Increase of public education investment and the value of human-capital-intensive firms : evidence from Chile.

(1)

University of Amsterdam, Amsterdam Business School

Master International Finance

Master Research Thesis:

Increase of public education investment and

the value of human-capital-intensive firms:

evidence from Chile

Student: Baeza Silva, Fabian

Supervisor: Matta, Rafael

Amsterdam, September, 2013

(2)

Abstract:

This thesis investigated the effect of human capital investments on firms’ market value. To measure this effect, I studied official announcements of increases in higher education as a proxy of human capital investment and tested the influence of factor on firms’ market price. For this purpose I selected two parametric event studies methods: the Cumulative Abnormal Returns and the Standardized Abnormal Returns. An additional non-parametric test, the Generalized Ranks, was used to check the results. The outcome of the industry-aggregated tests revealed that, in general, announcements of increase in public-education investment do not cause significant changes in the expected market value. This conclusion is similar at firm level. At this level I found some significant cases that correlate well to the announcements release. Both tests are consistent in firm proportion, but these results are too trivial and random to infer that the announcements have some effect on firm market value.

(3)

1 Introduction

During the last years in South America, especially in Chile, Brazil, and Colombia, series of political and social pressures have government officials constantly reviewing and updating the expectation of future public investments. This investigation seeks to find out whether official announcement signaling, particularly, increases in public education investment have a positive effect on the expected market value of

human-capital-intensive firms.

Relevant announcements are considered as the public statements on increases in the expected public education investment rate. Evidence suggests that when this happens, at the same time, the expectation fuels human capital by enhancing workers skills set in the future. Therefore, and based on efficient markets theory, it can be hypothesized that certain firms employing large proportions of human capital (a production factor) can obtain additional financial benefits due to the expected increase in public education investment.

This thesis is backed by several empirical and theoretical researchers in the field of financial economics. Schultz (1965), Becker (1962), Romer (1986), Lucas (1988),

Jagannathan and Wang (1996), Abowd et al (2005), Lev and Radhakrishnan (2005), Lev, Radhakrishnan and Zhang (2009), Le et al (2009) and Le Van et al (2010) among others, argued that when there is positive variation in the levels of specialization, this can increase the levels of human capital. Human-capital-intensive firms become more efficient and enabled to boost production. Consequently, while generating greater outcome levels, these firms create also financial efficiencies, contributing to the creation of extraordinary profits.

To examine this theory, I seek understanding from different angles: this report evaluates official announcements using two event studies methods. These

methodologies estimate the market returns using a linear model. The tests are applied to a group of public firms ordered by their industry codes. The selection of industries is made according to employee quantities, average salaries and other empirical evidence. The proposed event studies tests evaluate returns’ behavior around the dates near to the announcement of future education investments. Following this examination, I

(5)

derive the conclusions about the effect of public education investment announcements on financial returns, especially of human-capital-intensive firms.

The first parts of this paper (1.1 and 1.2) provide us with an introduction to central hypothesis and the main questions to be answered during this investigation. This part presents the reasons or motives to undertake the research of education investment as a value-generating factor. Additionally, in this part I also examine the basics of economic production models that are the pillars of firm’s value.

Pert two provides an overview of the literature used to gain insight in the theoretical, empirical and technical aspects of the investigation relevant events that might affect firms expected market value. Here are reviewed some important contributions to the field of human-capital investment, expected return approximation and event studies. Part three describes the investigation methodologies in detail. The emphasis of this part

is to gain understanding on the determination of abnormal returns using the market.

Furthermore, this part presents the statistical techniques to tests these abnormal returns for significance during a specified period. Here are presented the event studies systems suggested by Fama et al (1969), Patell (1976), Corrado and Zivney (1992), Campbell, Lo and MacKinlay (1997), Benninga (2009), and Kolari and Pynnönen (2010). These methods are formally introduced and applied in this study as well used to test the hypothesis.

Section four presents the statistics of the financial data and analysis of the levels of human capital held by Chilean industries according to average remuneration and employee numbers. The fifth part provides the results of the two parametric event study methodologies, the Cumulative Abnormal Returns (CAR) and Standardized Abnormal Returns (SAR) tests.

To verify the robustness of the previous results, the final part compares the parametric event studies results with a non-parametric test, Generalized Abnormal Returns Ranks (GRANK) and performs these tests on another three industries. After this checks, the paper ends with a critical conclusion on the obtained results, the issues found during the investigation or data collection and recommendation for further research.

(6)

1.1 Increase of public education investment and the value of human

capital-intensive firms: evidence from Chile

In order to understand the effect of education investment on financial returns in the Chilean market, one must look into the financial-economic aspects the firms in the sample and the aggregated industry levels. It is also important to learn how human capital can be decomposed in financial and economic factors such as the employment, salary, production and value levels to then study in depth a redefined sample of firms holding the largest levels of human capital.

To begin this research, I first assume that todays’ firm value is given by the expectation of its future financial performance. And that this performance can be measured in time by the share price levels of the company. Hence, if a firm’s financial results are expected to improve in the future, the value of its shares does efficiently reflect todays’

expectation and increase accordingly. Under this assumption, I would like to pose the main questions that motivates my research, Is there a significant effect in the expected returns of Chilean firms with high dependence on human capital around the days

government announces an increase in next year public education investment? To answer this question, it is necessary to first perform an analysis of the drivers of human-capital expectation.

According to available literature, investment in human-capital is, for example, the increase in efficiency or productivity of labor in a firm given by specialization. In other words, increases in human capital leads to increases in efficiency and productivity. For instance, in a production or services system one can simply decrease the level of waste by shifting to new technologies and processes to save money. In the same way, if X amount of product/services can be obtained in less time due to the increase in

productivity, the firm can reduce the time-cost for certain levels of X and save money as well. Both these increases generate additional marginal costs too, as proven by the studied researchers. The variations in labor specialization are highly correlated with the variations labor cost increase. For this reason, it is assumed that salary levels can used as proxy to measure the increase in human capital.

(7)

Another important logical assumption is that higher education can be used as an approximation for specialization in certain fields. It can be assumed that the financial resources going to external education providers (Universities, Colleagues, other centers of higher education or trainings), or going to internal training system, both with

plausible tax benefits, contribute to the development of a particular labor skill fundamental for a given industry.

In addition to this question, this paper seeks to clarify whether the possible effect of the investment in human capital is immediate or lagged. Therefore, this study should be also answering the question: If there are positive effects of public education investment on firms’ returns, are these immediately reflected in share prices? And should we have these two questions positively answered, I would like to investigate and respond the following query: Which firms’ returns vary significantly to the announcements of public education investment? And finally, is it possible to conclude that, generally, public education investment has a positive effect on Chilean firms’ value?

To investigate the announcements’ effect on returns, I selected two event studies methodologies. These procedures are to identify significance variations in the share price levels due to certain events. Basically, these are statistical methods of analysis of securities’ residuals or abnormal returns. The residuals are calculated by subtracting the market return estimation from the actual return. In this way, the residual

estimations obtained around the announcement day are tested for significance against the rest of year’s business-days.

The days around the announcement are known as the event window. Put another words, the event window represents the range of days with a median equal to day zero, or the day when the announcement takes place. Having this window identified, the procedure then evaluates the variation of returns within this window. Statistically different and positive variation is expected when announcements are positive. This premise follows the principles of the efficient-market hypothesis (EMH), a theory developed by

Professor Eugene Fama in 1960.

According to EMH, all new relevant information is quickly incorporated to the share price of the corresponding company, so there are no available arbitrage opportunities,

(8)

or at least these opportunities do not last much time because are quickly overexploited by arbitrageurs until the market reaches a new equilibrium.

Under this assumption, I expect that an announcement of an increase in public education investment is considered good news for firms that required high human capital levels. Not just because the implication of generating a more prepared and efficient workforce as we already discussed, but also because it could potentially

“reduce the cost of labor”. Dobbs et al, from McKinsey (McKinsey Global Institute, 2012), observed that today and in the future there is a persistent lack of specialized workforce being faced by the developed and the undeveloped economy. Additional human-capital financial benefits can be expected from an entity that expects more resources to be invested in specialization, making human capital premium lower.

For these reasons, I believe these types of announcement are likely to positively affect firms’ valuation, mainly of those firms that require high level of specialized staff or human capital as previously discussed. Subsequently, one can expect higher cash flows due to these two additional factors: (1) the enhancement in employee efficiency leading to savings in working processes and (2) likely reductions in costs of hiring due to

greater availability of asset and savings cost by utilizing less human capital.

Put another way, if the actual workforce acquires better skills and also the quality of education of beginners is improved, the performance of employees would be superior. This superiority would allow a company to save cost, increase production, improve the quality of processes, boost sales and finally generate abnormal returns.

This, I firmly believe, should lead some firms to outperform their peers in output and value, therefore fundamental analysis of human-capital-intensive firms should reflect the improvement in cash flows and this, at the same time, should drive an increase in the firms’ share price, if all other factors remain constant.

1.2 Growth models leading to value aggregation in firms with high levels of

human capital

In order to understand how education investment can be linked to financial value, one must first recognize the underlying relationship between human-capital-output and output-returns. As discussed in coming sections, it is proved that the enhancement of

(9)

human-capital produces a positive effect on output. This effect, subsequently provokes a proportional effect on expected returns because of the decrease of inefficiencies (lower costs) and increase in consume level, making clear that the increase in human-capital, i.e.: specialization of the individual via formal education, does sustain financial value creation.

This section explains how economic growth is among the drivers of firm value. The main argument here is that the sustained intensification of the production factors does naturally lead to the increase of the product quality and price. New efficiencies in any productive systems generate either savings or extraordinary benefits. I am convinced that education investment enhances the human-capital factor and that it provides firms with the potential to outperform peers and that this performance can be measured in means of abnormal financial returns.

Formally, one can look at the classical Solow growth function (Solow, 1956) to better understand this point. This function simply states that economic output is determined by physical and intangible capital:

( , )

Y = f L K Equation 1

where, Y is the real level of output, L is the quantity of homogenous labor and K is absolute the level of invested capital. Naturally, if L is increase one could certainly expect the output Y to increase as well making the firm more efficient and as I will prove in the next section, more profitable.

More recently, the study of Abowd et al (2005) presents a similar production output function in which, however, capital is subdivided into intangible and tangible capital.

( T, I, T)

ij j jt jt jt

y =F K K H Equation 2

Resembling the Solow model, this equation states that production is a function of

human and financial assets, invested physical capital KT_{i.e.: property, plant and}

equipment; intangibles KI_{i.e.: brands, patents, capitalized research and human capital}

HT,_{, attracted to firms by compensation or salary.}

In addition to these theories and pointing at the contribution of human capital to production, Blis and Klenow (2000) derived a similar economic output model where,

(10)

especially the human-capital factor is extended and an individual function for this variable is derived. Blis and Klenow enhanced this model further and included a technology factor.

The theoretic representation of this production model is Equation 3, where Y is the output for a company i at time t, K is the level of invested financial capital, A the level of technology and the factor H is labor. The factors α, represents the proportion of the resource (capital or labor) to be used in the production process. K, at the power α is the proportion of capital to be used for a given output level.

1

( )

it it it it

Y =Kα H A −α Equation 3

In Equation 3, if H increases, holding other things equal, one would expect to see an increase in production attributed to the enhancement of human capital. This, as shown in Equation 3, is the extension of the human capital factor.

= L

it it it

H h Equation 4

In a related analysis, Beyer (2005) interprets the factorHit stating thatLitrepresents

the total labor input, employees and salary, and hitthe level of qualifications necessary

for the execution of the labor.

Each of these model factors is further decomposed and explained by these academics, but this produces non-linear equations that are out of the scope of my research. The models used in this thesis are linear and is not convenient, for a matter of time and resources, to examine the effect of news updates in non-linear variables. For simplicity I focus on a linear regression model, the ordinary least squares method, to determine an adequate market return and measure whether updates on education investment exert some effect on the value of Chilean firms with high levels of human-capital.

Considering that these factors (tangibles, intangibles and human capital) are the fundaments for the determination of economic output, which at an aggregated level equals the gross domestic product (GDP) of a country, it can be argued that as output levels rise, commercialization of products and services also rises providing companies with additional cash flows or abnormal returns if the these same factors are improved

(11)

as it is in the case of specialization of labor via higher education, technical and professionals trainings.

Another proof of the relation of output and financial performance is found in a paper from the International Monetary Fund (Mauro, 2000). This research founds strong correlations between GDP and lagged stock prices. More interestingly, these

correlations are higher in advanced economies with higher market capitalization than in emerging markets, except for the case of Chile and a few other developing economies. Chile, Mauro shows us, has the highest GDP/lagged stock price correlation.

This can be explained by the time relation between present value and current output. Stock price is fundamentally the discounted value of cash flows. If these are expected to grow in line with GDP, the value of firms is higher today. According to Mauro’s study, it can be inferred that higher expected growth due to greater levels of human capital can lead to improved stock returns today.

Therefore, it is not surprising that when studying the correlations between GDP and share prices in the Chilean market I found that almost all industries have positive correlations factors above 0.5, with the exception of the Agricultural sector, the lowest and only below this level (0.3). At the top of the list I found that the Professional and Technical Services sector has correlation of 0.9037, strong enough to imply that expected GDP could be used as a prediction variable for the expected growth of this sector. In addition, I found that the correlation between GDP and stock price do not increase significantly if the lagged as suggested by Mauro.

2 Literature review

The investigation of the hypothesis, besides enhancing a particular interest on the effects of education investment on returns, can contribute to the study of the human capital investment as it attempt to insert an additional and important variable to the estimation of expected output and fundamental firm-value.

For this reason, I reviewed literature which I divided in three different subjects. The first is the theory from diverse economic papers backing the connection between production growth and firm value, and more importantly, the theories to understand

(12)

human capital investment. A second subject is the financial theory on returns analysis. This literature concentrates papers from authors who focused on understanding event studies. Put differently, these studies are used to understand how to measure the effect of relevant economic announcements on firm returns. The last investigated subject is the literature used to understand the education investment issues in Chile.

On the first study subject I found the research of Gary S. Becker (1962). Becker

contended that investment in human capital increases people’s earnings, which at the same time supports higher levels of consumption. From Becker’s work, one can derive the first of a series of implications pointing at human capital as a factor responsible for the positive expectation of firm growth that is fuelled by the increases in revenues. Becker empirical studies clearly indicated a necessary update in the way of analyzing production growth and value of firms.

In addition to Becker, other source of financial-economic literature that helped to set the bases for the analysis of human capital effect on individuals and countries’ income is Theodore W. Schultz (1965). Schultz wrote a paper on economic policy called “Human Capital Investment” where he argued that there is an irrefutable need to understand and measure the human capital factor to correctly understand a production growth

function. In it, Schultz acknowledged the difficulties of suchendeavor mainly because

the lack of data. Companies at those times did not keep registries or logs of such investments. However, the main argumentations of this work were directed to policy changes, which at those times seemed fairly necessary and left aside the need of a formal model to understand the mechanics of human capital as a factor.

Similar to Schultz, in a paper, “Increasing Returns and Long-Run Growth”, Paul Romer (1986) discussed the idea that an increase in human capital, specifically knowledge, can be measured and has a positive effect on returns. Romer elaborated a geometrical and non-dynamic model that includes this measurement of knowledge.

Another contribution to the understanding of human capital is the work of Robert E. Lucas (1988), titled “Mechanics economic growth”. In this paper, one follows-up the evolution of human capital investment and how this factor variance affects the expected economic growth of an economy. Lucas provides empirical evidence showing that, in developing economies, the increase in this investment tent to improve expected returns.

(13)

Jagannathan and Wang (1996), authors of “The Conditional CAPM and the Cross-Section of Expected Returns” to understand a factor models including labor and company size.

This paper conveys that the CAPM equation E(r) =rf + βrp, where is premium return rp =

(rm – rf), in its unconditional form, is not a good estimation for expected returns E(r).

Jagannathan and Wang criticize mainly the beta coefficient (β), which they believe is not supposed to be constant in time. Beta coefficient, they argue, should move as the

business cycle moves. Finally, these researchers explained that the CAPM model’s lack

of explanatory power can be attributed to the market return (rm) because the

determination of this return is done on a proxy portfolio, which according to the view of other investigators, represents only a small fraction of the entire market. The counter argument against the market return continues by saying that the portfolio representing the market returns, typically the NYSE or S&P500 or some other index of publicly traded stocks, is not large enough to be a good proxy for the market returns of the US economy.

Lev and Radhakrishnan (2005) developed a model to estimate organizational capital and provided evidence of the effect of this asset in market returns. This model is extended by the authors by including the saving in operating costs. Their work points out managers as the responsible for creating and maintaining these assets

(organizational capital). Therefore, it is a reflection of the managerial ability. The authors review managerial compensation and its relation to Organizational Capital in the last part of the paper. Moreover, Lev, Radhakrishnan and Zhang argued that

organizational capital, being an asset based on knowledge and information underlying the business operations, is not readily transferable across companies, even over

extended time periods. More importantly, the term organizational capital refers to the ability of firms to deliver and sustain super-normal performance.

The insights of these works allowed me to consider the use of different variables to determine a market prediction model. For instance, this investigation contemplates the use of annual intangible and PPEs investment as variables and included these in an augmented model to calculate expected returns.

In relation to Chilean economy, and to study some of the reasons why the country has gone into a social turmoil that triggered increases in public education investment this

(14)

paper looks at the work of Beyer (2005) who specifically discuss education and the effects of investment in human capital in a developing economy.

To understand the recent impact of investment in human capital on production and financial returns I reviewed the empiric work of Abowd et al (2005), “The relation among human capital, productivity, and market value”. In this article, Abowd et al analyzed two determinants of human capital, individual and firm factor. The individual factor is that the person self develops such as year of experience and level of education. The firm factor is that training and knowledge investment that firms provide to their employees. The authors found that in mainly in two industries, FIRE (Finance,

Insurance and Real Estate) and Manufacturing the correlation with growth are highly significant. In the event study, I use these two industries to first analyze the change in expected share price in relation with announcement in investment education.

The analyses that Abowd et al performed in their research seems very appealing because of it robustness and clear determination of factors that link human capital directly to financial performance. Unfortunately, the collection and analysis of these factors is mostly possible in advanced economies where census data on businesses is readily available. For this case, this data is not available, so only some main notions are adapted and used for this work.

More focus on accounting methods to measure intangible contribution of knowledge to financial returns is “Organization Capital” (2009), a research paper of Lev,

Radhakrishnan and Zhang. The factor “organization capital”, they explain, can be used as a measurement to capture firm’s fundamental ability to generate abnormal

performance. Lev, Radhakrishnan and Zhang use the accounting data from the financial statements to obtain the approximation variables. In their work, the author cite some relevant findings of Nakamura (2000) who estimated that the US investment in

intangibles was about 1trillion, almost equal to investment in PPEs (Property, Plant and Equipment) and who also explained that major growth in intangibles started in the mid 80’s when the “intangible companies” (biotech’s, software and internet companies) started (Nakamura 1999, 2000).

And more recently, Le Van et al (2010) focused in the investigation of new technology and human capital in Asian developing economies. Le Van et al found that there is a time

(15)

when investment in human capital replaces physical capital as the main engine of growth.

On the other hand, I have also found the contrasting conclusions of Fama and Schwert (1977). In their paper, “Human capital and capital market equilibrium”, the authors explain the limitations of considering human capital as a factor to estimate returns. Fama and Schwert found that there exist only a small correlation between the returns on human capital investment and securities expected returns. Therefore, they suggest, there is no clear reason to include this factor into fundamental analysis models.

Along those lines, Pantzalis and Park (1999) reported that small firms with low human capital intangibles outperform companies that have higher levels of this factor.

Both studies, however, are focused on a developed economy, namely the US. My

investigation will focus on emerging economies, with arguably higher return in human-capital investments, in order to see if the link between investments in education and returns is more pronounced in these countries.

Besides these articles on human capital investments, a second source of technical knowledge used in this thesis. The following literature was used to understand the test methodologies and be able to find out whether positive changes in education

investment news drive in some way the expected value of firms.

The main selected testing methodology was event studies. Leading empirical

investigations in this field are parametric and non- parametric tests. Parametric test are those that allow us to make inferences about a sample based on assumptions about the sample statistics such as the distribution of the outcomes. Normally, the parametric study events assume that abnormal returns have a mean or expected value equal to zero and dismiss the chance autocorrelation of the same errors. Fama et al (1969) studied the incorporation of new information to the financial market following the efficient market theory. The test models proposed by Fama et al concentrate the evaluation of the cumulative abnormal returns (CAR). This methodology represents a basic

evaluation method broadly used as a control method for most advanced models.

Another popular parametric methodology is called the Standardized Abnormal Returns (Patell, 1976). The model found in the paper “Corporate Forecasts of Earnings per Share

(16)

and Stock Price Behavior” delivers a student t-statistic to measure the significance of the outcomes. Patell finds that positive forecast of earnings lead to superior standardized abnormal returns (SAR) near to the announcement day. On the same page, the paper also shows that negative and no announcement affect expected returns in a negative and neutral way respectively.

SAR method proved to be efficient when measuring the effect of events news on financial returns. However, critics have pointed out that the methodology can over-reject the null hypothesis of no effect in financial returns because it ignores the firms’ cross-section correlation of the errors. Kolari and Pynnönen (2010), proposed an adjustment to the SAR test statistic to correct for this shortcoming. These researchers proved that the results from the corrected model are more robust and statistically improved.

Moreover, and in order to correctly perform financial analysis and implement the methodologies mechanics, I used contemporaneous technical literature as guidance for this work. “The econometrics of financial markets” (Campbell, Lo and MacKinlay - 1997) contains a detailed section with event studies theory, methodology and empirical

results. The analytic models used in the ordinary abnormal return tests were obtained from this literature. Similarly, Benninga’s (2009) “Financial Modeling” served greatly as a technical reference when studying the mechanics of event studies. Most financial models used in this thesis where replicated using Benninga’s formulations to improve efficiency in calculations of statistics. During this research, I also reviewed several

papers from the Organization for Economic Co-operation and Development (OECD 2010, 2012) to understand the economic position of Chile and the levels money invested in education compared to its European peers.

3 Adopted methodologies

The selected methodology to test the research hypothesis is event studies. This method is used to test the significance of residuals using two alternatives: the cumulative abnormal returns (CAR) and the standardized abnormal returns (SAR). These methods are explained afterward in this section. These tests are also performed later on another

(17)

three different industries to check the prime results’ robustness. These outcomes and benchmarks are provided in section six.

Furthermore, the testing examines only the firm returns of two industries: Financial, Insurance and Real State (FIRE) and Manufacturing (MANU). These industries were selected for their relevant aspects related to human-capital and convenience to compute returns estimation given by the high correlations with the market index.

The first human-capital criterion to consider these industries is the amount of

employees held by each. The two industries represent approximately one third of the Chilean labor force population with each holding about 15 percent of this total, but looking only at the sample, the proportions of FIRE and MANU change to 16 and 20 percent respectively. In addition to this, FIRE and MANU hold the largest number of firms. Together these two groups represent more than 40 percent of the total sample of public firms.

A third consideration is that these two sectors have also the highest levels of correlation with the market index (IGPA). This feature can make the testing of residuals easier and more accurate. By using the market model, the error terms of the return residuals estimations are smaller.

Another last factor supporting the selection of these industries is the results of previous research by Abowd et al (2005) done on the United States market. About et al found that human capital factors in these two industries do fairly well in explaining the financial returns.

3.1 Return estimation using the market model

To obtain abnormal returns per firm, it is necessary to first determine the actual return

of i.e.: firm i at time t. This is calculated rit = ln (sit/sit-1), where ln is the natural logarithm

of today’s share price (sit) divided by previous day share price (sit-1). Then, is needed

estimate the residual or abnormal returns ( rit ). This can be calculated using two

different approximations, the market model (Equation 6) and a multifactor estimation model (Section 3.2). After this, the estimated residual or abnormal return is subtracted from the actual return as shown in Equation 7.1.

(18)

The market model to estimate financial returns uses data from the Chilean general stock market index IGPA. This data, obtained from the Chilean Central Bank official database. Equation 6 states the simple beta model used to estimate returns:

IGPA

it i i t it

r =

α

+

β

r +

ε

Equation 6

Assuming that actual returns are independent and identic random variables with expected zero value and normally distributed through time, the variance of a firm

returns at time t is equal to the total variance of the firm and the covariance of returns is zero. These assumptions are formally shown here below:

[ ]

_it _i for all E r =

µ

t,

[ ]

2

var r_it =

σ

_i for all tand,

[

]

cov r rit, iu =0 for all t≠u

This definition of abnormal returns can be found in numerous financial economic articles where the calculation of abnormal returns can vary. MacKinlay et al (1997) present two methods, the market model as in Equation 6, and the constant mean method, which simply subtracts the fixed average for the returns in a given estimation window returns from the actual return.

For this investigation, I selected the market model to estimate returns because, and in line with MacKinlay, I found that there is no major difference in the results of the test using the market model based on the index IGPA to when calculating these expectations.

Lastly, this model choice is assumed to be improved by controlling for rIGPA_.

The calculation of the abnormal returns ARitis then based on the estimation of the

actual return ritminus the expected return rit as is shown here below:

AR_it = r_it –r_it, Equation 7.1

where rit =

IGPA i _{i t}r

α

+

β

and rtIGPAis the market index return calculated as the with the

actual return. This leads us to Equation 7.2, the model used I this thesis to calculate abnormal returns:

(19)

AR_it = r_it– (

α

i+

β

_ir_tIGPA)

Equation 7.2

In the rest of this paper and when referring to abnormal returns, this equation is used to determine these values for all event studies methodologies.

3.2 Two event studies methods to analyze the effect of education

investment on expected value

To determine whether abnormal returns around the days of the announcement of education increase are significantly different from the normally distributed returns, I used the two parametric event studies procedures and a non-parametric procedure to check the robustness of these calculations. The first test is the Cumulative Abnormal Return (CAR), which is implemented as suggested by MacKinlay (1997). This is a basic but powerful test that can be used as a control method to initially assess whether the abnormal returns are significant on the dates near to the announcement on education investment. The test results are presented in section five and include an industry aggregated test and individual firm test to measure if abnormal returns are statistically superior at lower level.

The second test is the Standardized Abnormal Returns (Patell 1987, Campbell and Lo and MacKinlay 1997, Kolari and Pynnonen 2010). The Standardized Abnormal Returns (SAR) method has proven to be robust in the study of aggregated and individual firm abnormal returns. This test, as applied in this thesis, corrects for the bias of event widow’s omitted variance. The final results are also adjusted with a factor derived from Kolary and Pynnomen (2010). This adjustment corrects the results to for cross-section correlation.

In addition to these tests, and to ensure robustness of the results, this study includes a third non-parametric abnormal returns evaluation method. This is the generalized RANK test (GRANK), derived from Corrado’s RANK test (Corrado and Zivney, 1993) and presented by Kolari and Pynonnen (2010). The results of this test are presented in section six.

The examination abnormal returns around assumes that this variable reflects a firms’ extraordinary performance in the calculation of a firm’s fundamental value. Under this assumption, this thesis analyses share prices accepting this factor appropriately

(20)

reproduce the current value of future cash flows. The event study methodologies in use measure then whether positive announcement generate significant positive variations in share prices around the days of the release.

3.2.1 Event and Estimation Windows

The event window is the time frame to study the possible occurrences of significant abnormal returns. This time span is set to 21 days, where day 0 (zero) represents the day of the announcement, day -1, -2…-10 are the days pre-announcement and day 1, 2…10 are the days post announcement. This event study uses daily share prices of

approximately 180 firms, from the second quarter of 2000 until the last quarter of 2012.

Source: Benninga, Simon, Financial Modeling 3rd_{Edition, 2009}

3.2.2 Cumulative abnormal returns

Event Studies testing is basically a parametric test use to proving that abnormal returns during a given event window are statistically different than the returns captured in the estimation period. I this investigation, the estimation period equal 221 days, a total 249 business days minus the event window of 21 days.

To test the full sample, the cumulated abnormal returns are standardized and aggregated through time and firm. The result of these aggregations is then tested

against the null hypothesis H0: there are no abnormal returns significantly different

than zero. 1 1 N it i AR AR n τ = =

∑

Equation 8.1 T0 T1 T1 + 1 0 T2 T2 + 1 T3

Start date for estimation

window

End date for estimation window

Start date for event window

Event date

End date for event window

Start date for post-event

window

End date for post-event window

ESTIMATION WINDOW The estimation window is used to determine the normal behavior of the stock wrt market factors. Most often we

use the regression Rit= a + bRmtto

determine this "normal" behavior.

EVENT WINDOW

We use data from this window, in conjunction with the a and b of the stock or stocks to

determine whether:

i) The event announcement was anticipated or leaked.

ii) The "post-announcement effect": How long it took for the event information to be absorbed

by the market.

POST-EVENT WINDOW Used to investigate longer-term company

(21)

Similarly, when testing specific industries, this study focuses on the statistical results of individual firms cumulative abnormal returns. Equation 8.2 shows the aggregation of

the event window results from τ1 to τ2 OR T+1 to T+2.

2 1 2 1 2 1 2 ( , ) CAR ( ,_i ) AR_it (0, _i ) t N τ τ τ τ τ τ σ = =

∑

_∼ Equation 8.2

For this test, compare the results of the market model including the news variable and he actual share price delta.

1 2 1 2 1 2 CAR ( , ) SCAR ( , ) ( , ) i i i

τ τ

σ τ τ

= Equation 8.3

H0: there are no abnormal returns on human-capital intensive firms after the

announcement of increase in education investment in the event window.

H1: there are abnormal returns on human-capital intensive firms after the

announcement of increase in education investment in the event window, A final and encouraging self-note for testing this methodology

Since this work is to be ready by the end of September 2013, the model and event study here presented could be tested by the end of the month when the Chilean government announces the investment in education for the coming year 2014.

3.2.3 Standardized Abnormal Returns

The Standardized Abnormal Returns (SAR) test follows a student-t distribution with 2 degree of freedom. The null hypothesis, as with other event studies, states that the mean μ is equal to zero. Therefore, the null hypothesis expectation for this test is no effect of news updated on share prices during the announcement days.

H0: there are no abnormal returns on human-capital intensive firms after the

announcement of increase in education investment.

H1: there are abnormal returns on human-capital intensive firms after the

announcement of increase in education investment.

(22)

The test is constructed by standardizing abnormal returns: AR AR SAR ( 2) (AR ) _i it it P AR i _it t T S

σ

C = = ∼ − Equation 9 (AR )i ARi it S =σ C Equation 9.1

The standardized abnormal returns (Patell, 1976) for individual firms (SARp) are

obtained dividing the abnormal returnsARitby the standard deviation of the abnormal

returns in the event window (-5…0…+5). The term

σ

ARi Cit represents the standard

deviation of non-event days adjusted for the increase in variance that arises because, according to Patell, the estimation of the standard deviation of abnormal returns does not consider the event window.

2 2 1 ( ) 1 1 , where ( ) mt mt it T mt mt t R R C T R R = − = + + −

∑

Equation 9.2 , 1 1 T mt _{IGPA t} t R R T = =

∑

Equation 9.3

To cumulate and test abnormal returns within the event window, Patell suggest adding

the individual ARit in the event period, including the square root of the M events periods

as average denominator. This is shown in Equation 11:

1 2 1 2 t-patell SAR ( 2) 4 T it t m t T m − = −   =   − −  

∑

∼ Equation 10

The news update we are testing occurs generally on the same day for all firms.

Therefore, an additional correction to the standardized abnormal returns is necessary in order to consider cross-section correlation bias.

A t-patell t-patell 1 (n 1)r = + − Equation 11

In this evaluation method, r represents the average of the estimation period ARit. This

because in the sample correlation from which the standardized normal returns is derived does not considers cross sections-correlations of firms within an industry. Put another way, SAR method (1976) does not account for correlations existing within the

(23)

firms when the event day is common for all the firms. However, this bias can be

diminished with 1 (+ n−1)ras shown in Equation 11. This adjustment is suggested by

Kolari and Pynonnen (2010) to account for the bias that can occur if the event is in the same day for all the firms.

4 Data and descriptive statistics

This thesis use financial and economic data from several sources. Share prices returns are the dependent or explainable variable. The data for this variable was extracted from DataStream; a database part of the financial information services of Thomson Reuters Corp. Additional financial data for Chilean public firms is, unfortunately, available only in a yearly frequency. Moreover, relevant human capital data, as for example, number employees, average salaries or paid salaries levels are either very limited or incomplete. Daily and quarterly economic data on gross domestic product and Chilean general stock price index IGPA (índice general de precios) was queried from the Central Bank of Chile’s statistical database. This data is used to determine the market model for the event study.

Education investment data was acquired from the Education Ministry records and also from the National Accounts Controller. This data was collected only for higher education investment (HEI), because, as it is assumed, this is the closest category related to human capital enhancement and the factor considered as public education investment. This variable includes all the public money invested in university scientific research, independent research programmes, scholarship of graduates course post-doctorate course and other allowances related to professional and technical traineeships.

4.1 Financial returns, the dependent variable.

The sample” Share Prices” is a selection of 220 public companies with approximately 3130 transaction days each. The starting point is the second quarter of 2000 until the fourth quarter of 2012. The statistics of this data sample are divided among 10

(24)

Active companies are considered those firms reporting during the full time period under study or at least for the last 3 years (2010 to 2012)

Out of the total sample, 30 entities were removed due to different problems with the data. For instance, many of these firms reported only for a limited period, no longer than a year. For this reason, these firms were eliminated from the selection. In a lesser extent, other companies’ shares price data that did not change during the whole analysis period (highly illiquid) were excluded too. The reminder firms, a total of 190 companies aggregated by industry were analyzed and its statistics are presented Table 1.

In this table, the mean of the returns seems fairly close to zero. This is confirmed by a hypothetical test which revealed the mean of the returns is, in fact, zero. Standard deviation of all industries moves between 0.2 and 0.3 (“Others, no information” is not considered).

FIRE and MANU industry represents more than 50 per cent of the sample, suggesting a practical number to carry out the event study and hypothesis testing. Also, the

correlation of industries return’s and the market index (IGPA) is, as suggested by Abowd, high. As shown in Table 4.2, at a correlation level less or equal to 0.25, the number of companies represents about 8 and 7 per cent of the sample for FIRE and MANU respectively.

Industry Number Mean Meadian Maximun Minimun Standard deviation Agriculture, Forestry, Fishing and Hunting 6 -0.000053 -0.000158 0.081727 -0.088946 0.204617 Construction 6 0.000103 -0.000078 0.079342 -0.092604 0.244497 Finance, Insurance and Real State 50 0.000335 -0.000008 0.091630 -0.078802 0.224516 Government, social and other services 6 0.000218 0.000000 0.186869 -0.139881 0.306242 Manufacturing 53 0.000224 -0.000041 0.087651 -0.092329 0.241524 Mining, Quarrying, and Oil and Gas 6 0.000488 -0.000024 0.145099 -0.128328 0.298247 Others, no information 1 0.000119 0.000000 0.054448 -0.011475 0.067849 Professional, Scientific, and Technical Services 2 0.000338 0.000011 0.075947 -0.062002 0.199913 Retail, Wholesale, Accomodation and Food 23 0.000135 -0.000032 0.072198 -0.065396 0.195994 Transportation and Warehousing 17 0.000233 -0.000063 0.127460 -0.115200 0.274395 Utilities 20 0.000358 -0.000065 0.084304 -0.069093 0.215732 Source: Self-made, data source: DataStream Thomson Reuters

Table 4.1 - Financial Returns Statistics by Industry. This table shows the average statistics for the observed finacial returns between years 2000 to 2012. All industries unders study are included in a sample of 190 companies. Averages were calculated first by year per company, and then the yearly results were averaged by industry.

(25)

4.2 Human capital explanatory variables

This section presents the characteristics of the Chilean labor market under formal contract. These variables are the average salaries and the number of employee per industrial sector. This economic data by industry was obtained from the Chilean tax revenue service (SII) and the Chilean National Institute of Statistics (IND). Statistical financial-economic data such as gross domestic product and market index (IGPA), variables used to compare the relation with human capital factors, was obtained from the Chilean Central Bank database.

From the studied financial theory, it is well know that market variables can help us to predict aggregated economic variables such as those presented here, GDP and Inflation. Alternatively, it can also be argued that the expectation of these economic factors can help us to approximate market levels in some degree. During this investigation I used average salaries and number of employees as factors to approximate the future market value of firms, especially of those firms with high levels of these.

Average correlation (rit , IGPAt) Average All > 25% > 50% > 75% > 25% > 50% > 75% Agriculture, Forestry, Fishing and Hunting 0.094211646 6 16.7% 0.0% 0.0% 0.5% 0.0% 0.0% Construction 0.235652779 6 66.7% 0.0% 0.0% 2.1% 0.0% 0.0% Finance, Insurance and Real State 0.168807511 50 30.0% 8.0% 0.0% 7.9% 2.1% 0.0% Government, social and other services 0.033433831 6 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Manufacturing 0.176992508 53 24.5% 9.4% 0.0% 6.8% 2.6% 0.0% Mining, Quarrying, and Oil and Gas 0.06157249 6 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Others, no information 0.016903719 1 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Professional, Scientific, and Technical Services 0.279584952 2 50.0% 0.0% 0.0% 0.5% 0.0% 0.0% Retail, Wholesale, Accomodation and Food 0.203317883 23 34.8% 13.0% 0.0% 4.2% 1.6% 0.0% Transportation and Warehousing 0.092813275 17 11.8% 5.9% 0.0% 1.1% 0.5% 0.0% Utilities 0.209348049 20 35.0% 15.0% 0.0% 3.7% 1.6% 0.0%

Total Firms 190 Source: Self-made, data source: DataStream Thomson Reuters

as percentage of category as percentage of total Table 4.2 - Correlation between industries and market index returns, full period 2000 to 2012 - This graph shows the proportion of firms according to their degree of correlation with the market index. Analysis on a sector level shows us that there is no industry with correlation with the market exceed 0.75. Utilities, Retail, Manufacturing and Finance industries present approximately exhibit high proportions of firms with correlations higher that 0.5. The industry with the greatest proportion of firms holding a correlation factor higher than 0.25 is Construction with 67 percent. Then Retail, Utilities, Professionals, Manufacturing and Finance (FIRE) hold about 30 percent at the same factor level. Only Social Services and Others do no present a relevant firm/index correlation factor. The second section of this table shares the same information but on a total level. Here the proportions per industry are determined against the total sample.

(26)

This analysis revealed that average salary growth has higher correlation factor with the unadjusted GDP growth level. An average correlation between these series was above 0.9 for the 10 industries under this study. This indicates that production output is clearly linked to wages, which at the same time is one of the main components of human capital. At the same time, unadjusted GDP growth is better correlated with some

industries, as expected, such as FIRE, Manufacturing, Construction, Agriculture and Retail. However, when using adjusted GPD figures, the correlations decrease

significantly. This may be attributed to the of lack positive or negative correlation between salaries and the adjusting variables i.e.: conversion rates, inflation, etc.

The graph below shows the share prices (A) and average salaries (B) variance indexes during the last thirteen years. A positive evolution both factors would suggest some positive level of correlation. Unfortunately, this is not the case. When quarterly data points are analyzed, the correlations between share price returns and average salary are small in the same period.

Retail, Wholesale, Accomodation and Food , 15% Finance, Insurance and Real State, 15% Government, social and other services, 15% Manufacturing, 12% Others, no information, 10% Construction, 10% Professional, Scientific, and Technical Services, 8% Agriculture, Forestry, Fishing and Hunting, 7% Transportation and Warehousing, 6%

Graph 4.1 – Labor, total market composition as at Q4 2012 Source: Self-made with data from SII (Chilean Tax Service Authority or Servicion Impuestos Internos)

Graph 4.2 – Labor, sample composition as at Q4 2012 Source: Self-made with data from SII (Chilean Tax Service Authority or Servicio Impuestos Internos)

Retail, Wholesale, Accomodation and Food 41% Manufacturing 20% Finance, Insurance and Real State 16% Transportation and Warehousing 8% Construction 7% Others 8%

(27)

Graph 4.3– Share price and Average Salary Evolution per Industry between 2000 and 2012 Source: Self-made, data source: DataStream Thomson Reuters and US Federal Reserve Economic Data

Table 4.3 shows the correlations between average salaries and market and economic variables Share Price, Index Price and Adjusted GDP. This comparison is performed to study the drivers of financial value expectation. This analysis shows that the correlation between average salaries and share price returns is not significant when evaluated on the same time period, but that these do increase if lagged later one year.

However, correlations do increase if we lag the Share Prices by one year. According to these calculations, the correlation of Average Salaries and the Market Index increases from -0.02 to 0.30 after the factor is lagged by one year. Manufacturing presents a similar correlation behavior with factor of moving from -0.17 to 0.12. These changes in correlation after a year lag are shown in tables 4.3 and 4.4.

Looking to Tables 4.3 and 4.5, the effect can be noticed. The analysis can help us

detecting the increase in the correlation between Market t-1 and Average Salaries t in

data sample as academic researchers has previously suggested. Table 4.4 presents the correlations of Average Salaries and the four-quarter (1 year) lagged variables.

These calculations suggest that standard deviation increases possibly signaling higher expected volatility in the series. However, correlations also change. As an example I take the average income correlations for FIRE and Manufacturing with the IGPA index, these are now positive and significant. The following table illustrates the lagged variables effect 0 0.5 1 1.5 2 2.5 3 3.5 2 0 0 0 -Q 2 2 0 0 0 -Q 4 2 0 0 1 -Q 2 2 0 0 1 -Q 4 2 0 0 2 -Q 2 2 0 0 2 -Q 4 2 0 0 3 -Q 2 2 0 0 3 -Q 4 2 0 0 4 -Q 2 2 0 0 4 -Q 4 2 0 0 5 -Q 2 2 0 0 5 -Q 4 2 0 0 6 -Q 2 2 0 0 6 -Q 4 2 0 0 7 -Q 2 2 0 0 7 -Q 4 2 0 0 8 -Q 2 2 0 0 8 -Q 4 2 0 0 9 -Q 2 2 0 0 9 -Q 4 2 0 1 0 -Q 2 2 0 1 0 -Q 4 2 0 1 1 -Q 2 2 0 1 1 -Q 4 2 0 1 2 -Q 2 2 0 1 2 -Q 4

A- Share price evolution Q2-01to Q4-12 Agriculture, Forestry, Fishing and Hunt ing

Constru ction

Finan ce, Insu rance and Real State Governmen t, social and ot her services Manu facturing

Mining, Quarrying, and O il and Gas Profes sional, Scient ific, and Technical Services

Ret ail, Wholesale, Ac comodation and Food Transportation an d Wareh ousin g Utilities 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2 0 0 0 -Q 2 2 0 0 0 -Q 4 2 0 0 1 -Q 2 2 0 0 1 -Q 4 2 0 0 2 -Q 2 2 0 0 2 -Q 4 2 0 0 3 -Q 2 2 0 0 3 -Q 4 2 0 0 4 -Q 2 2 0 0 4 -Q 4 2 0 0 5 -Q 2 2 0 0 5 -Q 4 2 0 0 6 -Q 2 2 0 0 6 -Q 4 2 0 0 7 -Q 2 2 0 0 7 -Q 4 2 0 0 8 -Q 2 2 0 0 8 -Q 4 2 0 0 9 -Q 2 2 0 0 9 -Q 4 2 0 1 0 -Q 2 2 0 1 0 -Q 4 2 0 1 1 -Q 2 2 0 1 1 -Q 4 2 0 1 2 -Q 2 2 0 1 2 -Q 4

B- Average Salaries evolution Q2-01to Q4-12

Agriculture, Forestry, Fishing and Hunting

Construction

Finance, Insurance and Real State Government, social and other services

Manufacturing Mining, Quarrying, and Oil and Gas

Professional, Scientific, and Technical Services Retail, Wholesale, Accomodation and Food

Transportation and Warehousing Utilities

(28)

4.3 - Average Salaries

Industry Mean Standard

Error (S P t-0, S a la ry t) (IGP A t-0, S a la ry t) (GDP t-0, S a la ry t) Agriculture, Forestry, Fishing and Hunting 0.01316 0.03548 -0.07803 -0.04990 -0.27741 Construction 0.01614 0.01971 -0.09772 -0.04541 0.14206 Finance, Insurance and Real State 0.01655 0.02639 -0.02489 -0.18903 -0.25830 Government, social and other services 0.01516 0.03059 0.10886 0.06006 0.04001 Manufacturing 0.01210 0.02800 -0.08053 -0.17424 -0.00450 Mining, Quarrying, and Oil and Gas 0.01548 0.05974 0.18995 0.03423 0.07811 Others, no information 0.02298 0.10029 0.09075 0.12645 0.03954 Professional, Scientific, and Technical Services 0.01022 0.05098 0.03687 0.08782 -0.15094 Retail, Wholesale, Accomodation and Food 0.01114 0.02237 -0.15476 -0.13372 -0.18304 Transportation and Warehousing 0.01416 0.02423 -0.00687 -0.04248 -0.25022 Utilities 0.01511 0.08294 0.22992 -0.04637 -0.16666

4.4 - Number employees

Error (S P t-0, Em plo ye e t) (IGP A t-0, Em plo ye e t) (GDP t-0, Em plo ye e t) Agriculture, Forestry, Fishing and Hunting 0.01934 0.13592 -0.14669 -0.12348 0.26504 Construction 0.02079 0.06283 -0.01010 0.15056 0.02157 Finance, Insurance and Real State 0.02181 0.12668 -0.05966 -0.02870 0.03830 Government, social and other services 0.00290 0.05433 0.07987 0.11588 -0.10147 Manufacturing 0.01208 0.05652 0.08535 0.11505 -0.00513 Mining, Quarrying, and Oil and Gas 0.01454 0.05607 -0.15013 -0.21675 0.07811 Others, no information 0.06812 0.29950 -0.17905 0.06466 -0.06101 Professional, Scientific, and Technical Services 0.02796 0.12936 0.07180 -0.16136 0.15882 Retail, Wholesale, Accomodation and Food 0.01100 0.03426 0.08884 0.16215 0.03314 Transportation and Warehousing 0.01031 0.03040 -0.02134 0.00133 -0.16616 Utilities 0.01547 0.14050 -0.55741 -0.12473 0.01891 Table 4.3 to 4.6 - "Series of human capital analysis with relation to Financial-Economic factors". These series porpuse is to ilustrate the basic satatistics per industry of average salaries and number of employees. In addition, these tables shows the change in Salaries and IPGA-Employees correations factors. It is shown that if the market factor is lagged by one year period the correlations with human capital variables increase suggesting that Average Salaries adjust post-change in IGPA.

Correlation with 0 lags Correlation with 0 lags

4.5 - Average Salaries

Error (S P t-4, S a la ry t) (IGP A t-4, S a la ry t) (GDP t-4, S a la ry t) Agriculture, Forestry, Fishing and Hunting 0.01316 0.03548 0.13588 0.10842 -0.20885 Construction 0.01614 0.01971 0.05819 0.13481 0.15903 Finance, Insurance and Real State 0.01655 0.02639 0.08096 0.30771 -0.29748 Government, social and other services 0.01516 0.03059 0.06979 0.14758 0.19754 Manufacturing 0.01210 0.02800 0.02995 0.17613 -0.07397 Mining, Quarrying, and Oil and Gas 0.01548 0.05974 -0.07764 0.08037 -0.03383 Others, no information 0.02298 0.10029 -0.45371 -0.13386 0.22248 Professional, Scientific, and Technical Services 0.01022 0.05098 0.14531 0.04754 -0.02394 Retail, Wholesale, Accomodation and Food 0.01114 0.02237 0.08585 0.05513 -0.09156 Transportation and Warehousing 0.01416 0.02423 -0.07808 0.07711 -0.26827 Utilities 0.01511 0.08294 -0.14800 -0.00085 -0.17596

4.6 - Number employees

Error (S P t-4, Em plo ye e t) (IGP A t-4, Em plo ye e t) (GDP t-4, Em plo ye e t) Agriculture, Forestry, Fishing and Hunting 0.01934 0.13592 -0.20406 -0.20983 0.19816 Construction 0.02079 0.06283 0.06515 0.05119 -0.17235 Finance, Insurance and Real State 0.02181 0.12668 0.06696 -0.11715 -0.00264 Government, social and other services 0.00290 0.05433 0.17004 0.14213 -0.13222 Manufacturing 0.01208 0.05652 0.17840 0.11630 0.01969 Mining, Quarrying, and Oil and Gas 0.01454 0.05607 0.06300 0.18325 -0.15048 Others, no information 0.06812 0.29950 0.07959 0.03910 0.09852 Professional, Scientific, and Technical Services 0.02796 0.12936 -0.00739 0.05218 0.19730 Retail, Wholesale, Accomodation and Food 0.01100 0.03426 -0.10981 -0.02472 -0.07349 Transportation and Warehousing 0.01031 0.03040 -0.01987 -0.03921 -0.20614 Utilities 0.01547 0.14050 0.16407 0.12285 -0.09929 Table 4.5 and 4.6 "Average Salaries and Employees per industry correlated with lagged (1 year) financial-economic factors" Source: Self-made, data source: Chilean Tax Revenue Service and Chilean Central Bank

Correlation with 4 lags Correlation with 4 lags

(29)

4.3 Announcement of public education investment

In the last ten years I found a total of forty-one relevant news releases where Chilean government officials announced changes in the expected public education investment for the following year. The announcements normally appeared in the local and

international press each year near the end of the month of September. The relevancy of the news is determined as follows. The news must indicate amounts or percentages of total GDP to be invested in education. The news is a public declaration of the Finance Minister. The news comes from a main source, a recognized local news web site. Second sources were accepted only if no other prime source was found. In these collection, only three second sources are used and two of these only as support for the main source. This list also includes opinion of the leaders of the higher educational system with regard to the assigned fiscal money to higher education.

Since most official declarations from the Finance Minister were given at the end of

September after business hours, the default testing day (day 0) is the 1st_{of October for}

each year under study because is the first business day after the announcement . In the tests performed during this investigation, I took ten days before and after day zero. This time frame, which is found each year between the 13-14 Septembers to the 12-13 of October, represents the event window. It is during

this period when significantly different abnormal returns are expected with a direction (positive or negative) depending on the announcement. This direction is defined as the announcement’s “Impact” in Table 4.

The impact is given as suggested by the studies of Fama et al (1969) and MacKinlay (1997). If the announcements were positive (1), the expected abnormal returns should be positive as well with alike sign and significant test value. If there were no announcements made (0), then the expectation is zero and with no significant test value. Lastly, if the announcements were negative (-1), the

Year News Expectation

2000 2 1 2001 2 1 2002 1 0 2003 4 -1 2004 2 -1 2005 2 1 2006 3 1 2007 2 1 2008 5 1 2009 3 1 2010 4 -1 2011 6 1 2012 3 1 Total 41

Table 4.7 - Public Education Investment Announcements. This table shows the amount of announcement found in press

end the expected impact ( 1 = positive, -1 negative, 0 neutral)

S o urce: Self-mad e with co llect ed relat ed anno ucement s

(30)

expected test value is significantly negative as the estimated returns. The following table shows the days when the announcement was found in the local news and whether this was positive (1), negative (-1) or neutral (0).

4.4 Public education investment figures

5 Results

This section presents the results of the two abnormal returns tests, CAR and SAR. The methodology applied in this analysis is to employee a default day 0 (zero), as the day of

the official announcement. Initially, assuming this day to be the 1st_{of October for each}

year, the date when the ministry of finance announces the investment plans for the coming year. The selected event window goes from -10 days and +10 days after day zero. The estimation period, the business days pre-event, is 229 business days, (i.e.: 250 – 21 days).

Both these tests, as previously discussed, were applied to the returns of the two industries, FIRE and MANU. These industries represent about 40 per cent of the total sample or almost 100 firms. All year with available data are used for this analysis, with

Year HEI (billions CLP)* HEI growth (HEIg ) GDP Growth (GDPgt -1) HEI as % GDP HEI - GDP Growth 2000 249.32 2001 263.36 5.5% 4.5% 0.63% 1.0% 2002 268.94 2.1% 3.4% 0.59% -1.3% 2003 272.09 1.2% 2.2% 0.56% -1.0% 2004 251.85 -7.7% 3.9% 0.48% -11.6% 2005 258.59 2.6% 5.9% 0.43% -3.3% 2006 300.22 14.9% 5.6% 0.44% 9.3% 2007 301.22 0.3% 4.9% 0.37% -4.6% 2008 321.83 6.6% 4.9% 0.36% 1.7% 2009 385.53 18.1% 3.2% 0.41% 14.9% 2010 420.92 8.8% -1.3% 0.44% 10.1% 2011 451.77 7.1% 5.8% 0.41% 1.3% 2012 496.33 9.4% 5.9% 0.41% 3.5% Mean 0.05737(*) 0.04075(*) 0.00459(*) 0.01662(*)

S o urc e : S e lf-m a de with da ta fro m M inis try o f Educ a tio n, DIP R ES C hile a nd C hile a n C e ntra l B a nk da ta ba s e .

Table 4.8 - Public higher education investment (HEI) Chile 2000 to 2012 - This table shows the level of higher education investment used as a proxi of enhancement of human capital. This table also shows HEI as a percetange of total non-seasonally-adjusted GDP and in the last column, the difference between HEI and GDP growth.

(31)

exception of the years 2001 and 2002 because in these periods 10 approximately did not report share prices.

5.1 Statistical test results by industry and method

The total aggregated results for the FIRE and MANU are presented in Table 5.1 for CAR test and 5.2 for SAR test. These tables are divided into two industries. For each industry I show in the test statistic which is the results of the event study test. CAR follows a normal distribution and SAR test is distributed on a student-t distribution. The column Mean represents the average of the residuals for the period and the column SE AR is the standard error of the firms’ residuals.

The null hypothesis, H0: “there are no abnormal returns on human-capital intensive

firms after the announcement of increase in education investment in the event window” is tested against the CAR test. This test assumes a normal distribution under the given parameters. Under this distribution and at 5 per cent significance, the critical value is equivalent to ±1.64. Since no value in column “Test Statistic” for FIRE and MANU is outside this range, there is not enough evidence to reject the hypothesis. It can be then concluded that under the given parameters, the announcement of increase in education investment does not present statistically significant effects on firm value as measured by the CAR test.

Year Test Statistic Mean SE AR Test Statistic Mean SE AR 2001 -0.27118779 -0.00106182 0.08222409 -0.45073670 -0.00182040 0.08481310 2002 -0.00888569 -0.00003905 0.09229201 -0.13062244 -0.00046789 0.07522145 2003 0.31790195 0.00174320 0.11515239 0.06575480 0.00028815 0.09202482 2004 0.32946528 0.00251063 0.16002679 0.27094179 0.00379632 0.29424273 2005 -0.70117345 -0.00188458 0.05644265 -0.44381743 -0.00121888 0.05767325 2006 -0.23067656 -0.00073385 0.06680679 0.01692270 0.00005441 0.06751834 2007 0.11587744 0.00030162 0.05466091 -0.08991709 -0.00030827 0.07199703 2008 -0.27987095 -0.00136327 0.10229220 -0.50707896 -0.00268681 0.11127049 2009 0.09998050 0.00038128 0.08008505 -0.06291772 -0.00017088 0.05703354 2010 0.07969704 0.00027123 0.07146738 0.13095068 0.00064250 0.10303511 2011 -0.09559160 -0.00028493 0.06259577 0.00632189 0.00002196 0.07293914 2012 0.18372269 0.00068743 0.07857477 0.17653351 0.00042483 0.05053661 Source: Sefl-made, derived from event study analysis of Chilean securities. Data queried from DataStream database.

Finance, Insurance and Real Estate Manufacturing

Table 5.1 - CAR test statistics - This table shows the main statistics of estimated abnormal returns (AR). The calculation of these AR's uses the Chilean market index (IGPA) on model to estimate the return for the period. This aggregation is done for all firms in the Finance, Insurance and Real Estate (FIRE) and Manufacturing industry.