, upon further investigation, we found that 100% of the probable data errors identified by the standard deviation filter at the x=6 level were all missing values. This means that if we dropped all missing values before applying this filter at this level, our results would be very different as this filter would be dropping 7.7% extreme events for no drop in data errors.Applying this filter on the whole data led to the removal of less outlier than applying it on a per stock basis. Using the x=6 level (whole data) appeared to yield the best results 58.4% of probable data errors were filtered out while no probable extreme events were dropped. For more detailed results, refer to Exhibit 3. However, even in this case, 100% of the probable data errors identified were missing values. This means that if we were to drop all missing values before applying this filter, this filter would identify 0% of the probable extreme events or probable data errors. Thus, the question arises if we are really over filtering at this level? If yes, then should xData cleaning is an e xtremely arbitrary process which makes it quite impossible to objectively decide the level of optimal filtering, which is perhaps, the reason behind limited research in this area. This limitation of research in this particular field and inevitably this paper should be noted. That being said, even though some filters chosen were more arbitrary than others, we have made an attempt to objectively analyze the impact of each filter applied. The issue of missing values for any of the variables should be taken into consideration because they are data errors and if we were to ignore them, they would distort our analysis because they interact with the various filters applied. Alternatively, we could have dropped them before offset our analysis, but we dont know if researchers would choose to change them to the closest value for instance or filter them out, therefore, its interesting to see how the filters interact with them.Our analysis indicates that when it comes to the optimal amount of data cleaning, rule of thumb filters furthere better than statistical filters like trimming, winsorization and the standard deviation method. This is because statistical filters assume that any extreme value outside a specified window must be a data error and should be filtered out but as our analysis indicates, extreme events dont necessarily lie within this specified window. On the other hand, rule of thumb filters set logical thresholds, rather than just removing/limiting observations from each tail of the distribution. The outcomes of different filters which are shown in exhibit 1, 2 and 3 are represented graphically below.Figure 2. Box plot of outcomes of all the data cleaning methodsAs shown in section 4.2 and the graph above, Winsorization whether on a stock basis or on the whole data, tends to filter out a large portion of probable extreme events. Thus, it is not a robust filter if we want to retain maximum probable extreme events and should be probably avoided if possible. As far as the standard deviation filter is concerned, as shown in section 4.3, applying it at the x = 6 level, whether on a per stock or whole data basis, seems to perform well but it is not of much value if combined with a missing values filter and all other scenarios tested, actually dropped more probable extreme events than data errors. Therefore, it is not advisable to simply drop outliers existing at the tails of distributions without understanding the cause behind their existence. This leaves us with the rule of thumb filters. We combined the filters that performed optimally spread and additional return filter which accounts for reversals, along with a filter for removing the missing values. This resulted in dropping one hundred and two i.e. 63.4% of all probable data errors without removing any probable extreme events. At this point, a payoff has been made in order to not drop any probable extreme events, we have foregone dropping some extra probable data errors because ove r scrubbing is a serious-minded form of risk.15 This highlights the struggle of optimal data cleaning, because researchers often dont have the time to check the reason behind the occurrence of an outlier, they end up removing probable extreme events in the quest to drop probable data errors. Thus, the researcher has to first determine what optimal filtering really means to him does it mean not dropping any probable extreme events albeit at the expense of keeping some data errors like done in this paper, or does it mean giving precedence to dropping maximum amount of data errors, albeit at the expense of dropping probable extreme events? In the latter case, statistical filters like trimming, winsorization and standard deviation method should also be carefully used.The limitations of this paper should also be recognized. Firstly, only two hundred outliers were analyzed due to time constraint, maybe, future research in the area can look at a larger sample to get more insightful resul ts. Secondly, other variables can also be looked at in addition to depth, volume, spread and returns and more popular filters can be applied and tested on them. Moreover, a different definition can be used to define an outlier or to select the sample for e.g. the two hundred outliers could have been selected randomly or based on their level of extremeness but close attention must be paid to avoid sample biases.Future research in this field should perhaps, also focus on developing more objective filters and method of classifying outliers as probable extreme events. It should also look into the impact of using the above16two approaches of optimal filtering on the results of empirical research for e.g. on robust regressions, to verify which approach of optimal filtering performs the best.Table 1 Outcome of Rule of Thumb Filters AppliedTable 2 Outcome of Winsorization Filters AppliedTable 3 Outcome of Standard Deviation Filters AppliedFigure 3 Kernel distribution before and after log t ransformation3.1 Depth3.2 Effective Spread3.3 Quoted Spread3.4 Volume Figure 4. Kernel Distribution before and after log transformation of transaction price ReferencesBollerslev, T./Hood, B./Huss, J./Pedersen, L. (2016) Risk Everywhere Modeling and Managing Volatility, Duke University, workings Paper, p. 59.Brownlees, C. T/Gallo, M. G. (2006) Financial Econometric Analysis at Ultra-High Frequency Data Handling Concerns. SSRN electronic diary, p. 6Chordia, T./Roll, R./Subrahmanyam, A (2000) Market Liquidity and Trading Activity, SSRN Electronic Journal 5, p. 5Dacorogna, M./Mller U./Nagler R./Olsen R./Pictet, O (1993) A geographical model for the daily and weekly seasonal volatility in the foreign exchange market, Journal of International Money and Finance, p. 83-84Dacorogna, M (2008) An introduction to high-frequency finance, Academic Press, San Diego, p. 85Eckbo, B. E. (2008) Handbook of Empirical Corporate Finance SET, Google Books, p. 172https//books.google.co.uk/books?isbn=0080 559565Falkenberry, T. N. (2002) High Frequency Data Filtering, S3 Amazon,https//s3-us-west-2.amazonaws.com/tick-data-s3/pdf/Tick_Data_Filtering_White_Paper.pdfGoodhart, C./Figliuoli, L. (1991) Every gauzy counts in financial markets, Journal of International Money and Finance 10.1Green, C. G./Martin D. (2015) Diagnosing the Presence of Multivariate Outliers in Fundamental Factor Data using calibrate Robust Mahalanobis Distances. University of Washington, Working paper, p. 2, 8Hussain, S. M (2011) The Intraday Behaviour of Bid-Ask Spreads, Trading Volume and Return Volatility Evidence from DAX30, International Journal of Economics and Finance, p. 2Laurent, A. G. (1963) The Lognormal Distribution and the Translation Method Description and Estimation Problems. Journal of the American Statistical Association, p. 1Leys, C./Klein O./Bernard P./Licata L. (2013) Detecting outliers Do not use standard deviation rough the mean, use absolute deviation around the median, Journal of Experimen tal Social Psychology, p. 764Scharnowski, S. (2016) Extreme Event or Data Error?, Presentation of Seminar Topics (Market Microstructure), Mannheim, PresentationSeo, S. (2006) A Review and Comparison of Methods for Detecting Outliers in Univariate Data Sets, University of Pittsburg, Thesis, p. 6Verousis, T./Gwilym O. (2010) An improved algorithm for cleaning Ultra High-Frequency data, Journal of Derivativ