The Journal of American Business

The Omniscience Model: Lagged Correlation

Dr. Jeffry Haber, Iona College, NY

Patrick Hardiman, Iona College, NY

ABSTRACT

The Omniscience Model is an attempt to harness the nearly limitless flow of real-time financial information with the geometric gains in computing power to predict future stock prices. A previous paper (1) reviewed various mathematical and statistical functions from a variety of fields to determine which would be applicable to stock price prediction. Correlation was one metric selected that had potential. Correlation is the relationship between two streams of data and can be any value between (and including) -1.00 and +1.00. The extremes (-1.00 and +1.00) represent perfect correlation, whereas values close to 0.00 indicate non-correlation. The sign (positive or negative) indicates whether the relationship is direct or inverse (respectively). In a direct relationship as one stream increases the other stream would increase as well. In an inverse relationship as one stream increases the other decreases. In an investing context correlation is important because it represents the inference that can be made by utilizing one stream of data to predict the value (or change in value) of a second stream. The problem with correlation is that it is not investable. Both streams of data are for the same temporal period, which means that by the time you have finished the calculation the ability to invest has passed based on the information provided by the correlation. This paper seeks to develop the concept of lagged correlation – that is, where one stream of data is for period t and is used to calculate the correlation with a second stream of data at period t+x for possible use as a filter in The Omniscience Model. Correlation is a ubiquitous metric in investing, usually used in the context of how a fund or security behaves in relation to an index, benchmark or sector. Many fund sponsors will tout that their fund is “uncorrelated” with stocks, which would position their fund as a risk diversifier for a portfolio. Since correlation is a widely-used statistic, it is not surprising that it has also been widely researched. Most specifications of correlation are for extended periods, such as 10 or 15 years. Studies have shown (2) that correlation calculated over the long-term might not hold in shorter segments within this longer term. The typical calculation is based on returns (as opposed to prices). There is sound statistical reasoning for this, but at the same time it is also potentially hazardous when the data looks like a horizontal (or vertical) line when plotted. (3) As conventionally used, correlation is a descriptive metric, which helps an investor understand how the fund or security reacts, given what has already happened to another fund, security, benchmark or index. To be usable as an investing metric correlation needs to be predictive rather than descriptive. A version of correlation is perfectly suited to this task. Commonly labeled “cross-correlation,” or as we prefer to call it “lagged correlation,” it involves off-setting on stream of data. Peterson states “Linear correlation coefficient as a function of time lag is the “cross-correlation function.” (4) It is easiest to describe correlation with a cause-and-effect narrative description. When it gets cold, people dress warmer. There is a direction of cause-and-effect, first it gets colder then people dress warmer. It doesn’t get cold because people dress warmer. If one data stream was temperature and the other was level of warm clothing warm, we would expect to see a high correlation. Some cause-and-effects do not cause the change that can be observed so close in time. Since weather forecasts are available for the next day, people could anticipate the level of clothing to wear so that cold temperatures are met with warmer clothing with nearly no lag. Rainfall and river flow are two data streams that would seem to have a cause and effect relationship with some lag. (5) Other cause-and-effect relationships that would seem to have a lag include television ads and the sale of a product and tweets about a product and sales (6) and pressure and state indicators and North Sea fish population. (7) While these examples are where a cause-and-effect relationship is a reasonable hypothesis, cause-and-effect is not a requirement to apply correlation or lagged correlation. For instance, if two companies compete in the same industry and are roughly the same size with similar operating characteristics, it would not a stretch to think that are population favors or disfavors their products their stock prices might move in similar fashion (think Pepsi and Coca Cola). The increase in Pepsi’s stock price would not cause the increase in Coca Cola’s stock price, but it is not crazy to think that both stock prices would move in the same direction if they are susceptible to similar market changes. For any two streams of data, one set becomes the predictor and the other the target. The target is offset by some amount of time, call it “x,” so that the predictor at time t will be used as one stream and the target stream will be based on t+x. The lagged correlation model (with a lag of k) is described by (8) When making a standard correlation calculation it does not matter which stream is first and which is second. When calculating lagged correlation, the order does matter.

Full text

American Depository Receipts: An Analysis of the Underlying Stock Returns

Dr. Chih-Chieh (Jason) Chiu, Rider University, Lawrenceville, NJ

ABSTRACT

This paper examines option listing effects on American Depository Receipts and their local counterparts. I analyze a sample of 42 option listings from 15 different countries and obtain negative price effects on the day after option listing. One possible explanation for the findings is that option listing relaxes the short sale constraints for the underlying securities. Do option listings have an impact on the underlying securities? According to Arrow-Debreu uncertainty theory, option listings move an incomplete stock market towards a state of equilibrium by increasing the number of assets to hedge against different states of nature. Black and Scholes (1973), however, see options as redundant assets and price them as such in their renowned options model. Consequently, numerous options models assume independence between stock and options market. Without a concrete theoretical direction, I examine the option listing effects on the underlying American Depository Receipts (1)(ADRs) and their local counterparts in the home country. I find a negative price effect on the first day post option listings. The negative abnormal returns are -0.66% and -0.64% for ADRs and their local counterparts respectively. Empirical research has shown price effects of option introductions on the underlying assets. The milestone paper by Conrad (1989) suggests that option introductions have a permanent effect on stock prices. Using a sample of 96 option listings for the period of 1974 to 1980, she finds option introductions have a positive price effect starting three days before introductions. She attributes the increases to the heightened demand for stocks from traders. The positive price effect from option listing, however, has been dampened by the introduction of options on the S&P 500 futures. Using a sample of 300 option listings from 1973 to 1986, Detemple and Jorion (1990) observe positive price effects and reduced volatility due to option listings. More importantly, they find a decline of positive price effect after the introduction of option on S&P 500 futures in April 1982. Their findings are consistent with market completion hypothesis that options market stabilizes financial markets. Researchers around the world seem to offer support for the findings in U.S. markets. Positive price effects around option listings are documented by Watt, Yadav and Draper (1992) for the U.K., Aitken, Frino and Jarnecic (1996) for Australia and Draper, Mak and Tang (2001) for Hong Kong. In contrast to the previous findings, Sorescu (2000) reports negative price effect for options listings that took place after 1981 that cannot be explained by either market completion hypothesis or regulatory changes. The author expands previous research on option listing effects by covering 2051 listing events from 1973 through 1995. He proposes that the negative price effect is due to option listing lowers the short sale constraint by reducing the cost of short selling that was traditionally done with the underlying securities. Mayhew and Mihov (2004b) test the robustness of Sorescu (2000) results and find negative returns after 1981. The negative trend, however, disappears when the authors use Conrad (1989) event study methodology to calculate the abnormal returns. The authors further investigate the short selling constraint hypothesis by examining trading volumes of stocks and options before and after the listings. They find no significant changes in volume before and after option listings. The lack of findings in Mayhew and Mihov (2004b) implies there is no increase in demand for the underlying securities from the option traders. American Depository Receipts (ADRs) have been excluded from previous studies on option listings because their underlying stocks trade outside the U.S. The fact that ADRs have underlying stocks, and perhaps options, trading in the home market does not necessarily mean that one could hedge the ADRs efficiently. For example, when covering a short position in an ADR with home market options, a trader would be exposed to additional currency risks that could increase the short selling costs. ADR options can be seen as a mechanism that allows for relaxation of the short selling constraints by lowering the short selling costs. My results show that the abnormal returns on the day after option listing are negative for ADRs and their local counterparts, which is consistent with the short sale constraints relaxation hypothesis. The data sample consists of option listings on ADRs from the period of 1981 through 2003. Out of the original 97 option listings, 23 listings have option listing dates fewer than 100 days apart from ADR listing dates and are excluded from the sample as their estimation windows are too narrow for any meaningful analysis. Additional listings were rejected due to data gaps or data unavailability that leaves us with a total of 42 listings in the data sample. The 42 listings represent 15 countries, including Chile, Finland, France, Germany, Ireland, Israel, Italy, Japan, Korean, Netherlands, South Africa, Spain, Sweden, Switzerland and United Kingdom. Companies from UK dominate the sample with 12 listings, followed by Japan with 6, and Netherlands and France each with 4.

Full text

The Dead-Weight of Public Sector Companies in India: A Case Study

Dr. Anuja Gupta, Rutgers University, Camden, NJ

ABSTRACT

The Indian economy followed a socialistic model of economic development from 1947 to 1991, in which the government played a very large role in economic activity in the country. While liberalization started 25 years ago, there is still considerable dead-weight in the economy in the form of persistently loss-making public sector enterprises. Privatization, which was one of the planks of liberalization and a potential solution to this dead weight problem, has not made much progress in India. In this paper, we analyze the case of one such loss making public sector unit (PSU) - Mahanagar Telephone Nigam Limited (MTNL) with a view to gaining insight into the dire financial situation of many such PSUs. Further, we conduct a comparative analysis of MTNL with a private player in the same industry thus highlighting the glaring inefficiencies and poor performance of the PSU. We argue that there is no economic or strategic logic of carrying this dead weight in the economy. The assets tied up in these extremely inefficient companies could be put to much better use in India, a country with growth needs and ambitions. The Indian economy was following a “socialistic” model of government and economic development after its independence in 1947. This led to an economy that was replete with government control, and with direct participation of the government in economic activity through the ever-expanding public sector. In 1991, India faced a balance of payments crisis which led to a bailout by the International Monetary Fund (IMF) (Makhija, 2006). This, of course, came with stipulations to open up the economy, and get rid of excessive regulations. Thus, the process of liberalization started in 1991 with the government working on deregulation, delicensing and privatization. While reasonable progress was made in deregulation and delicensing, very little headway has been made in the area of privatization. As of 2016, there are still 244 public sector units (PSUs) under central government control. Only 14 PSUs have been fully privatized. In this paper, we highlight the dire financial position that many PSUs are in, by examining one such loss making PSU in-depth. We study the public sector telecommunications services provider Mahanagar Telephone Nigam Limited (MTNL). We argue that there is no strategic or economic imperative to continue operating enterprises like MTNL. Privatization is the only way to release the assets under control of these PSUs which can then be put to other productive uses in the economy. In the sections below, we first provide an overview of the current state of the public sector in India. We then present a case study on MTNL and discuss why privatization is the only way forward. In India’s socialistic or planned approach to economic development, five year plans were rolled out to define the objectives that the government would tackle in the five year window. The government was envisaged as playing a large role in setting up public sector enterprises in areas where the private sector would not (due to long gestation periods), could not (owing to the heavy capital requirements), or should not (owing to their strategic importance) participate. Examples of these were areas are infrastructure development, iron and steel, and national defense related industries, respectively. Thus, while there were 5 PSUs in 1951, this number gradually expanded to 244 in 2016 (see Figure 1 for overall growth of PSUs over the years. All Figures are in Appendix 1). Today, PSUs employ 1.4 million people, which is 4.8 % of the total workforce employed in the organized sector in India; they control 25 % of the total productive assets of the country, and contribute 14% to the country’s GDP¹. This is a much larger presence of the public sector than many other market economies such as USA where employment in public sector was a mere 2% of total employment. Many of India’s PSUs operate in areas such as technology, tourism, and financial services that are not of strategic or national importance, and are left open to private players in most open economies. Loss-Making PSUs: In 2016, there were 244 operating PSUs. One third of these (78) were loss-making units. The total losses stood at Rs.28,756 crores ($4.4 billion), which translates into a loss of Rs.118 crores ($18 million) per operating PSU in the past fiscal year. The top ten loss making units account for 80% of these losses. Some of them are Steel Authority of India, BSNL, Air India, MTNL, Hindustan Cables, Hindustan Photo Films Manufacturing Ltd. Many of these are operating in industries where private sector players now dominate the market. Sick PSUs: Further, there are PSUs that have been declared “sick”. A unit is defined as sick if its losses in any financial year are 50% (or larger) than its Net Worth (averaged over the preceding four years). As of 2016, there are 65 sick PSUs that have been referred to the Board for Industrial Financial Reconstruction (BIFR). Although the reasons for their sickness vary, most of them face tough competition from the private sector and are burdened with obsolete plants and machinery, heavy interest on debt, surplus manpower and lack of adequate working capital. Of these 65 cases, only four have been approved for winding up operations. 48 PSUs were approved for revival by the Government or holding companies (till March 2014) envisaging total fund and non-fund based assistance of Rs. 40,937 crores. This comprises of cash assistance of Rs. 10,940 crores and non-cash assistance of Rs. 29,997 crores². Some of the PSUs recommended for revival are Cement Corporation of India, Hindustan antibiotics Ltd, Bengal Chemicals and Pharmaceuticals.

Full text