The determinants of investment planning and the reasons for differences between planned and actual investment : evidence from airline companies

UNIVERSITY OF AMSTERDAM

The determinants of investment planning and the reasons for

differences between planned and actual investment – evidence

from airline companies

Master Thesis

Supervisor: Tomislav Ladika MSc Finance

Amsterdam Business School Faculty of Economics and Business

1st _{of July 2017}

Statement of Originality

This document is written by Hendrik Ropers who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

The determinants of corporate investment are a well-covered and discussed area in the academic literature. However, due to a lack of data, little is known about the key drivers of the foregoing planning process and the deviation from these plans. Using the committed purchases of new airplanes from the 10K-statements of 33 airline companies from the years 1994-2014, I investigate which factors drive the investment plans. Furthermore, I analyze what possible reasons are for companies to change these plans and deviate from their commitments. The main focus lies on the role of investment opportunity, managerial overconfidence, incentive compensation and financial constraint. My results show that investment opportunity matters for the level of planned investment. Incentive compensation is a driver of planned investment and over-planning, managerial overconfidence seems to determine over- and underinvestment. Financially constrained firms tend to have higher deviations from their investment due to their worse access to investment funds.

Key Words: Corporate Investment, Investment Planning, Managerial Overconfidence,

Financial Constraint, Incentive Compensation

Acknowledgment: I acknowledge the data collection by Dr. Tomislav Ladika and thank him

Table of Content

1. Introduction ... 1

2. Theoretical background and literature review ... 4

2.1. Determinants of Corporate Investment: Q-Theory, financial constraint and other factors ... 4

2.2. Incentive Compensation and Corporate Investment ... 6

2.3. Managerial Overconfidence and Corporate Investment ... 8

3. Data, Empirical Measures, and Methodology ... 10

3.1. Sample and Data ... 10

3.2 Hypotheses ... 10 3.3. Empirical measures ... 12 3.3.1. Dependent Variables ... 12 3.3.2. Explanatory Variables ... 13 3.3.3. Controls ... 15 3.4. Empirical design ... 16 3.4. Descriptive statistics ... 17 4. Results ... 18

4.1. Determinants of Planned Investment ... 18

4.2. Determinants of Deviations from planned Investment ... 21

4.3. Financial constraint ... 22

5. Robustness Checks and Limitations ... 24

5.1. Robustness Checks ... 24

5.2. Limitations... 25

6. Conclusion ... 26

7. References ... 28

8. Appendix ... 40

1

1. Introduction

How do firms plan their investment? Does financial constraint play a major role in firm’s investment planning process? And do higher incentivized and overconfident managers tend to plan more investment, but then deviate stronger from these plans? There is a great body of literature which examines the determinants of corporate investment, but little evidence about the drivers of the managerial planning process of corporate investment. The existing theory emphasizes that with higher growth, greater investment opportunities and lower financial constraint companies tend to spend more of their funds on investment (e.g. Tobin (1969), Campello, Graham and Harvey (2009), Fazzari, Hubbard and Petersen (1988)). The theory focusses here mainly on the determinants of undertaken investments. However, an important part of firm’s investment is left out in most of the analyses: the step of the foregoing process of investment planning. The majority of companies only reports data about their undertaken investments and hence provides only data for the research of actual corporate investment. In order to examine the planning process, data on the planned investment are needed. Therefore, I will use the unique dataset of 33 airline companies, since this industry has to report in their financial statements their plans of purchasing new airplanes for the next 5 years. The reason for that is that airline companies have to plan for the long-term and hence give commitments to airplane manufacturers in order to make the planning process for both sides more sufficient. Such a reporting for long-term investment planning is hard to find in other industries. Building on panel data for the years 1994-2014, I use the commitments of the airlines as a proxy for planned investment in order to investigate how investment opportunities, incentive compensation, managerial overconfidence and financial constraint affect these plans. Furthermore, by creating a proxy for the deviations from the investment plans based on the actual capital expenditures, I analyze which of the mentioned factors predicts deviations from these investment plans. The academic literature suggests many possible determinants that could explain planned purchases of new airlines. These determinants are mainly based on the theory of actual corporate investment. However, some other factors might also play an important role in the planning process. First of all, according to the Tobin’s Q theory (Tobin (1969), the level of investment opportunity is a main driver of corporate investment and should therefore also matter for investment planning. Although there is a lot of controversy about the explanatory power of Tobin’s Q, it is one of the major factors of corporate investment (e.g. Erickson and Whited (2000)) and should hence be tested for. My hypothesis for the relationship is that firms with higher investment opportunities plan more investment for the next period.

2

Another possible factor that could determine the level of investment plans is the design of managerial compensation, in particular the incentive component. Theory predicts that incentive compensation does not only play an important role in financial policy like dividend pay-out or leverage (e.g. John and John (1993)), but there is also some evidence for a correlation with corporate investment. Academic research shows that compensation design can affect the level of investment (Kang, Kumar and Lee (2006). One of the main explanations is provided with the agency theory, which assumes information asymmetry between shareholders and CEOs (Jensen and Meckling, 1976). As a result, incentive compensation is used as an instrument to mitigate agency problems and make managers act on behalf of shareholders. This suggests that managers plan more investment in order to increase shareholder value. However, higher investment plans could also lead to higher deviations from these plans since CEOs neglect important factors that could influence their investment. Hence, my second hypothesis is that firms with stronger incentivized CEOs plan more investment for the next period and have higher deviations from these plans.

Another determinant that is suggested by theory to be an important factor of corporate investment is the level of managerial overconfidence. A CEO who expects the company to grow in the future tends to plan more investment than a less confident CEO. Furthermore, these CEOs tend to overestimate future projects, both of the own companies or new projects financed through investment. Hence, it is likely that these CEOs have to deviate from their investment stronger than less confident CEOs. Theory predicts managerial overconfidence to play a main role in investment decision-making (e.g. Malmendier and Tate 2005, 2013). My hypothesis states that companies with more confident CEOs plan more investment for the next period and have higher deviations from these plans.

The fourth factor I investigate is the role of financial constraint. As mentioned, theory predicts that with more financial constraint the level of investment decreases. In my research, I investigate whether this relationship is also true for the level of investment planning, and furthermore I check if financially constrained firms deviate more or less from their plans. My hypothesis regarding the deviation is that financially constrained firms deviate less from investment since they have to plan more accurate on their investment spending.

In order to test my hypotheses, I construct measurements for the variable of interests. For investment opportunity, I use Tobin’s Q as a proxy, based on the suggestions from the academic literature. The level of managerial overconfidence is proxied by a variable following Campbell et al. (2011) that measures how deep the options of CEOs are in the money in order to see how

3

positive they expect the future performance of the firm to develop. Incentive compensation is measured by the Dollar amount that a CEO receives from an increase in the company’s stock price by 1%. This measurement captures all equity parts of the compensation and is, therefore, a good proxy for the level of incentive salary. Furthermore, I identify more and less constrained firms based on the SA-index of Hardlock and Pierce (2010), which is only built on the exogenous factors of firm size and firm age. Using these measurements, I regress the committed investment for the next year on the generated variables in order to detect the determinants of planned investment, with varying specifications. After, I run the same regressions with the deviation from the planned investment, measured as the difference between Capital Expenditure and the committed investment. This helps me to find determinants of why companies do not stick to their investment plans, differing between negative and positive deviation. Lastly, I regress my two dependent variables based on two subsamples regarding the level of financial constraint.

My results show that investment opportunity is a statistically significant determinant of the level of planned investment. More investment opportunity leads to more planned investment. These findings are robust for the committed investment for the next period and the period after. However, this result changes in the other direction when including further airline-specific variables. It is not clear if this is due to the variables itself, or a problem of a too small sample size.

For the incentive compensation, I find weak evidence for a positive impact of incentive compensation on the level of committed investment. Furthermore, my results show that incentivized CEOs tend to plan more investment compared to what they actually undertake and is in line with previous theory.

The role of managerial overconfidence cannot be finally determined based on my results. On the one hand, managerial overconfidence tends to have no significant impact on investment plans. However, based on a smaller number of observation and including airline-specific the results suggest that there might be a negative impact of overconfidence on corporate investment planning. Regarding the deviations, managerial overconfidence affects both positive and negative deviations from planned investment positively.

For the financial constraint, I provide evidence that more financially constrained firms tend to have more positive deviations than less constrained firms.

Since there has been little evidence about the determinants of investment and the reasons for differences between planned and actual investment, my research would investigate a field

4

which adds an important topic in the theory of investment. Additional to what determines investment itself, it contributes by looking closer at the planning process of companies and examines which factors predict deviation from this planning.

The paper is structured as followed: In section 2, I review the relevant academic literature regarding corporate investment, with a particular focus on financial constraint, incentive compensation, and managerial overconfidence. In section 3, I describe and summarize my sample data, and introduce into the empirical approach of the research. Here I focus on the derivation of my hypothesis, the construction of the variables and the empirical model. In section 4, I explain and discuss the results and put them in a theoretical context. Following in section 4, I test and discuss the robustness of my results Section 5 concludes.

2. Theoretical background and literature review

2.1. Determinants of Corporate Investment: Q-Theory, financial constraint and other factors

The question of what determines the investment level of a firm is a highly relevant topic in financial theory. There has been a lot of research done about it, and following I review the main concepts and theories. Starting with the findings of Modigliani and Miller (1958) who show that capital structure and corporate investment decisions are not related under certain conditions, the academic theory provides multiple approaches and explanations. One major factor of a firm’s investment level is predicted by its investment opportunities. Tobin (1969) provides with his Q-theory as first an explanation that firm’s investment could be predicted with the ratio of its market value to the replacement cost of capital. He uses this measurement as a proxy for unobservable investment opportunities. In the academic theory, Tobin’s Q became a common and widely used measure for investment opportunity, modified in many ways. However, this theory only holds under certain (perfect) market conditions, since information asymmetry can lead to adverse selection and moral hazard, as indicated by Akerlof (1970). Hence, Tobin’s work became the starting point for much further research about the determinants of corporate investment. One of the most important and famous theories in the field of corporate investment theory is the pecking order theory (Myers, 1984). It suggests that external financing is more expensive than internal financing, and hence can be seen as an indicator of the financial situation of a company. Based on that, a major factor that affects corporate investment decisions is investigated by Fazzari, Hubbard and Petersen (1988) who find that financial constraint impacts firm’s investment. Their analysis shows that the sensitivity

5

of investment to internal capital is stronger for more constrained firms and weaker for less constrained firms which implies that firms tend to invest more when they have more internal funds available.

Financial constraint means that a firm has difficulties to obtain external finance. As a result, financially constrained firms have limited access to capital and are therefore not able to provide the firm with sufficient liquidity and capital for possible investment with positive net present value. Farre-Mensa and Ljunqvist (2013) define financial constraint in a way that raising new funds from external sources becomes both more expensive and difficult, which can lead in an extreme case to a situation in which such firms are completely shut out from access to capital. However, being financially constrained does not only affect corporate investment but also firm’s efficiency (Bond, Söderbom and Wu, 2010) and innovation) and innovation (Savignac, 2008).

Erickson and Whited (2000) show that the results of Fazzari et al (1988) are consistent with models in which financing is frictionless. However, Kaplan and Zingales (1997) findings contradict Fazzari, Hubbard and Petersen. Their results show that cash flows are correlated with firm’s investment regardless the level of financial constraint. Differently from Fazzari, they do not use dividend payouts as a measurement for financial constraint but reclassify firms by using discussions of financial constraints in firm’s 10-K statements. Their second main argument is that there might be a bias in Fazzari’s findings, since cash flows can be correlated with investment opportunities and hence their results include possibly a measurement error related to the application of Tobin’s Q. Furthermore, also Gomes (2001) finds that financial constraint does not significantly matter for cash flow effects. Their results, as well as the ones of Kaplan and Zingales, support the idea that Fazzari’s findings in 1988 are mainly due to problems with the Q- measurement. Clearly et al. (2007) argue that these contradictory findings are due to a lack of correct empirical proxy for financial constraint. Another evidence for the impact of being financially constrained on corporate investment is provided by Campello, Graham and Harvey (2010) who find that constrained firms have deeper cuts on tech spending, employment and capital spending. Lamont (1997) uses a “natural experiment”, provided by the oil crash in 1986, to show that oil companies had to cut their investment in non-oil-related departments due to the exogenous shock. This supports the idea that financial constraint matters for corporate investment.

Regarding the use of Tobin’s Q, there is a deep and open discussion in the academic research about how to use it. George et al. (2011) argue that Tobin’s Q is a good measurement for

6

investment opportunity since it allows to proxy the expected value of future projects' profitability. Erickson and Whited (2000) find that the measurement error, as found by Kaplan and Zingales (1997), leads to an attenuation bias in regressions, which means that the coefficient of the Tobin's Q would asymptotically tend to 0. This finding also applies on R-squared which tends to fall in case of measurement errors to 0. Hayashi (1982) suggests in his paper that since the better fitting marginal Q is not observable, the market-to-book value of companies can be used as a measurement of the average Q.

The academic literature provides a wide range of other determinants for corporate investment. For example, Peters and Taylor (2016) examine the role of intangible assets. Their findings show that the amount of intangible assets of a company matters for its investment decisions. However, they are not first ones to examine the relationship between corporate investment and intangible assets (Gourio and Rudanko, 2014, Baker, Stein and Wurgler 2003). Almeida et al. (2004) find that more cash holdings increases firms’ investment opportunity and therefore could be another determinant of investment planning. Furthermore, Brown and Petersen (2011) find that liquidity and investment decisions are linked by showing how cash holdings are used to smooth R&D investment. Further literature investigates the impact of discount rates on investment. Lamont (2008) finds that investment responds to changes in discount rates and, additionally, that investment can increase due to unexpected profits. Furthermore, Denis and Sebilkov (2009) find that there is a significant relationship between cash holdings and the level of investment. Ruiz-Porras and Lopez-Mateo (2011) show in their paper that the degree of separation of ownership and control, approximated by agency costs encourages more investment decisions. Their findings are one example for the role of corporate governance on firm’s investment.

In this review of the relevant literature, I showed how financial constraint and investment opportunity have an empirically proofed impact on corporate investment. Furthermore, I introduced into several other factors that matter for investment decisions. My research will contribute here since the planning process of corporate investment has not yet been investigated with regard to financial constraint and investment opportunity.

2.2. Incentive Compensation and Corporate Investment

When it comes to the planning process of corporate investment, there are some factors besides the financial ones that play an important role. One of them is the function of executive compensation design, especially the incentive part of it. The theoretical background of the

7

relationship between corporate finance decisions and CEO compensation design is the agency theory (e.g. Jensen and Meckling, 1976). Information asymmetry between managers and shareholders leads to uncertainty for the investor about the quality of firm’s investment decisions, and hence one argument of academic literature is that incentive compensation can be one way to minimize ex-ante the agency costs of shareholders. The standard optimal contracting framework (e.g. Holmstrom 1979) suggests that the investment decisions of managers are endogenously and jointly determined in a second-best incentive arrangement. As the main goal, incentive compensation should result into better quality of firm’s investment decisions since managers would only undertake projects which are positive for shareholder’s value.

There is a great body of academic research which supports the idea that equity-based CEO compensation and corporate financial decisions are related. This includes on the one hand decisions regarding dividend policy, capital structure, and other financial parameters: John and John (1993) for example find a positive empirical relationship between firm leverage and executive compensation leverage. Tong (2008) finds evidence that changes in CEO ownership are related to firms’ future returns. Agrawal and Mandelker (1987) find a positive correlation between CEO ownership and changes in firm variance and leverage, due to the reduction of agency costs. Lambert, Lannen, and Larcker (1989) show that incentive compensation matters for the dividend policy of a firm. There is also evidence that incentive compensation improves firm performance in general. Mehran (1995) provides with his findings evidence that stock options as a part of CEO compensation can mitigate agency conflicts and hence argues with based on his empirical results that compensation structure matters for firm performance. Jensen and Murphy (1990) and Hall and Liebman (1998) found similar results.

On the other hand, and even more related to the research in this paper, there is also evidence for an empirical relationship between equity compensation and corporate investment policy. Eisdorfer, Giaccotto and White (2013) show in their paper that similarities between the CEO compensation leverage ratio and firm’s leverage ratio impact the quality of firm’s investment decisions.They find that CEOs with higher debt-like compensation components tend to under-invest and CEOs with more equity-based compensation tend to over-under-investment. Datta, Iskandar-Datta and Raman (2001) findings suggest that equity-based compensation and corporate acquisitions have a strong positive relationship, not depending on the type of acquisition. The research of Kang, Kumar and Lee (2006) support their finding by investigating the relationship between corporate investment and managerial incentive structure. Their results show that long-term investment is significantly related to incentive payment of CEOs through

8

equity-based compensation. Furthermore, they build on Mehran et al. (1998), who found that liquidation decisions are influenced by CEO incentive plans. Kang, Kumar and Lee’s (2006) findings provide an interesting determinant: They argue based on their results, the standard Q-theory as introduced in the previous section, appears to miss in its equation the factor of compensation design.

Based on this review, it becomes clear that incentive compensation matters for many financial decisions and policies in a company. The academic literature suggests furthermore a relationship between corporate investment and compensation structure. However, most of the research does not directly investigate the effect on corporate investment, and especially not the planning process. Hence, I will contribute with my paper to the existing research.

2.3. Managerial Overconfidence and Corporate Investment

In the previous section, I introduced into the main literature regarding the link between executive compensation and corporate investment. The role of equity compensation for the level of firm’s investment suggests that there could be another determinant of planned investment: The level of managerial overconfidence. When managers are (partially) incentivized with stock options and other equity compensation, CEOs have the choice about when to exercise these options. The timing of the exercise indicates how confident a CEO is about the future development of the firm’s well-being and ultimately the stock price. Overconfident managers expect a positive performance of the firm, and since investment is a main driver of financial performance, theory investigated this relationship previously. Following, I review the relevant literature that relates managerial overconfidence to corporate investment.

The basic concept of managerial overconfidence builds on findings from behavioral economics. The idea of a homo oeconomicus who only behaves rationally is just a theoretical model. There is a great range of literature which suggests that manager, in general, tend to overestimate themselves and think they perform better than the average (Alicke, 1995). However, the measurement of this effect is difficult since this is an unobservable characteristic (compare section 3.3). Nevertheless, there are different approaches for this, which are used in literature to set overconfidence in relation to corporate finance theory. Kent, Hirshleifer and Subrahmanyam (1998) for example find that overconfidence leads to negative investment outcomes, stock mispricing and excess volatility in their research about investor overconfidence. Oliver (2005) shows in his research that there is a correlation between CEO

9

overconfidence and firm’s capital structure. His findings suggest that companies with more confident CEOs tend to have a higher debt ratio.

Ben-David, Graham and Harvey (2007) use panel data of about 7,000 observations of probability distributions from top financial executives for the stock market and find that firms with overconfident CFOs use lower discount rates for valuing cash flows. Furthermore, their results show that overconfidence is related to more investment, more debt, and less dividend-payout. Deshmukh, Goel and Howe (2009) find similar results for the payout policy. A paper about overconfidence and mergers and acquisitions is published by Doukas and Petmezas (2007). They show that more confident CEOs tend to undertake acquisitions projects more quick and frequent due to their too positive and superior self-estimation.

Two very popular researchers in the field of overconfidence are Malmendier and Tate who published a couple of academic papers regarding the relationship of overconfidence and corporate finance and investment policy. In 2005a, they find that overconfidence has an effect on the investment-cash flow sensitivity. This means that more confident CEOs tend to plan investment projects more likely with internal funds. Furthermore, they show that highly overconfident CEOs’ investment is significantly more responsive to cash flow. In their paper of 2006, they provide evidence about the impact of overconfidence on corporate financial policy. Building on the assumption that overconfident CEOs expect their company to be undervalued, they show that these CEOs are significantly less likely than less confident CEOs to issue equity. The two authors find further fields of corporate finance which are impacted by managerial overconfidence (e.g. 2003, 2015).

Campbell et al (2011) analyze the role of CEO optimism for forced turnovers. Additionally, they show that CEOs with optimism over the interior optimum tend to overinvest. A research about the impact of overconfidence on the estimation of project risk is done by Goel and Thakor (2008). They provide empirical evidence that CEOs can underestimate project risk due to overconfidence. Based on their results, this can improve the level of investment and lower the likelihood of forced turnovers.

This introduction into the main literature about managerial overconfidence suggests that this is a highly relevant topic when it comes to corporate investment. For my research, which contributes to the previous papers by analyzing the planning process of firms, overconfidence seems to be an important factor.

10

3. Data, Empirical Measures, and Methodology 3.1. Sample and Data

As discussed before, there has yet not been a lot of research done about the planning process of firms regarding their investment decision. This is mainly due to a lack of data since most companies only report what they invested in their previous year, but nothing about their future plans. My thesis aims to fill this gap by using unique data from airline companies as they have to report their commitments for airline purchases. I assume these commitments to be a good proxy for general investment plans since investment in new airplanes is the main way for airlines to adjust to their strategic and financial growth. In particular, I gather data about the investment planning of airline companies and variables regarding their financial firm characteristics, taken from the company’s 10-K annual report. The raw dataset which I will use is already collected by research assistants of Dr. Tomislav Ladika. This dataset contains the airplane purchase plans for the next five years of 33 airline companies for the sample period 1994–2014 (see Table 1). I investigated a few comparable industries (e.g. shipping companies) which also report investment commitments, however, the reporting of planned investment per year is not as precise as for the airline companies and therefore not sufficient for my research. Additionally, I gathered further data to the given sample for my research. At first, I collected data for the company characteristics, which include financial data from the database Compustat and industry-specific data about airline companies like the size of the airplane fleet and the available flight miles. Additionally, I gathered compensation data for the airline companies from Execucomp. Here I focused on basic measurements of payment, data about equity compensation and information needed for measuring managerial overconfidence.

3.2 Hypotheses

Based on my review of the relevant academic literature in section 2, I developed 4 hypotheses that I want to test. My first hypothesis addresses the basic determinants of planned investment. Theory predicts that companies with higher investment opportunities should tend to plan more investment into new airplanes than companies with fewer investment opportunities. The basic economic reasoning behind this is that the main goal of firms is usually to maximize the firm and/or shareholder value. Assuming that higher investment opportunities are a proxy for more availability of realizable projects with positive net present value, higher investment opportunity should be positively correlated with investment plans. Hence, my first hypothesis is:

11

Hypothesis 1: Companies with higher investment opportunities plan more investment for the next period.

The next determinant that I am analyzing is the role of CEO compensation design, in particular, the incentive component of compensation. As reviewed in the section before, theory suggests that incentive compensation is an instrument for shareholders to minimize agency costs. Hence, managers are incentivized to act on behalf of shareholders and therefore try to increase shareholder value. As a result, managers could attempt to find more positive investment projects and hence plan more investment. Furthermore, these CEOs might plan more investment than they actually undertake, since they underestimate certain factors that change their plans due to their willingness to increase shareholder value. This argumentation leads to my second hypothesis:

Hypothesis 2: Companies with more incentivized CEOs plan more investment for the next period and have higher deviations from these plans

The third hypothesis contains the assumed relationship between CEO overconfidence and planned investment and the deviation from it. Since more confident CEOs expect the firm value to increase in the long-term, the economic intuition suggests various things: At first, these CEOs are more likely to believe that the investment projects that the company investment chances. Relating this to the airlines, this could mean for example that a manager expects to win new routes or hubs for the airline and therefore sees the reasoning for purchasing new airplanes. This managerial overconfidence would lead to relatively higher investment plans than for companies with less overconfident CEOs. Furthermore, it seems likely that these firms have higher deviations from their plans since unexpected changes in the firm or industry might have been underestimated. Hence, my third hypothesis is:

Hypothesis 3: Companies with more confident CEOs plan more investment for the next period and have higher deviations from these plans

Lastly, I expect financial constraint to be a further determinant in my research. On the one hand, I assume financial constraint firms to plan less for new investment. This would be in line with corporate investment theory which predicts that financially constraint firms invest less in general. On the other hand, I expect that financially more constrained firms have to plan more exact due to their financial situation. Since these firms have likely more need to save earnings and be more careful with their funds, the chances of great changes in their investment plans seem less likely. This, I expect financially constrained firms to have lower deviations from their planned investment, which ultimately results in my last hypothesis:

12

Hypothesis 4: More constrained companies plan less investment for the next period and have lower deviations from these plans.

3.3. Empirical measures

The specific definitions of the variables can be found in the appendix in Table A1.

3.3.1. Dependent Variables

i. Planned investment

The variable ‘Planned investment in year x’ measures the amount which the airlines are planning to invest in year x. Although I have data for the next five years, I use only the data for the next period. This has to reasons: On the one hand, the lack of data for the other periods is bigger, and on the other hand, I expect the deviations from planned investment to increase over time and therefore I assume that the comparison between planned and actual investment becomes noisier. However, for robustness tests, I investigate the effects of the planned investment for the next two periods.

ii. Deviation from planned investment

In order to test which factors matter for the deviation from planned investment, I generate a variable “Deviation” that measures how much the actual investment is different from the amount of planned investment in new airplanes in the period before. Hence, this is measured as the difference between Capex in year 1 and the investment committed for year 1 in the year before. I use Capex a proxy for the actual investment in new airplanes. I am aware of the fact that the Capex includes other investment than only new airplanes, but for airlines, the majority of Capex are purchases of new airplanes. However, this noise in the measurement could affect my results and hence presents a severe limitation of my research. In section 5, I go more into detail of this problem.

The deviation from the committed investment could be both positive and negative, depending on whether the firm spent more or less on new airlines than they committed. Since I want to differ between these two cases of over- and underinvestment, I create two specifications of the deviation: “Negative deviation”, if the committed investment was higher than the Capex, and “Positive deviation” in case of an increase of the actual investment compared to the commitment.

13

3.3.2. Explanatory Variables

a. Investment Opportunity

In order to measure the investment opportunities of a firm, I apply the academic findings regarding Tobin’s Q theory: The original proxy provided by Tobin (1969), the marginal Tobin’s Q, measures unobservable investment opportunities by a ratio between the market value of an additional value and the replacement costs of this investment. Later, Fazzari et al. (1988) adjusted Tobin’s Q by using the average of it. In consequence, academic research provides a great amount of discussion and contradiction regarding this approach (e.g. Kaplan & Zingales, 1997; Rauh, 2006; Chen & Chen, 2012).

In general, the idea behind using Tobin’s Q as a measure of investment opportunities is that companies invest only in projects with a positive net present value. Hence, this suggests that a firm only undertakes investments when the created value of the of it is higher than the firm’s costs of capital for this project. The capital markets anticipate this and therefore the investment decision affects the company’s stock price even before undertaking the investment. Hence, it can be supposed that measuring the investment opportunity of a firm is possible by measuring the ratio between its market value and the replacement costs of capital, in other words with Tobin’s Q. In academic theory, Tobin’s Q is a common and widely used measure for investment opportunity.

b. Managerial Overconfidence

The academic literature suggests various ways for measuring managerial overconfidence. The main problem of measuring overconfidence is that this is unobservable, and hence a fitting proxy has to be generated. The approaches for this differ in the academic research: One approach is to create variables based on firm investment, as for example Malmendier and Tate (2005) do it in their research. Campbell et al. (2011) follow this approach by classifying managers as high optimistic when the firm is in the upper quintile of the firm’s industry-adjusted investment in two following years. Other approaches identify overconfidence based on net stock purchases Malmendier and Tate (2005) or based on the CEO’s portrayal in the media Malmendier and Tate (2008).

For my research, I follow Campbell et al. (2011) who develop a proxy for managerial overconfidence based on the work of Core and Guay (2002) approximation. They measure overconfidence by calculating the option moneyness of the CEO’s option in order to detect the behavior of the CEO’s option exercise policy. Managers who hold options, that are at least 67%

14

are considered to be overconfident. Intuitively, this makes sense since holding options longer than necessary suggests that a manager expects the firm’s share price to increase and therefore detects overconfidence. Hence, I follow their approach and create the variable “Average option moneyness” as defined in Appendix A1. Furthermore, I also follow the cutoff of 67% to identify optimistic CEOs.

c. Incentive Compensation

For measuring incentive compensation, I follow Bergstresser and Philippon (2005), who measure the dollar gain for a 1% increase for the CEO in shareholder value (dollar gain from +1%). This measure, defined as “OnePercent” provides a proxy for incentive compensation since it captures the real financial benefit that managers get from acting in the interest of shareholders. For robustness test, I follow Fahlenbrach and Stulz (2011) in order to create a second measurement: “Percentage Ownership” measures the ratio that the CEO owns through possessing stocks and stock options of the company.

d. Financial Constraint

In order to detect the level of financial constraint, there are many different measurements provided by academic literature. A popular measurement for financial constraint is the level of dividend payout ratio, following Fazzari et al. (1988), Almeida & Campello, (2007) and several other studies. Almeida, Campello & Weisbach (2004) use the cash to cash flow sensitivity in order to detect constrained firms, meaning that those firms have a positive relationship between cash stocks and cash flow. Another well-known way to measure financial constraint that is applied in this paper is the KZ-index, which measures constrained firms based on financial accounting data. The KZ-index, introduced by Kaplan & Zingales (1997), is created based on regression findings, contains the variables cash-flow over total assets, long-term debt over total assets, total dividends over total assets, liquid assets over total assets and Tobin’s Q. Furthermore, Erickson and Whited (2000) proposed that a firm is constrained when it´s total assets are continuously at the lower half (50%) of the firms in a sample dataset. This means that they also use firm size as a measure of financial constraint.

For my research, I follow Hadlock & Pierce (2010) who find that after controlling for firm`s size and age only leverage and cash flow identifies the level of financial constraint. However, due to endogeneity reasons, they argue that the only exogenous variables firm size and age can be used to measure financial constraint. Based on that, they build the SA-index which is only based on these two factors. The idea behind it is that smaller companies tend to be younger and are therefore more likely to face the problem of information asymmetry when finding external

15

funds. The question if firm size is exogenous and therefore an appropriate choice for an independent variable in a regression can be answered with yes, as firm size and age are given and not dependent on investment (Carreira & Silva, 2010; Hughes, 1994). I follow their argumentation in order to have a measurement of financial constraint, in particular meaning that firms with a higher SA-index are suggested to be more constraint than firms with a low SA-index. The SA-index is calculated as followed:

𝑆𝐴 − 𝑖𝑛𝑑𝑒𝑥 = 0,737 × S + 0,043 × 𝑆2 _{– 0,040 × A (1)}

In equation (1), S measures firm size and A firm age. Here, firm size is measured as the natural logarithm of total assets. For measuring firm age, there are different options: The preferred way would be to use the number of years since the company is listed since this is intuitively a good proxy when talking about financial constraint. However, Compustat and other databases do not have sufficient information about this. Therefore, I hand-collected information about the year of incorporation in order to calculate firm age following (Carreira & Silva, 2010).

3.3.3. Controls

a. Firm-specific variables

There are different factors that can influence both planned and actual corporate investment. Theory suggests that these variables do not only affect investment, but also the explanatory variables. Hence, it is important to control for them. My firm-specific controls include the Market-Book Ratio and Log assets in order to control for the size of the airlines. Furthermore, I control for the performance and therefore profitability of the firms by including the Return on Assets and cash flows in the regression. All of these fours variables control for possible differences in investment opportunities. As suggested and mentioned in the literature review, Peters and Taylor (2016) found that Tobin’s Q works better when Intangible Assets are included in the regression. Hence, I follow his approach and control for Intangibles as well. Additionally, it is controlled for the Book Leverage of the airlines. Lastly, it is also controlled for all firm and time fixed effects.

b. Industry-specific variables

For the airline industry, there are a couple of interesting factors that might affect the investment in new airplanes. One possible determinant of corporate investment could be the price for fuel (kerosene). With an increasing price for kerosene, airlines could tend to invest less into new airplanes since the profitability of the general business might decrease. Hence, years with higher oil prices could experience a decrease in new airplanes purchases and

16

therefore it is controlled for the fuel price. Furthermore, the average age of the airplane fleet is a factor that has to be controlled for. In theory, airlines with relatively old airplanes have to replace sooner their airplanes than companies with younger fleets. Lastly, I control in the regressions for the size of the airplane fleet. The reasoning behind this is similar to the airplane age since airlines with fewer planes might have to buy more new airplanes (or: companies with more airplanes grow faster and buy new ones).

3.4. Empirical design

In order to test my hypotheses, I developed different steps within my research. First of all, I conduct a general investigation of the main drivers of the airplane purchase plans. Therefore, I use a panel data model which measures the different effects of variables on the planned investment.

𝑃𝑙𝑎𝑛𝑛𝑒𝑑 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝑥_𝑖,𝑡 + 𝛽₂𝑥₂ + 𝛼_𝑖+ 𝛿_𝑡+ 𝜀 (2) In equation (2), the dependent variable is the planned firm investment in the next period, and 𝑥_𝑖,𝑡 stands for different factors that might influence firm investment and represents therefore the independent variables. This includes also certain airline industry specific variables. Additionally, I control for different factors as specified above (𝑥₂). Finally, I control for firm (𝛼𝑖) and time (𝛿𝑡) fixed effects. This is necessary to avoid that shocks impact the results and

therefore the true causal relationship. Based on this general examination, I have a closer look at the relationship between the planned investment and managerial overconfidence, investment opportunity and incentive compensation. Therefore, I use the proxies for these three as explanatory variables, defined as described in the previous section.

In a second step, I investigate the deviation of the airlines from their planned investment. Therefore, I use the variable “Deviation” as a proxy in order to test my hypotheses regarding the deviations from planned investment. The regression equation is similar to the first one, but with a “Deviation” as dependent variable. Since the deviation from the planned investment can be either positive or negative, I differ here between positive and negative differences to see what drivers over- and underinvestment.

𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝑥_𝑖,𝑡+ 𝛽₂𝑥₂+ 𝛼_𝑖 + 𝛿_𝑡+ 𝜀 (3) The first two steps of my research are constructed for testing the first three hypotheses regarding investment opportunity, incentive compensation, and managerial overconfidence. In a third step, I have a closer look at the impact of financial constraint on the above results by

17

splitting my sample into two subsamples based on the level of financial constraint. This subsample is created based on the SA-Index, described in the previous section. The upper half of the sample based on this index is considered as more financially constrained, and the lower half as less constrained. The samples are created in a way that the observations are split based on the median of the SA-index for each year. Building on that, I can test my fourth hypothesis.

3.4. Descriptive statistics

As mentioned above, my sample dataset builds on the raw dataset about airlines’ commitments for purchasing new airplanes within the next years. In Table 2, Panel A, I provide summary statistics regarding the key financial variables for my sample of airlines. As can be seen, the sample covers mainly large airlines, which can be derived from a median asset value of $8.4 billion and a mean asset value of $18.7 billion. The median market capitalization is $1.2 billion and the mean $4.3 billion. Panel B of Table 2 provides more particular data about my key dependent variables. The median committed investments for the next period. in new airplanes is about $601 million, and the mean is $1.4 billion. It is worth noticing that these values are in the same range as the values for the Capital Expenditures, which suggest that the Capex provides a relatively good proxy for creating the deviation measure. Furthermore, the statistics provide an intuitive finding: As further the time horizon goes into the future, as more the committed investment decreases. This suggests that airlines try to plan their investment in new planes for the short-run or at least attempt to keep the commitments for future periods on a low level. Additionally, I report the summary statistics for the developed variable “Deviation”. As mentioned, “Deviation” deviation is calculated as the difference between Capex and planned investment. Hence, “Negative” measures the deviations when the Capex, meaning the actual investment, was lower than the committed, and “Positive” the other way around. Therefore, one can derive from the summary statistics that there were about the same number of over- and under investments. (209 vs. 239). In graph 1, the distribution of the deviation is illustrated. As can be seen, most airlines under- or overestimate their investment in a range of about $500 million, which is in line with the median values for “Negative Deviation” (“Positive Deviation”) of -$1,538 million ($329 million). The curve suggests that there is a slight skewness to the left, meaning that more airlines tend to invest less than they planned. This is relatively surprising since the Capex captures more investment than just the investment in new airlines, and hence is ex-ante expected to be higher than the planned investment. Furthermore, by analyzing the raw data, it stands out that the commitments for the next year always differ: In particular, one

18

would expect that the committed investment in year t for the year t+2 would be the same as the committed investment in year t+1 for the year t+1. However, it can be seen that there is always a difference in these numbers, which means that the airlines adjust their investment for the same period of time.

Panel 3 of Table 2 summarizes compensation measures. In the table, one can see the mean and the median for components of annual compensation, bonus, the equity portfolio of the CEOs and equity portfolio incentives. The median (mean) of the total compensation in the sample is about $2.4 million ($4.3 million). This total compensation consists less of a basic salary, but more of equity-based payment from options and stocks (about 67% regarding the mean). Additionally, the mean of 1834.35 of the “Dollar gain from +1%” increase in stock price shows that CEOs are highly motivated to increase stock price and hence behave in the interest of shareholders.

In Table 3, I summarize industry-specific statistics for the airlines. The median airplane age of the airlines is 9 years, which is relatively young. The airlines have in the median 272 aircraft in services, of which in the median 121 airlines are leased and 131 are owned. In figure 2 I provide the development of the kerosene price. This is worth mentioning, since the kerosene price fluctuated during the sample period strongly, with a maximum in 2012 of about $3 per gallon and a minimum of $0.40 per gallon in 1998. From 1994 to 2014, the kerosene price tripled. This supports the idea that the kerosene price should be included in the regression model since it is a major driver of airlines’ expenses.

4. Results

4.1. Determinants of Planned Investment

In this section, I discuss the results from analysis of the determinants of planned investment. Table 4shows the results from the panel data regression, with the committed investment for the next year as the dependent variable for all specifications. Column (1) provides an analysis of the impact of Tobin’s Q. The coefficient for Tobin’s Q is highly statistically significant at the 1% level and economically large with 11,595. The economic magnitude suggests that with an increase in Tobin’s Q of 1, the investment plans for the next year increases by Since Tobin’s Q is a ratio, this large magnitude makes sense as an increase of one in a ratio is a great change in investment opportunity. It can be concluded, that investment opportunity matters for planned investment and therefore this result is in line with the first hypothesis of this research. This finding follows the standard corporate investment theory and fits the results

19

of previous research. The second column adjusts the regression by using “OnePercent” as a measurement for incentive compensation as the explanatory variable. As one can see, the coefficient is statistically significant at the 10% level, and has an economic magnitude of 0.02. It implies that if a CEO gets one dollar more for an increase in stock price of 1%, the investment plans increase about $20,000. Putting this in a more realistic magnitude, the economic magnitude seems significant: If a manager gets more incentive payment based on equity, and therefore profits from an increase in stock price of 1% with $500 more, the level of committed investment increased by $10,000,000. This suggests that airlines with CEOs who are more aligned with shareholders plan more investment into new airplanes. The third column provides the results with the third variable of interest as the explanatory variable, the overconfidence proxy “Average Option Moneyness”. The coefficient appears to be insignificant, meaning that being overconfident does not have an empirical impact on the level of investment planning. Even the negative direction of the coefficient does not support previous empirical findings, meaning that overconfident CEOs should tend to plan more investment due to their positive expectations, and hence this result level suggests that the hypothesis has to be rejected. The results in column (4) strengthen these relationships: Here all of the three measurements are included in the regression, and the main implication of the findings do not change. The levels of statistical significance suggest that investment opportunity (now at the 10% level) and incentive compensation (at the 1% level) matter for corporate investment planning, whereas overconfidence stays insignificant. The explanatory power and the economic magnitude of the two significant coefficients, meaning that the impact of Tobin’s Q decreases now to a level of 5,910, whereas the magnitude of the incentive measurement “OnePercent” increases from 0.02 to 0.03 which is highly economically significant.

In the first four columns, the model controlls mainly for financial characteristics of the firm. Worth mentioning is here the coefficient of cash flows, which is in line with the theory in the columns (2), (3) and (4), meaning that has strong economically and statistically effect on the level of airplane purchase commitments. The adjusted R-squared for these four regressions has the highest value in column (4), which suggests that by including all three variables of interest the model explains more of the variance in the dependent variable.

The main change from columns (1) – (4) to the columns (5) – (8) is the adjustment of control variables. Besides the financial measurements, the model includes now the size of the airline fleet as well as the average age of the airplanes. Before analyzing the results of these regressions, there is one main remark to be made: Due to a lack of data regarding the

industry-20

specific variables, there are way fewer observations for the columns (5) - (8). As a result, I assume that the findings from these specifications have to be handled carefully, however it might still give an implication for the research done in this paper.

When controlling for the mentioned airline-specific characteristics, the results change in a couple of ways: Both investment opportunity and incentive compensation do not have a statistically significant level when including them as a single explanatory variable (besides the controls). Furthermore, managerial overconfidence has a statistically significant negative impact on the level of planned investment. The finding would suggest that overconfident managers plan $126 million less on new airplanes when increasing the average moneyness of their options by one. The theoretical implications are not straightforward, since this finding would mean that the overconfident CEOs think that less investment leads to higher firm value. This would make sense in times of crisis or for example structural problems in the whole industry. However, since I control for time fixed effects, this seems unlikely.

In column (8), where the three measurements are jointly included, the results suggest a high statistical and economical significance level of investment opportunity. The impact of investment opportunity seems to be positive for the level planned investment, and both of the two other measurements are not having a significant effect.

Based on the overall regressions results, there can three first conclusions be made: On the one hand, Tobin’s Q as a measurement of investment opportunity seems to be a statistically and empirically significant driver of planned investment. Although column (5) does not support that implication, the first hypothesis regarding the positive correlation between investment opportunity and planned investment can be supported, since the findings in column (5) are likely to be biased due to a small sample size. As a second finding, managerial overconfidence is more likely to have no significant impact on investment plans, based on the results from columns (3) and (4). However, based on a smaller number of observations and including airline-specific characters in column (7), the results suggest that there might be a negative impact of overconfidence on corporate investment planning. This would not be in line with academic theory and my hypothesis, which implies that overconfident CEOs plan more investment. The third finding provides evidence for a positive impact of incentive compensation for CEOs on the level of their planned investment.

Finally, I would conclude here that the results from columns (1) to (4) support stronger the findings of previous literature. However, a further investigation with more data on airline-specific characteristics could support the weaker finding of columns (5) to (8).

21

4.2. Determinants of Deviations from planned Investment

Following, I discuss my regressions results regarding the determinants that affect the deviation from planned investment. In Table 5, the dependent variable is the Deviation from the planned investment for the next year, measured as the difference between Capex. As mentioned, it is divided into positive and negative deviations in order to see effects for under- and overinvestment. In this regression, I do not control for airline-specific measurements due to non-sufficient data for these variables.

Column (1) to (4) report the regression results with the positive deviation as the dependent variable. Positive deviation means that the company invested less in new airplanes as they committed to. The results in column (1) and (2) do not suggest empirical evidence for an impact of investment opportunity or incentive compensation on the reduction of investment plans. Column (3) captures the effect of managerial overconfidence. Here, the coefficient for the average percent moneyness of 93.86 is highly statistically significant at the 1% level. The implication of this finding is that as more overconfident CEOs are, as more they tend to underestimate their investment plans. Theoretically, this is not a straight-forward finding, since one would assume that overconfident do not experience to underestimate their committed investment, but rather the other way round. The theoretically more intuitive finding is given in column (7) and (8). The statistically and economically highly significant coefficient of 139.06 and 159.50 on the negative deviations suggest that overconfident managers plan too much investment. As a result, it seems like they have to adjust their plans due to unexpected or ignored factors, which would be in line with academic theory.

The results from column (4) suggest that Tobin’s Q has a weakly statistically significant negative impact on a reduction from the original investment plans at the 10% level significance level. The economic magnitude of this effect is significant. This finding provides weak evidence for the theoretical implication that an increase in investment opportunity leads to more deviation from planned investment. The deviation can be seen here as both negative and positive since in column (8) the results suggest that investment opportunity also matters for the negative deviation. The theoretical reasoning or implication behind is not straight-forward. The fact that higher investment opportunities matter for the positive deviation is in line with the theory since it suggests that a company that can increase its investment opportunity uses this change in its financial situation to invest more than planned. This could be especially interesting for financially constrained firms. However, the finding that investment opportunity impacts an

22

increase of investment compared to the committed investment negatively, is harder to explain and not in line with previous theoretical findings.

Regarding the impact of incentive compensation, the results are in line with the previous theory. As mentioned, incentive compensation does not matter for positive deviation. However, looking at the specifications (6) and (8), one can see that the “OnePercent” measurement matters for the negative deviation. In particular, an increase of one dollar more for an increase in stock price of 1% for a CEO increases the deviation from planned investment by the investment plans increase about $40,000 ($30,000). In other words, using the before mentioned example, $500 more in for an increase in stock price of 1% leads to an overestimation of $20 million $15 million. This suggests that compensation based on incentives leads to higher investment plans (based on the findings in the section before), but also to higher deviation towards lower actual investment. Putting this in the theoretical framework, it supports previous findings and implies that CEOs try to increase the shareholder value and therefore their own compensation by planning more investment which they might not be able to undertake.

Among all specifications, the R-squared is the highest for the models in which all three variables of interest are included, which suggest that these models fit well. To conclude, these results suggest three things: Investment opportunity impacts deviations from committed investment negatively, meaning that an increase in the level investment opportunity influences both an increase and a decrease of Capex relatively to the committed investment. Furthermore, managerial overconfidence affects deviations from planned investment positively, meaning that overconfident CEOs seem to plan worse. However, this seems to be the case for both under- and overinvestment. Lastly, Incentive Compensation matters only for negative deviations from planned investment. This suggests that incentive CEOs tend to plan more investment compared to what they actually undertake.

4.3. Financial constraint

In this section, I discuss how financial constraint affects the results analyzed above and what that means for the relevance of financial constraint for both investment planning and the deviations from it. As described in the methodology part, I split the data based on the SA-index of Hardlock and Pierce (2010) in two samples. This provides me with two subsamples of each 171 observations, defined as “less constrained” and “more constrained”. Furthermore, I report the I report the findings for the whole sample. For all the results, I do not include the

23

specifications for managerial overconfidence in the table, since these result did not show a significant impact. Table 6 repots the results of my empirical research.

For the whole dataset, i.e. the specifications (1) – (3), the committed investment increases in Tobin’s Q. The coefficient is significant at the 10% level and suggests findings consistent with the traditional Q-Theory: Investment opportunity is a main driver of corporate investment decisions. For the positive deviation, meaning an increase in investment compared to the committed plans, both the measurement for incentive compensation and investment opportunity have insignificant effects. However, the results for the negative deviation which implies a reduction of investment compared to the committed investment, show empirical significant results. Tobin’s Q has a negative impact on the deviation, whereas incentive compensation seems to reduce the actual investment compared to the original plans. Both effects are economically significant. These results and their implications are in line with the findings discussed in the previous section.

When splitting the sample into the two subsamples, there can only some sufficient findings be made. Incentive compensation measurement as a determinant is only significant for the less constrained group. That suggests that managers of financially constrained firms are not able to increase shareholder value in the way they want. As can be seen in the results in section 4.1 and also for the less constrained managers, there is empirical evidence for the fact that CEOs try to plan more investment in order to increase their own salary. Due to less access to the capital markets, this is not possible for financially constrained managers. This makes sense in connection with the positive coefficient for Tobin’s Q which implies together that investment opportunity matters for planned investment, but cannot be used for financial constraint firms.

Regarding the deviations from planned investment, one can see that for the constrained firm the coefficient of investment opportunity is highly significant and negative for the positive deviation. On the other hand, investment opportunity does not matter here for the less constrained firms. This finding can be interpreted as follows: Since financially constrained firms have less access to capital markets, they have in general more problems to commit investment. As a result, when investment opportunities increase for a constrained firm, it suggests that they immediately use this option to invest in new airplanes. This effect is likely to be weaker for unconstrained firms since they are not as dependent on external factors and therefore have fewer deviations from their plans.

24

5. Robustness Checks and Limitations 5.1. Robustness Checks

In this section, I show how robust my results found and described above are to changes in my baseline empirical.

In a first step, I adjust my empirical model by changing the sample time period. In particular, I investigate the determinants of corporate investment planning in a longer This means, I change the dependent variable from “Commitments for purchases in one year” to the period after, meaning “Commitments for purchases in two years”. I do this in order to see if my previous results might be affected by the planning horizon of the airlines. Table 7 shows the results for it. The impact of investment opportunity on the level of planned investment stays economically and statistically significant for the specifications without airline-specific measurements. Furthermore, it seems interesting that the results in columns (5) – (8) suggest similar implications regarding the negative correlation between managerial overconfidence and planned investment. That could suggest that these findings are not necessarily biased because of a too small sample size, but maybe due to a more important role of the industry-specific measurements.

Regarding the impact of incentive compensation on, the robustness test does not show statistically significant findings. This supports, on the one hand, the argumentation that the evidence before only provides weak evidence. On the other hand, it could also show a new finding, meaning that managers are more short-term focused and therefore incentive compensation only matters for the next period. The results for managerial overconfidence are in line with the findings for the committed investment in one year.

In Table 8, I use the deviation from planned investment in 2 years as the dependent variable to compare my findings from Table 5. There are a couple of remarks to make: On the one hand, investment opportunity appears to not matter anymore for deviations. However, my hypothesis did not have any expectations about the relationship between investment opportunity and deviations from planned investment. Incentive compensation keeps statistically significant influence on the level of negative deviations from planned investment. This suggests that incentive CEOs tend to plan more investment compared to what they actually undertake. However, now it also seems to have a positive effect on the positive deviations. This would suggest that incentivized CEOs tend to plan carefully their investment since their incentive compensation leads to deviations in which the company ultimately invests more than planned. Hence, these two results contradict each other.