Prudential regulation in an artificial banking system

Curto, José Dias; Quinaz, Pedro Miguel Mateus Dias Working Paper Prudential regulation in an artificial banking system Economics Discussion Papers, No. 2016-27 Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Curto, José Dias; Quinaz, Pedro Miguel Mateus Dias (2016) : Prudential regulation in an artificial banking system, Economics Discussion Papers, No. 2016-27, Kiel Institute for the World Economy (IfW), Kiel This Version is available at: https://hdl.handle.net/10419/142246 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Received March 21, 2016 Accepted as Economics Discussion Paper June 20, 2016 Published June 23, 2016 © Author(s) 2016. Licensed under the Creative Commons License - Attribution 4.0 International (CC BY 4.0) Discussion Paper No. 2016-27 | June 23, 2016 | http://www.economics-ejournal.org/economics/discussionpapers/2016-27 Prudential Regulation in an Artificial Banking System José Dias Curto and Pedro Miguel Mateus Dias Quinaz Abstract This study constitutes an exploratory analysis of the economic role of banks under different prudential frameworks. It considers an agent-based computational model populated by consumers, firms, banks and a central bank whose out-of-equilibrium interactions replicate the conjunct dynamics of a banking system, a financial market and the real economy. A calibrated version of the model is shown to provide an intelligible account of several recurrent economic phenomena and it can be a privileged ground for policy analysis. The authors’ investigation provides a relevant methodological contribution to the field of banking research and sheds new light into the role of banks and their prudential regulation. Specifically, the results suggest that banks are key economic agents. Through their financial intermediation activity, credit institutions facilitate investment and promote growth. JEL C63 G28 Keywords Agent-based computational model; financial intermediation; prudential policy; bank regulation Authors José Dias Curto, University of Lisbon, Lisbon, Portugal, [email protected] Pedro Miguel Mateus Dias Quinaz, University of Lisbon, Lisbon, Portugal Citation José Dias Curto and Pedro Miguel Mateus Dias Quinaz (2016). Prudential Regulation in an Artificial Banking System. Economics Discussion Papers, No 2016-27, Kiel Institute for the World Economy. http://www.economicsejournal.org/economics/discussionpapers/2016-27 1 1. Introduction Are banks key drivers of economic performance? Which are the impacts of different macro and micro-prudential regulations in economic growth? These seminal questions have resurfaced, in the aftermath of the 2008’s financial debacle, as a central topic in banking research and as an important source of concern for policy makers. In this study, we develop an agent-based computational model through which an exploratory analysis of the economic role of banks under different prudential frameworks is conducted. The years that preceded the 2008 crisis were characterized by a blatant phenomenon of reckless mortgage lending. Loan-underwriting criteria were so lenient that “subprime” borrowers with no capacity to redeem their debt had easy access to credit. These mortgages were pooled together and used to back collateralized debt obligations (CDOs). As soon as the housing markets collapsed, “domino effects” spread throughout the extremely interconnected financial system, quickly exposing severe fragilities. The price of mortgage-backed securities dropped astoundingly and allegedly safe CDOs were, in the end, valueless. The sale of such securities or its use as collateral became increasingly difficult. In an effort to rally liquidity, banks started to sell assets at fire-sale prices, giving rise to “price spiral” phenomena that, in turn, reduced banks’ capital thanks to “mark-to-market” accounting rules. When banks started doubting the solvency of their counterparties, wholesale agents stopped rolling over short-term debt, causing the failure of banks that had vigorously relied on non-stable sources of funding (Economist, 2013). The financial system proved to be incredibly fragile. Banks had expanded more than ever before, leveraging beyond reasonable levels and holding insufficient capital to absorb losses. It is therefore no surprise to note that the ex-post analysis of the failed mechanisms that led to the crisis has focused on banks’ raison d’être and on the regulatory instruments that can increase and promote financial stability. From a theoretical standpoint, banks have long been viewed as fundamental economic agents. Diamond and Dybvig (1983) suggest that banks are useful as liquidity providers since they provide depositors with liquidity insurance against idiosyncratic shocks that affect their consumption needs. In turn, Diamond (1996) suggests that banks provide monitoring services and help decrease asymmetric information between investors and firms. Banks also provide maturity transformation, a process whereby short-term liabilities are converted into long-term assets (Freixas and Rochet, 2008). The existing literature postulates that agency problems 2 (Dewatripont and Tirole, 1994) and the risk of a systemic crisis (Santos, 2000) are the main reasons why banks should be subject to regulatory requirements. Across time, several were the instruments (one of the most prominent ones being the deposit insurance scheme proposed by Diamond and Dybvig (1986)) suggested by academia to mitigate or prevent the negative effects of bank bankruptcies. In spite of this, and due to its function in reducing risk-taking and providing a buffer to absorb losses, the regulatory limelight always fell more heavily in bank capital (Allen and Gale, 2007). In this scope, Basel III/CRD IV rules, which are at present mandatory for Euro area banks, represent the latest regulatory effort to reinforce the robustness of the banking system. Besides the enhancement of minimum capital requirements to at least 10.5% of risk-weighted assets, it is worth highlighting Basel III’s introduction of a counter-cyclical capital buffer. However, the latest Basel Accord, is not completely consensual. Carmassi and Micossi (2012), for instance, argue that Basel III could follow a much simpler path in route to a more stable financial system simply by abandoning the risk weighting approach. The strong debate surrounding prudential policies, coupled with the blatant failure of macroeconomic models in the 2008 financial crisis, suggests that the study of regulatory policies is not within the grasp of the existing nucleus of mainstream economic knowledge. The notions of interdependence, networks, trust and expectations, which are key to understanding financial crisis, appear not to be “features of modern macroeconomic models” (Kirman, 2010: 501). To address these shortcomings, it is crucial that economists start considering the economy as a complex evolving system (Taylor, 2007). This can be done through agent-based modelling, a technique that allows researchers to simulate and comprehend out-of-equilibrium economic dynamics (Arthur, 2006). Agent-based computational economics encompasses, as illustrated by Tesfatsion (2006), a group of technologies that are best used in the analysis of complex, evolving systems composed by a multiplicity of agents who interact on the basis of behavioral rules that are exogenously defined. The incentive for this approach is rooted in the fact that complex systems, of which economies are flawless examples, can display emergent phenomena that cannot be foreseen or understood by any agent on an individual level. Instead of modeling economies under the omniscience assumption, according to which the decisions of agents are pre-coordinated, agents are programmed to follow simple behavioral rules that may or may not result in equilibrium. Through computer simulations, the behavioral patterns of the system can be 3 scrutinized and understood. Once a computer program able to replicate the phenomena of interest is developed, researchers can use it as a laboratory in which they can conduct policy experiments. On this vein, it is the purpose of this paper to assess, through the use of a rich, yet tractable agent-based model, the effectiveness and efficiency of prudential regulatory measures in their effort to maintain a stable and sound banking system. We add to the existing body of literature by creating a stylized agent-based computation model1 that simultaneously replicates the conjunct dynamics of a financial market, a banking system and the real economy, mimicking several stylized facts of the latter and grasping, to a certain extent, the implications of banking regulation in economic performance. The conceptual framework employed draws heavily from Takahashi and Okada (2003) and Tedeschi et al. (2012). For tractability, we adopt a parsimonious approach that drifts away from fully specified models such as the Eurace project (Raberto et al., 2011; Cincotti, 2012). Specifically, our model attempts to depict, in an admittedly stylized form, the inner workings of a small economy composed by consumers, firms, commercial banks and a central bank. Consumers are responsible for providing the work force needed for production and for driving internal demand through consumption. Firms are profit-maximizing entities that produce the consumption goods and are thus responsible for driving supply. The primordial function of banks is to collect deposits from their clients and to give out loans to consumers and companies. Banks are, in addition, subject to prudential regulations established by the Central Bank (i.e., minimum capital ratios and minimum loan underwriting standards). Following Ashraf et al. (2011), the model is calibrated to U.S. data. Simulations are performed repeatedly, for several time periods and under diverse parameter settings in order to assess how credit institutions impact macroeconomic performance and how performance, in turn, is impacted by diverse regulatory frameworks. The model we present aims to be a valuable methodological contribution to the field of banking research, shedding new light on the role of banks and bank regulations by leveraging on the 1 The computer program under the agent-based computation model has been developed in NetLogo, an agentbased programming language and integrated modeling environment. 4 advantages of agent-based computational modelling. In spite of being too simple to befit the policy-making endeavor, it does generate five results of broad-spectrum qualitative interest. The first result corroborates the view that credit institutions are crucial economic agents in so far as their existence significantly improves the performance of the economy. Banks foster growth by providing credit and expanding firms’ investment opportunities, which are usually limited to each company’s ability to endogenously generate cash flows. The capacity of the economy supply-side to meet aggregate demand is thus significantly improved when a banking system is in place. The role of credit institutions is also relevant in mitigating the negative externalities of firm failures. By providing financing to new entrants and sustaining incumbent firms by preventing capital erosion, banks are often able to alleviate the effects of shocks. Our second result reveals that the macroeconomic impact of credit institutions and bank regulation frequently contradicts the traditional dogmas behind micro-prudential policies. In particular, the experiments conducted show that aggregate macroeconomic performance deteriorates as banks become safer (i.e., as stricter capital requirements are imposed). This is because decreased credit availability hampers the creation of value by reducing the speed at which supply is able to meet surges in demand. Furthermore, increased leverage does not always entail augmented firm bankruptcies. If credit levels are not excessive, bank funding is able to prevent the depletion of firms’ production capacity in periods of low demand, thus contributing to a reduction of bankruptcies in the long run. Finally, increased capital requirements do not always mean that banks are less likely to fail. Indeed, micro-prudential policy is shown to entail calibration risk: while increasing capital requirements increases the immediate loss-absorbency capacity of banks, it also decreases their ability to endogenously generate capital through profits. The third result shows that the role of banks (in particular if risky) is more relevant when the economy is performing significantly below optimal levels (i.e., when the economy displays large output gaps). According to our policy experiments, the output gap is closed faster if banks are allowed to require lower financial autonomy ratios from customers and are subject to lower capital adequacy ratios. In other words, credit is proved to be more important when the output gap is larger. A corollary of this result is that using the credit-to-GDP ratio as a trigger to the implementation of the counter-cyclical capital buffer might be, if taken single-handedly, an insufficient approach. Since the same level of leverage may be too high or too low depending on the current 5 state of the economy (i.e., the magnitude of the output gap), the establishment of a countercyclical capital buffer must also consider the impact of credit-constraining measures in GDP growth and an assessment of the level of debt that can be sustained by the economy without causing the build-up of system-wide risk. More importantly, our fourth result argues, under the premises of our model, that the countercyclical capital buffer should never actually be implemented. Indeed, our experiments fail to find evidence in support of the rationale behind macro-prudential tools. This conclusion is not surprising if considered conjunctly with previous insights, which showed that riskier banks promote growth in fundamentally all scenarios. Finally, our fifth result suggests that effective resolution is crucial. Specifically, resolution of credit institutions acts to the benefit of the economy by ensuring depositors are never bailed-in to a great extent and by getting rid of “zombie-banks” that are not able to support investment. The paper is organized as follows. Next section discusses the conceptual framework of the model and describes the protocol and behavioral rules imposed on its interacting agents. The calibration procedure is also clarified in this chapter. Section 3 describes the core results of our policy experiments, providing insights into the role of banks and the impact of micro and macro prudential policies in economic performance. It also addresses the robustness of the results by examining a scenario where the negative consequences of bank failures are magnified. Section 5 summarizes our concluding remarks 2. A Stylized Model of the Banking System Conceptual Framework The conceptual framework employed draws heavily from Takahashi and Okada (2003) and Tedeschi et al. (2012). The model depicts, in an admittedly stylized form, the inner workings of a small economy composed by I consumers, E companies (also referred to as firms), B commercial banks and a central bank. Consumers are indexed by =1,…, , companies are indexed by =1,…, and banks are indexed by =1,…, . Since there is no government, there are also no taxes nor public expenditure. In this economy, there is only one kind of product valued by the population – the consumption good. Consumer goods are perishable and have their priced fixed at unity. 6 Two types of production factors exist: labor and capital. Consumers provide labor and receive compensation from companies in return. Based on their disposable income, net worth and bank deposits, consumers determine how much to spend on the consumer good. Companies decide how much to produce based on their own production capacity and expectations of demand. In this process, firms can invest in infrastructures and technology, increasing their productivity. In general, supply and demand (both internal and external) for the consumer good do not balance, giving rise to demand rationing (excess demand) or production waste. Economic agents can also invest in financial assets (i.e., bonds), with each debt security generating interest income at each time step. This asset is finite and exogenous to the economy, and can therefore be interpreted as a debt security issued by a sovereign country. Each consumer and company can buy and sell the financial asset in the market. All transactions must be mediated by bank deposits, and proprietors are not able to lend and borrow among themselves. Each bank grants credit and accepts deposits from their customers, with depositors being allocated randomly across banks so that each credit institution starts off with the same number of clients. Transaction settlements are executed by changing the holder of the bank deposit. For a more comprehensive view of the economy’s agents, a cursory description of their core behavioral features and characteristics is provided below. 1.1.1. Consumers Consumers are responsible for providing the work force needed for production (for which they receive their wage ()) and for driving internal demand through consumption. They are members of the co-operative firms in which they are employed and are thus entitled to a share of their profits. In addition, consumers are also a big part of the financial asset market since they are able to buy and sell debt securities. At each time step , the patrimony of consumer  is composed by  units of the financial asset and a bank deposit (,). Since companies may not always be able to fulfill their commitments to their employees, circumstances might happen when the consumer is also owed a part of its salary (). To finance the acquisition of the asset, consumers can make use of bank loans (). The value of the financial asset is marked to market at each time step: 7 = (1) Hence, the net worth (i.e., the net equity) of each consumer equals: =+,+, (2) The financial autonomy ratio of consumer  can thus be represented as: = ++ (3) Finally, each consumer displays, at each time step , a balance-sheet similar to the one depicted in Figure 1. 1.1.2. Companies (also referred to as Firms) Companies are profit-maximizing entities of a co-operative nature (i.e., owned by their employees) that produce the consumption good and are thus responsible for driving supply. The production activity entails the usage of two production factors: labor and capital. While the amount of labor is fixed (i.e., each firm has  employees), the amount of capital can vary based on investments made by the firm in order to increase its production capacity. Companies are also assumed to be able to invest in the financial asset. At each time step , company  owns  units of the financial asset, a bank deposit (,) and  of productive capital. In order to finance the acquisition of the asset or increase the production capacity through investment, companies can make use of bank loans (). Finally, and since companies may not always be able to fully remunerate employees for their work, circumstances might happen when delayed wages are accumulated (∑,    , where , represents the amount of delayed wages between company  and consumer ). Taking into account that the value of the financial asset is marked to market at each time step, the net equity of each firm equals, at time step :   =    + , +   ,  ,   (4) The financial autonomy ratio of company  can be represented as: 14 5. Assessment of companies’ financial position After trying to sell their production in the market, companies assess their cash in-flows and start making their payments. At this point, each firm’s revenues3 are given by the remuneration generated by each unit of bonds owned and the minimum between real demand and the company’s production:    =min∑ 1  =1 +  1 ;   +×  ,+×  × (16) where  is the coupon rate of the debt security and  represents its nominal value. On the other hand, cash outflows4 are given by:  =  +×+,×,+  (17) where  × represents personnel costs, × represents funding costs and   represents the loan’s principal repayment. Whenever cash outflows are greater than cash inflows plus the company’s deposits, the firm gives priority to the payment of wages. Bank loans are assumed to be paid last. Amounts not paid are registered as liabilities and need to be paid in the following time step. In an effort to re-equilibrate its financial position, the company will also try to sell any bond it owns. Finally, the company net worth is updated as follows: = +    (18) 3 Firm’s revenues also include interest received on deposits and potential cash-inflows stemming from the sale of debt securities. At this point, however, it is not possible to know whether the revenue generating capacity of the company’s bank is enough to fulfill the totality of its commitments. Interest revenue is thus added to the firm’s net worth after the computation of the bank’s financial position (please refer to stage 8). In addition, proceeds from the sale of debt securities are added to the firm’s bank deposits whenever a transaction is settled (please refer to stage 10). 4 Firm’s cash-outflows also include potential payments stemming from the purchase of debt securities and from the distribution of dividends. Disbursements associated with the acquisition of financial assets are deducted from the firm’s bank deposits whenever a transaction is settled (please refer to stage 10). Cash-outflows associated with dividend payouts are deducted from the firm’s net worth at the end of each time step (please refer to stage 11). 15 6. Companies’ bankruptcy and entrance As already stated, banks will liquidate debtors as soon as they miss  payments. In such situations, the remaining company’s assets are ceased by the bank and the player leaves the market. The model assumes a simple entry mechanism based on one-to-one replacement. As such, firms that go bankrupt are automatically replaced by new players. In accordance with the empirical literature (Bartelsman and Scarpetta, 2005), new entrants are usually smaller than existing firms. Specifically, the stock of capital of new firms is drawn from a uniform distribution with =min(,,…,) and =median (,,…,). 7. Assessment of consumers’ financial position At this stage, the income of consumer  in period  () 5, is a function of his wage and the amount of bonds it owns. This can be represented as: =+××+× , (19) Note that the consumer might not always receive the totality of his wage, as the company for which it works for might be in financial distress. On the other hand, the cash outflow6 of each consumer at time step  is given by:  =+×, (20) where  represents consumption and ×, represents funding costs. Whenever cash outflows are greater than income plus the consumer’s deposits, the consumer gives priority to the liquidation of consumption expenses. Bank loans are assumed to be paid last. Amounts not paid are registered as liabilities and need to be paid in the following time 5 Consumer’s income also includes interest received on deposits and potential cash-inflows stemming from the sale of debt securities and from firm dividends. At this point, however, it is not possible to know whether the revenue generating capacity of the consumer’s bank is enough to fulfill the totality of its commitments. Interest revenue is thus added to the consumer’s net worth after the computation of the bank’s financial position (please refer to stage 8). In addition, proceeds from the sale of debt securities are added to the consumer’s bank deposits whenever a transaction is settled (please refer to stage 10). Cash inflows associated with firm dividends are added to the consumer’s net worth at the end of each time step (please refer to stage 11). 6 Consumer’s cash-outflows also include potential payments stemming from the purchase of debt securities. Such disbursements are deducted from the consumer’s bank deposits whenever a transaction is settled (please refer to stage 10). 16 step. In an effort to re-equilibrate its financial position, the consumer will also try to sell any lot of land it owns. Finally, the consumer’s net worth is updated as follows: = +  (21) 8. Assessment of banks’ financial position At each time step, banks will assess their cash-flows in order to determine the robustness of their financial position. Cash inflows7 consist fundamentally of interest income and are thus a function of the debtors’ ability to service their debt and the rate of remuneration of the bank’s claims on other (foreign) credit institutions:    =  × ,    +  × ,     +  ×  (22) In accordance with commercial banks’ business models, cash outflows8 consist of interest expenses related with the remuneration of deposits and the payment of interest on interbank loans:    =  × ,    +   × ,     (23) Liquidating distressed borrowers is needed for banks to raise liquidity, thus playing a crucial role for banks’ financial stability. As already stated, banks will liquidate debtors (consumers or companies) as soon as they miss  payments. In this process, credit institutions take over any bond still owned by the debtor9 and immediately try to sell it on the market. The transfer of the collateral’s ownership implies a direct write off of the bad loan. 7 Banks’ cash-inflows also include potential revenues stemming from the sale of debt securities. Such proceeds are added to banks’ deposits whenever a transaction is settled (please refer to stage 10). 8 Banks’ cash-outflows also include potential payments stemming from the purchase of debt securities. Such disbursements are deducted from the banks’ bank deposits whenever a transaction is settled (please refer to stage 10). 9 Note that the debtor will not have any asset at this point, since all deposits/cash have already been used to try and pay off its debts. 17 Whenever cash inflows are greater than cash outflows plus the bank’s cash holdings, it is clear that the credit institution is able to fulfill the totality of its commitments. The net worth of both companies and consumers is thus increased by their respective remuneration on deposits: =+× , (24) =+× , (25) Finally, the bank net worth is updated as follows: = +    +× (26) where  stands for the nonperforming loans that were written off by bank  in time step  and  represents the number of debt securities foreclosed during the same time step. 9. Banks’ bankruptcy and entrance Bankruptcy is assumed to happen to banks whenever credit institutions are not able to fulfill the totality of their commitments (i.e., whenever cash outflows are greater than cash inflows plus the bank’s cash holdings) or whenever their capital ratio is below zero. The model thus captures the possibility for both liquidity and solvency related default events. In the event of bankruptcy, the credit institution enters into resolution. As part of this process, non-performing loans (i.e., loans with >0) are foreclosed. Debtors that were not able to fully pay their obligations to the bank see their holdings confiscated. Since the collateral’s value may not be enough to fully compensate non-performing loans outstanding, the bank’s capital might be further reduced during this process. After all the non-performing loans are foreclosed, the capital position of the bank is assessed. Losses are first and foremost absorbed by equity capital. Depositors are bailed-in whenever equity capital is negative. Since all depositors have the same level of seniority, losses are distributed among all clients proportionally to the amount of deposits they own. Similarly to the process described in point 6, a recapitalization of the bank using foreign capital is assumed. Since only healthy assets are now left in the bank, this does not constitute a farfetched scenario. The amount of the recapitalization is equal to the maximum between the endowment initially given to all banks and the amount of capital needed for the bank to display 18 a capital adequacy ratio 12% in excess of minimum requirements. When computing the latter, a haircut ( ) is applied to the value of the financial assets. 10. Purchase and sale of the financial asset Supply and demand for the financial asset are generally determined by how attractive the bond is relative to other financial assets (loans). For each agent, the attractiveness of the debt security in period , denoted by , is given by: =+(2 ) (27) where   = × + (  )  (28) and ~  0  1 (29) The first term on the right hand side of equation (28) represents the expected income accrued from owning one unit of the financial asset. The second term captures the expected capital gain perceived by the trend chasers augmented by , the strength of the trend chasing attribute. Expected capital gains in equation (28) reflect the change in the asset price over the last   periods. Even though all traders are trend chasers, the capital gain term is weighted by a random variable () to capture divergent expectations. The gains from holding debt securities (first term of equation (27)), minus loan interest (denoted by the third term), measure the net marginal gain from owning an additional bond. In reality, other factors affect purchase and sale decisions: the second term of equation (27) represents such factors. This second term further helps traders forming different expectations. The owner attempts to sell one lot of land when it judges land to be less attractive (i.e., less than ), or when it is pressed to sell land due to a financial distress situation. For ≥1, supply of land () is such that: =1 0  < >0 <0  (30) 19 Economic agents, unless they are debtors of nonperforming loans, will try to make a purchase of one unit of the financial asset if having the additional unit is attractive enough: =1 0  >   (31) After defining its interest in buying the financial asset, every agent needs to determine if it needs bank financing. Buy-side players’ intention to acquire debt securities is limited to the purchase of one unit. After each agent determines its market positioning, supply and demand meet randomly, resulting in transactions. The short side of supply or demand determines the actual quantity of units traded. Agents that manage to settle their transactions see their financial position updated accordingly: =×+× (32) where  represents the amount of debt securities purchased and  represents the amount of debt securities sold. In general, prices are assumed to be sticky. The price therefore responds to the difference between supply and demand (), with  representing the speed of the price adjustment: =+ (33) where   =  ×∑∑ max(∑,∑) (34) 11. Distribution of profits At the end of each time step, firms, in accordance with their cooperative nature, distribute dividends. The payout ratio is a function of both the firm’s current cash holdings and the outflows that the company expects to face in the next period. Specifically, expected disbursements reflect each firm’s forecasted personnel and investment costs. Mathematically:   =max  ,   ×   +× []   ×,0 (35) 20 where []  represents next period’s expected demand,   represents next period’s production capacity and  designates each firm’s propensity to fund new investments with equity (i.e., the percentage of cash withheld to fund new investment opportunities or, in other words, the target capital structure of greenfield projects). Naturally, and in case the distribution of profits is consentaneous with the liquidity position of each company, the net worth of each of its associates must be updated with their respective share of the dividend:   =  +  + (36) Finally, each company’s net worth is updated as follows: = (37) 21 Model calibration As demonstrated above, the model has multiple individual agents. In order to restrain the number of parameters and to facilitate calibration, ex ante symmetry between agents was imposed. In spite of the homogenous initial setting, the economy develops heterogeneous behaviors and results due to the impact of feedback effects and noise. The calibration procedure follows the methodology described in Ashraf et al. (2011) and assumes each time step represents one month. The 39 parameters of the model were classified as consumer parameters, firm parameters, bank parameters, financial market parameters and general parameters. These are listed in Table 3 along with the assigned values under the baseline scenario. The calibration of these parameters encompassed three different stages. During the first stage, one subgroup of parameter values was chosen based on empirical U.S. data or the values used in previous studies. The results of the extensive simulations performed identified specific parameters that tended to converge, on average across simulations, to a specific value. The identification of such parameters comprised the second stage of the calibration process. Finally, and during the third stage, the values of the outstanding parameters, for which no suitable empirical evidence was found, were chosen in order to (i) guarantee that the model replicates some of the most well know stylized facts of real economies and (ii) make the model’s median outcome across simulations of specific outputs (e.g., GDP) loosely match the properties of U.S. data. 2.3.1 First stage of the calibration Consumer parameters. Following Takahashi and Okada (2003):  The consumer’s propensity to consume out of disposable income () was set to 0.95;  The consumer’s propensity to consume out of net wealth () was set to 0.01;  The initial wage of each worker () was set to 500. Analogously to the volatility of the shock to aggregate demand defined by Tedeschi et al. (2012), the volatility of the shock to each agent’s consumption () was set to 0.1. Firm parameters. In accordance with Tedeschi et al. (2012):  Capital productivity () was set to 0.1; 22  Each firm’s risk aversion coefficient when submitting loan applications () was set to 1. Bank parameters. As defined by Tedeschi et al. (2012):  Each bank’s risk aversion coefficient used for the purposes of defining the spread on granted loans () was set to 1;  The standard maturity of every loan () was set to 12 (i.e., 1 year). As foreseen in the Basel III capital accord, the minimum capital adequacy ratio ( ) was set at 8% (capital buffers not included). The risk-weight () on non-risk-free assets is approximated by the average Basel II RWA density10 reported by Le Leslé and Avramova (2012). The risk-weight attributable to claims on other credit institutions () is set at 20%, which is the weight foreseen in CRD IV for claims on banks belonging to the best credit quality step. Finally, and based on Banco de Portugal’s definition of credit-at-risk11, the number of missedpayments necessary for the bank to trigger the liquidation of a debtor was set at 3 (i.e., 90 days). Financial market parameters. FollowingTakahashi and Okada (2003):  Each agents’ propensity to form divergent expectations regarding the financial asset price (as measured by  and ) were set to 0.5 and 0.2, respectively;  The strength of the trend chasing () was set to 1;  The attractiveness threshold for the financial asset () was set to 0.01;  The price adjustment speed () was set to 0.1. General parameters. The long-term growth expectation of internal and external demand () was set at 0.26% based on the monthly growth rate of the US economy between 1947 and 2014 according to the data series available at the Federal Reserve Bank of Saint Louis. To estimate 10 Percentage of RWAs over total assets. 11 The concept of credit at risk was initially defined by Banco de Portugal in Instruction No 22/2011. Credit at risk corresponds to the following elements as a whole: (a) Total amount of outstanding loans with principal installments or interest overdue for a period of 90 days and over. Non-contracted current account claims should be considered as credit at risk 90 days after an overdraft is recorded; (b) Total amount of outstanding restructured loans not covered by the preceding sub-paragraph, whose installment or interest payments, overdue for a period of 90 days and over, have been capitalized, refinanced or their payment date delayed, without an adequate reinforcement of collateral (this should be sufficient to cover the total amount of outstanding principal and interest) or the interest and other overdue expenses that have been fully paid by the debtor; (c) Total amount of credit with principal installments or interest overdue for at least 90 days, but on which there is evidence to warrant classification as credit at risk, notably a debtor’s bankruptcy or winding-up. 23 the growth rate, and taking into account that our model does not consider inflation (i.e., the price of the consumer good is fixed at unity), the data series for the real gross domestic product12 was used. 2.3.2 Second stage of the calibration During the calibration phase of the model, the price of the debt security () was found to show mean reverting behavior. As such, its initial value was set to 2055 (i.e., the median value of the cross-run averages of the security’s price). 2.3.3 Third stage of the calibration After the first and second stages, 20 parameters still required calibration. To define them, the behavior space of the model was analyzed in detail. The model was run several times while systematically varying its settings and recording the results of each run. This "parameter sweeping" process allowed us to explore the model's potential behaviors and identify which combinations of settings caused the phenomena of interest. Behaviors of interest were identified whenever the model was able to (i) replicate some of the most well know stylized facts of real economies and (ii) make the median outcome across simulations of specific outputs (e.g., GDP) loosely match the properties of U.S. indicator variables. In particular, 1000 simulations of 40 years (i.e., 480 time steps) were performed. For each run, the average of each indicator variable was computed using only the last 35 years to get rid of transients and capture only the system’s stochastic steady state. Finally, the median of the simulation averages was computed. The only exception to this methodology was the calculation of GDP related variables, which were assessed based on a time series composed of the crossrun average of the economy’s domestic product at each time step. With regards to the real values of U.S. data used, it is important to note that:  GDP’s volatility is the standard deviation of the detrended HP-filtered log GDP series for the period 1947-2014. As suggested by Ravn and Uhlig (2002) for quarterly data, the value of the multiplier parameter was set at 1600. The autocorrelation of this variable was computed by estimating an AR(1) process over the same time period.  The default rate is the value-weighted average of rated corporate bond issuers that 12 Inflation adjusted (billions of chained 2009 dollars), seasonally adjusted, quarterly value of the goods and services produced by labor and property located in the United States. 30 3. Simulation and Results As pointed out in the introduction of the dissertation, this study focuses on developing a model that replicates the conjunct dynamics of a financial market, a banking system and the real economy, mimicking several stylized facts of the latter and grasping, to a certain extent, the implications of banking regulation in economic performance. Specifically, and based on the model described in section 3, policy experiments are designed to answer the following seminal questions: 1. Are banks key drivers of economic performance? 2. What impacts do different micro-prudential regulations have in economic growth? 3. Are macro-prudential policies effective promoters of financial stability? In line with the methodology used to calibrate the model, each experiment encompassed 1000 simulations of 40 years (i.e., 480 time steps). For each run, the average of each indicator variable was computed using only the last 35 years to get rid of transients and capture only the system’s stochastic steady state. Finally, the median of the simulation averages was computed. The only exception to this methodology was the calculation of GDP related variables, which were assessed based on a time series composed of the cross-run averages of the economy’s domestic product at each time step. 4.1 The role of banks From the theoretical standpoint described in section 2.1.1, and following the seminal paper of Diamond and Dybvig (1983), banks play a crucial role in economic growth by pooling liquidity from depositors (who prefer to have their savings placed in liquid instruments) and channeling the funds to firms (who require long term, large sum investments in order to generate returns in the future). This view is corroborated by the policy experiments of our model. To assess the importance of banks in economic performance, a comparison was made between the most relevant economic aggregates resulting from simulated economies with and without banks. Specifically, we compare the model results in the baseline calibration with the model results in a scenario where banks are shut down. Operationally, the behavior of banks was adjusted in order to require from any loan applicant a financial autonomy ratio larger than 100% (which is naturally not 31 feasible). Banks were thus turned, in this scenario, in mere operators of the economy’s payments systems (which are, nevertheless, a crucial infrastructure of modern economies). From the results laid down in Table 4 it is clear that the existence of banks improves economic development. Indeed, all measures display a manifest degradation in median performance when banks are suppressed. The reason for these phenomena is related with the inner workings of the modelled economy. Since no aggregate equilibrium relationship is forced between the agents’ actual and expected demand, out-of-equilibrium dynamics are created at the level of each specific agent. Since there are no market-clearing mechanisms, the economy spontaneously self-organizes towards a state in which demand persistently exceeds supply. Economic growth in our model is thus a function of the firms’ capacity to increase production capacity through investment. The inexistence of banks is a clear obstacle to this process, since in this scenario companies’ investment capacity is limited to their ability to endogenously generate capital. As such, the capacity of the economy to meet aggregate demand is severely hampered when there is no banking system, as reflected in the difference of the output gap (here defined as the difference between potential GDP if all demand was met and actual GDP) in both scenarios (3.31% in the baseline calibration versus 7.35% in the setting with no banks). This mechanism is also the reason why GDP volatility is smaller in environments where credit institutions are shut down (0.86% in the baseline calibration against 0.59% in the setting with no banks). The fact that capital investments cannot be leveraged through bank financing means that production responds less quickly to increases in expected demand. Firms adjust more slowly and smoothly to increases in consumption and exports, which in turns reduces GDP’s volatility. In addition, firm failures are, in our baseline calibration, mostly attributable to potential gaps between expected and actual demand. These gaps can create an unexpected shock to profits which, associated to firm’s leverage, may render institutions unable to fulfill their commitments. It would thus be expected that an economy without banks (and hence without credit) would result in decreased default rates. In a very interesting result, however, the default rate of firms is higher under the scenario where credit institutions are shut down (1.49% in the baseline calibration against 5.20% in the setting with no banks). This curious feature of the model is attributable to the fact that firms face a 32 monthly capital depreciation rate. Without banks to finance reinvestment, firms are often unable to secure the amount of cash needed to prevent capital erosion. Ultimately, the depreciation of capital reduces firms’ production capacity to the extent of making them unable to achieve break-even, in which case firms default due to their inability to pay wages. Based on this, it is possible to conclude that banks do not only foster growth, they also contribute to the maintenance of the current level of wealth as measured by an economy’s production capacity. Table 4 – Banks vs. No Banks Banks No Banks Output gap 3.31% 7.35% GDP volatility 0.86% 0.59% GDP autocorrelation coefficient 0.93 0.92 Firms' default rate 1.49% 5.20% Banks' default rate 0.57% - Source: Authors 4.2 The impact of micro-prudential policies The impact of micro-prudential regulation is gauged by considering alternative scenarios where banks are either riskier or safer than in our baseline calibration. In the risky scenario:  Loan underwriting policies are looser. By imposing a minimum financial autonomy ratio of just 0.3, banks implement lenient credit quality requirements when assessing firms and consumers, thus significantly increasing the universe of loan-eligible agents;  The required capital adequacy ratio is diminished to just 2%, which significantly increases the capacity of credit institutions to provide loans. In the safe scenario:  Loan underwriting policies are stricter. By imposing a minimum financial autonomy ratio of 0.9, banks implement stricter credit quality requirements when assessing firms and consumers, thus significantly decreasing the universe of loan-eligible agents;  The required capital adequacy ratio is augmented to 15%, a much more creditconstraining scenario than the one imposed in the baseline calibration (8%). 33 Table 5 – Safe Banks, Regular Banks and Risky Banks Safe Banks Regular Banks Risky Banks Output gap 3.66% 3.31% 3.26% GDP volatility 0.74% 0.86% 1.19% GDP autocorrelation coefficient 0.93 0.93 0.94 Firms' default rate 1.83% 1.49% 1.66% Banks' default rate 0.57% 0.57% 1.14% Source: Authors As Table 5 clearly suggests, aggregate macroeconomic performance deteriorates has banks become safer. Indeed, the median output gap is higher in scenarios where prudential policies are looser (the output gap stands at 3.26% in the risky scenario, which compares to 3.31% in the baseline calibration and 3.66% in the setting with safe banks). This result corroborates our prior conclusions, according to which increased credit availability improves the creation of value by increasing the speed at which supply is able to meet surges in demand. In the same vein, it is also clear that the mechanisms contributing to GDP volatility are magnified when credit is less constrained. On the other hand, it is interesting to recognize that the default of firms responds non-linearly to changes in the micro prudential framework. While the scenario with safe banks displays the highest default rate (1.83%), thus confirming that credit is important for firms to stay afloat in periods of low demand, the scenario with risky banks is not the best performer (the default rate is 0.17% higher than in the baseline calibration). This result suggests that there is an inflexion point after which less credit constraints actually contribute to a degradation of firms’ financial robustness. Beyond a certain indebtedness level, and in a manifestation of the financial accelerator effect postulated by Bernanke et al. (1996), increased firm leverage (see figure 9) makes credit institutions more likely to face distress (due to increased client bankruptcies and credit losses), which in turn increases companies’ likelihood of failure (due to foreclosures by failing banks and credit rationing by low-capitalized institutions). Also remarkable is the non-linear tradeoff between firm defaults and the output gap. Given that new entrants are smaller than incumbent companies, it would be expected that more firm failures would entail decreased economic performance. This thesis is repudiated by the risky scenario, where the greatest amount of defaults coincides with the lowest median output gap. Such a result clearly indicates that firm failures are less likely to slow down the economy when 34 firms have easy access to credit, since this source of financing allows entrants to quickly catch up with production requests. Figure 9 – Credit-to-GDP ratio (cross-run average) .1 .2 .3 .4 .5 .6 .7 .8 .9 100 150 200 250 300 350 400 450 Credit-to-GDP - Regular Banks Credit-to-GDP - Risky Banks Credit-to-GDP - Safe Banks Source: Authors As expected, the decrease in the capital-adequacy ratios of banks is automatically reflected in increased default rates among credit institutions. On the other hand, increased capital-ratios are not automatically reflected in safer institutions, as seen by the fact that the median bank bankruptcy rate is virtually the same in the scenarios with safe and regular banks. This result indicates that micro-prudential policy entails calibration risk: while increasing capital requirements increases the loss-absorbency capacity of banks, it also decreases their ability to endogenously generate capital through profits. In addition, and despite the fact that bank and firm failures are much higher in the risky scenario than in the baseline setting, value-creation actually improves. This phenomenon stems from the fact that there is an important circuit-breaker that prevents the modelled economy from suffering terribly with bank-bankruptcies. The effective resolution of credit institutions, which are terminated immediately after entering into technical insolvency (i.e., after displaying capital adequacy ratios below 0), means that depositors are not usually bailed-in to a great extent. As such, bank bankruptcies represent salutary events for the economy, which performs better after getting rid of “zombie-banks” that are not able to support investment due to the fact that they do not comply with micro-prudential requirements. 35 Relevant insights can also be drawn from analyzing the behavior of the output gap in time. During the initial stages of the simulations, the gap between production capacity and aggregate demand is large in all scenarios (please refer to Figure 10). In the setting with risky banks, the ability of firms to tap onto external sources of funds is translated into increased leverage and bankruptcies, which in turn contributes to a high output gap when compared to the remaining scenarios. However, as demand grows, firms become able to sustain supplementary levels of debt. The output gap is thus closed much quicker in the scenario where banks are riskier. Figure 10 – Output gap (cross-run average) .00 .04 .08 .12 .16 .20 .24 .28 .32 .36 100 150 200 250 300 350 400 450 Regular Banks Risky Banks Safe Banks Source: Authors In the long run, growth in the economy is ultimately capped by the growth of consumption and exports. Since the economy does not, on average, experience extraordinary spurs of growth, firms are usually able, in all scenarios and towards the end of the simulations, to raise the necessary capital (either endogenously or through bank financing) to meet demand. As such, the output gap in different prudential frameworks tends to converge to the same value as time goes by. 36 Figure 11 – Firm defaults at each time step (cross-run average) .0 .1 .2 .3 .4 .5 .6 .7 100 150 200 250 300 350 400 450 Firm Defaults - Regular Banks Firm Defaults - Risky Banks Firm Defaults - Safe Banks Source: Authors In addition, and since labor costs are fixed in our model, the increase in the size of firms means they are more easily able to sustain additional financing costs, which in turn is reflected in a substantial decrease of the number of firm defaults towards the end of the simulations (please refer to Figure 11). A corollary of these results is that the role of banks (in particular if risky) is more relevant when the economy is performing significantly below optimal levels (i.e., when the economy displays large output gaps). Finally, and with respect to micro-prudential policy, it is also relevant to disentangle between the effects of changes in the capital framework and modifications to the loan underwriting criteria of banks. To analyze this, a sensitivity analysis is conducted. Starting from the model’s baseline calibration, we vary, ceteris paribus, the minimum capital requirements of banks and the minimum financial autonomy ratios demanded by credit institutions from their customers. In order to guarantee that the stochastic properties of the model do not affect results, the simulation is performed using the same seed value for the random number generator. Figure 12 and Figure 13 display the differences in the output gap between each scenario and the baseline setting (depicted by the red line). 37 As can be seen, changes to the capital requirements have a severe impact on economic performance. In a clear validation of previous results, it is clear that the output gap is persistently higher (lower) when minimum capital requirements are greater (smaller). On the other hand, it is also clear that changing the loan underwriting policies has a reduced impact on economic activity. In spite of this, and while there are no relevant differences in median performance between the baseline scenario (median output gap of 2.74%) and the scenario where the minimum financial autonomy ratio is set at 30% (median output gap of 2.91%), it can be concluded that increasing the strictness of loan underwriting criteria tends to degrade macroeconomic performance (median output gap of 4.09%). Figure 12 – Output gap under varying capital requirements (seed -204145716) -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 50 100 150 200 250 300 350 400 450 CAR = 0.02 CAR = 0.08 CAR = 0.14 Figure 13 – Output gap under varying financial autonomy requirements (seed -204145716) -.15 -.10 -.05 .00 .05 .10 .15 50 100 150 200 250 300 350 400 450 FAR = 0.3 FAR = 0.7 FAR = 0.9 Source: Authors 38 4.3 An exploratory analysis of macro-prudential policies As already discussed, the recent financial crisis exposed a vicious circle whereby difficulties in the banking system can prompt a recession in the real economy that then feeds back on to the financial sector. This phenomenon suggests that banks should increase their capital buffers in periods where systemic risk is greater. The countercyclical buffer was designed to guarantee that the level of capitalization of the banking sector is consistent with the macroeconomic environment in which credit institutions operate. It should be enforced when the Authorsities acknowledge that the levels of credit in the economy are excessive and associated with systemic risk, and is targeted at guaranteeing that banks have sufficient capital to withstand potential future losses in case of a bust. To test the effectiveness of the counter-cyclical capital buffer, we depart from the baseline scenario by establishing a setting in which the regulator enforces an additional capital surcharge for banks whenever the credit-to-GDP ratio of the economy rises above 35% (which is the average value of the indicator in the baseline cross-run average time series). The buffer is then maintained at least for one year, with the possibility of extension should the economy’s leverage remain above the threshold. The results laid down in Table 6 indicate that imposing a capital buffer of 2.5% (which is the value foreseen in Basel III) or 5% does not have a significant impact on the overall dynamics of the economy. The only visible benefit stemming from this measure is a very mild reduction in the median amount of firm defaults. However, and similarly to the results obtained in the “safe banks” scenario, this is not reflected in persistently lower output gaps (in fact, the opposite occurs). As already explained, this is underpinned by the fact firm defaults appear not to have a significant impact on economic growth (in spite of new entrants being smaller than incumbent firms). Indeed, increased access to credit allows new players to quickly catch up with production needs, a phenomenon that is able to completely mitigate the slowdown effect of firm bankruptcies. 39 Table 6 – Regular banks and banks subject to macro-prudential policies Banks subject to a CCB of 2.5% Banks subject to a CCB of 5% Regular Banks Output gap 3.32% 3.36% 3.31% GDP volatility 0.86% 1.08% 0.86% GDP autocorrelation coefficient 0.92 0.96 0.93 Firms' default rate 1.43% 1.48% 1.49% Banks' default rate 0.57% 0.57% 0.57% Source: Authors In spite of this, it is possible to conclude that using credit-to-GDP ratios as a trigger to the implementation of the counter-cyclical capital buffer might be, if taken single-handedly, an insufficient approach. The results of our model suggest that credit is most important when the output gap is larger. In practice, this means that high levels of leverage may be beneficial if related with the need to close a large gap between the economy’s supply and demand (i.e., when firms have abundant investment opportunities). Since the same level of leverage may be too high or too low depending on the current state of the economy, the establishment of a counter-cyclical capital buffer must also take into account the impact of credit-constraining measures in GDP growth and an assessment of the level of debt that can be sustained by the economy without causing the build-up of system-wide risk. 4.4 Endogenizing bank failures As already described, bank failures are a beneficial phenomenon in our baseline model. A bank that enters into resolution is naturally not complying with minimum capital requirements and is thus not allowed to make new loans. The assumption that this agent can be replaced through foreign investment by a credit institution that can finance economic activity from the onset could be signaled as one of the reasons why our model reveals improved economic performance with risky banks. To address this shortcoming, we endogenize the cost of bank failures. Like before, depositors of failed banks are still bailed-in, thus absorbing the losses of the institution and raising its capital to zero. Under the new setting, however, depositors further see a part of their deposits being converted into equity so that the new institution complies with minimum capital requirements (specifically, the capital adequacy ratio is set 12% in excess of the minimum in each scenario). 46 Appendix I. ADF test (no intercept nor trend) – Cross-run average time series of log GDP – Baseline calibration Null Hypothesis: GDP has a unit root Exogenous: None Lag Length: 9 (Automatic - based on SIC, maxlag=17) t-Statistic Prob.* Augmented Dickey-Fuller test statistic 1.438923 0.9629 Test critical values: 1% level -2.570466 5% level -1.941578 10% level -1.616194 *MacKinnon (1996) one-sided p-values. Augmented Dickey-Fuller Test Equation Dependent Variable: D(GDP) Method: Least Squares Date: 02/26/15 Time: 21:17 Sample: 60 480 Included observations: 421 Variable Coefficient Std. Error t-Statistic Prob. GDP(-1) 6.26E-06 4.35E-06 1.438923 0.1509 D(GDP(-1)) 0.166225 0.048820 3.404837 0.0007 D(GDP(-2)) 0.327729 0.049231 6.656964 0.0000 D(GDP(-3)) -0.155027 0.051637 -3.002276 0.0028 D(GDP(-4)) 0.054448 0.053077 1.025835 0.3056 D(GDP(-5)) 0.178412 0.052348 3.408165 0.0007 D(GDP(-6)) 0.037580 0.052991 0.709172 0.4786 D(GDP(-7)) 0.045778 0.052473 0.872409 0.3835 D(GDP(-8)) 0.155157 0.049588 3.128924 0.0019 D(GDP(-9)) 0.139903 0.049362 2.834199 0.0048 R-squared 0.809321 Mean dependent var 0.002565 Adjusted R-squared 0.805146 S.D. dependent var 0.001343 S.E. of regression 0.000593 Akaike info criterion -12.00004 Sum squared resid 0.000144 Schwarz criterion -11.90402 Log likelihood 2536.009 Hannan-Quinn criter. -11.96209 Durbin-Watson stat 2.021119 47 II. ADF test (intercept and trend) – Cross-run average time series of log GDP – Baseline calibration Null Hypothesis: GDP has a unit root Exogenous: Constant, Linear Trend Lag Length: 3 (Automatic - based on SIC, maxlag=17) t-Statistic Prob.* Augmented Dickey-Fuller test statistic -10.68448 0.0000 Test critical values: 1% level -3.980006 5% level -3.420533 10% level -3.132959 *MacKinnon (1996) one-sided p-values. Augmented Dickey-Fuller Test Equation Dependent Variable: D(GDP) Method: Least Squares Date: 02/26/15 Time: 21:18 Sample: 60 480 Included observations: 421 Variable Coefficient Std. Error t-Statistic Prob. GDP(-1) -0.011720 0.001097 -10.68448 0.0000 D(GDP(-1)) 0.090017 0.047722 1.886286 0.0600 D(GDP(-2)) 0.230838 0.046490 4.965323 0.0000 D(GDP(-3)) -0.186049 0.047606 -3.908075 0.0001 C 0.170527 0.015883 10.73666 0.0000 @TREND("60") 1.97E-05 2.02E-06 9.742456 0.0000 R-squared 0.821892 Mean dependent var 0.002565 Adjusted R-squared 0.819746 S.D. dependent var 0.001343 S.E. of regression 0.000570 Akaike info criterion -12.08725 Sum squared resid 0.000135 Schwarz criterion -12.02963 Log likelihood 2550.366 Hannan-Quinn criter. -12.06448 F-statistic 383.0099 Durbin-Watson stat 2.005924 Prob(F-statistic) 0.000000 48 III. ADF test (intercept and trend) – Exemplificative time series of log GDP (seed -397426811) – Baseline calibration Null Hypothesis: SEED___397426811_ has a unit root Exogenous: Constant, Linear Trend Lag Length: 1 (Automatic - based on SIC, maxlag=17) t-Statistic Prob.* Augmented Dickey-Fuller test statistic -1.513395 0.8240 Test critical values: 1% level -3.977372 5% level -3.419250 10% level -3.132200 *MacKinnon (1996) one-sided p-values. Augmented Dickey-Fuller Test Equation Dependent Variable: D(SEED___397426811_) Method: Least Squares Date: 02/18/15 Time: 19:51 Sample (adjusted): 3 480 Included observations: 478 after adjustments Variable Coefficient Std. Error t-Statistic Prob. SEED___397426811_(-1) -0.014250 0.009416 -1.513395 0.1308 D(SEED___397426811_(-1)) -0.147429 0.027361 -5.388184 0.0000 C 13995.62 11664.01 1.199898 0.2308 @TREND("1") 55.72262 24.32668 2.290598 0.0224 R-squared 0.075482 Mean dependent var 2217.037 Adjusted R-squared 0.069631 S.D. dependent var 38776.71 S.E. of regression 37402.32 Akaike info criterion 23.90519 Sum squared resid 6.63E+11 Schwarz criterion 23.94008 Log likelihood -5709.339 Hannan-Quinn criter. 23.91890 F-statistic 12.89992 Durbin-Watson stat 1.137362 Prob(F-statistic) 0.000000 Please note: You are most sincerely encouraged to participate in the open assessment of this discussion paper. You can do so by either recommending the paper or by posting your comments. Please go to: http://www.economics-ejournal.org/economics/discussionpapers/2016-27 The Editor © Author(s) 2016. Licensed under the Creative Commons License - Attribution 4.0 International (CC BY 4.0).