Empirical evidence | EPFL Graph Search

Empirical evidence for a proposition is evidence, i.e. what supports or counters this proposition, that is constituted by or accessible to sense experience or experimental procedure. Empirical evidence is of central importance to the sciences and plays a role in various other fields, like epistemology and law. There is no general agreement on how the terms evidence and empirical are to be defined. Often different fields work with quite different conceptions. In epistemology, evidence is what justifies beliefs or what determines whether holding a certain belief is rational. This is only possible if the evidence is possessed by the person, which has prompted various epistemologists to conceive evidence as private mental states like experiences or other beliefs. In philosophy of science, on the other hand, evidence is understood as that which confirms or disconfirms scientific hypotheses and arbitrates between competing theories. For this role, it is important that evidence is public and

Essays in Monetary Policy and Asset Pricing

Benoit Vincent Sylvain Cornet

This thesis uses machine learning techniques and text data to investigate the relationships that arise between the Fed and financial markets, and their consequences for asset prices.The first chapter, entitled Market Expectations and the Impact of Unconventional Monetary Policy: An Application to Twitter Data, is an answer to (Greenlaw et al., 2018), who show, by looking at a large set of monetary policy announcements made between November 2008 and December 2017 - FOMC meetings, release of minutes and speeches of the Fed Chair - that long term yields tended to increase following these events. By doing so, the authors challenge common wisdom according to which the central bank intervention in the aftermath of the financial crisis lowered long term rates (Gagnon, 2016). Using machine learning and twitter data, this chapter develops a novel measure of market expectations of monetary policy, and shows that the increase in yields was simply due to a marginal adjustment of market expectations following announcements being less dovish than expected.The second chapter, entitled Informational Feedback Loop, Monetary Policy Decisions and Asset Prices Dynamics, investigates the consequences of a Fed that uses (1) its own private signal and (2) fed funds futures to take its monetary policy decision. Fed funds futures aggre- gate private information received by financial markets participants - traders - but they also depend on traders' expectations about the Fed's behavior, which makes futures endogenous in the central bank decision. The theoretical model shows that the surprise generated by monetary policy announcements and the subsequent adjustment in short term U.S. treasury yields depend on the precision of the signals received by each agent. When the signal received by traders is more precise than the central bank's, the latter relies more on fed funds futures to take its decision, and the surprise and adjustment of short term yields are smaller. By contrast, long term yields adjust only because the announcement provides traders with new information about the state of the economy, by revealing the central bank's private signal. Finally, when the Fed is averse to financial markets volatility, it tends to put some weight on fed funds futures even if they are not informative about the state of the economy. The empirical part of the paper provides some evidence supporting these channels, by using a topic and tone approach (Hansen and McMahon, 2016) to extract the precision of the signals received by the central bank and traders from FOMC minutes and tweets respectively.

EPFL2022

Financial Stability and the Macroeconomy

Corinne Dubois

The financial crisis of 2007-2009 drew attention to the essential role of banks for the macroeconomy and to the importance of having a resilient financial sector. A vulnerability in the financial sector spills over to the real economy and can drive it into a deep recession. Following the financial crisis, policy makers have issued rules and regulations aimed at strengthening the banks. My dissertation considers some of those measures, namely bank liquidity regulation and deposit insurance, and tries to assess their effectiveness in stabilizing the financial sector and the economy in general. The first chapter (joint with Prof. Luisa Lambertini) develops a model with regulated banks and a hedge fund to analyze the behavior of wholesale funding and the macroeconomic consequences of liquidity regulation. Banks raise deposits and subordinated wholesale funding from the hedge fund. Wholesale funding amplifies shocks: it is curtailed in economic downturns to avoid leveraging up and risk-taking by banks, further depressing credit and economic activity. By making banks safer, liquidity regulation increases wholesale funding and bank loans at the steady state. However, flat liquidity regulation requiring banks to hold a constant fraction of liquid assets can have unintended consequences and increase macroeconomic volatility. Our model further shows that cyclically adjusted regulation stabilizes the response of the economy to shocks. We also provide empirical evidence suggesting that liquidity helps contain the contraction in wholesale funding and loans during times of funding stress. The second chapter explores the macroeconomic and welfare implications of harmonizing or joining deposit insurance in the Eurozone. I develop a two-country model to represent the core and periphery countries in the Eurozone. The model features fragile banks that can be subject to endogenous and costly runs. The two countries are financially integrated through an international interbank market. I introduce deposit insurance and compare different insurance regimes in steady state and along the business cycle. The model emphasizes that financial integration implies co-movement in asset prices and the transmission of shocks across the border. I find that when deposit insurance has lower coverage in the periphery than in the core, financial stability is undermined in both regions due to higher asset price volatility and a higher probability of deposit runs. Hence, either harmonizing or joining deposit insurance leads to substantial welfare benefits for both countries. In the third chapter, I develop a macroeconomic model with multiple equilibria and bank runs to analyze deposit insurance. A bank run is a sunspot equilibrium, and it is caused by a self-fulfilling panic of all depositors. The model features a unique no-run equilibrium when macroeconomic conditions are good, but an unexpected negative shock may open up the possibility of a run equilibrium. A bank run is a rare but costly event that drives the economy in a long and severe recession. I analyze the introduction and build-up of a deposit insurance fund and show that it reduces the cost and probability of bank runs. The dynamic simulation of the model reveals that deposit insurance diminishes macroeconomic and financial instability following a technological shock.

EPFL2018

Traffic Anomaly Detection and Diagnosis on the Network Flow Level

Marc Philippe Stöcklin

Monitoring traffic events in computer network has become a critical task for operators to maintain an accurate view of a network's condition, to detect emerging security threats, and to safeguard the availability of resources. Conditions detrimental to a network's performance need to be detected timely and accurately. Such conditions are observed as anomalies in the network traffic and may be caused by malicious attacks, abuse of resources, or failures of mission-critical servers and devices. Behavior-based anomaly detection techniques examine the traffic for patterns that significantly deviate from the network normal activities. Such techniques provide a complementary layer of defense to identify undesired conditions which traditional, signature-based methods fail to detect. These conditions may, for example, emerge from zero-day exploits, outbreaks of new worms, unanticipated user behavior, or deficiencies in the network infrastructure. This thesis is concerned with the challenge of detecting traffic anomalies with behavior-based methods from flow-level network traffic measurements while providing interpretable alert information. We address the problem from two opposite perspectives by analyzing network behavior and individual host behavior. Learning the normal behavior of network activities and detecting relevant deviations thereof is a complex task since behavior changes may also occur under legitimate conditions and should not be reported as anomalies. Due to the absence of explicit detection rules, behavior-based methods moreover provide less precise information as to the causes of aberrant events. Network operators, however, critically depend on meaningful detection results to timely react to alerts by defining effective countermeasures or ruling out potential false alarms. The first part of this work introduces a novel detection scheme which mines for anomalies in the network behavior observed from traffic feature distributions. We study how various types of anomalies may be detected while providing sufficient information to administrators for their characterization. Based on the observation that networks have multiple behavior modes, we propose a method to estimate and model the modes during an unsupervised learning phase. Observed network behavior is compared to the baseline models by means of a two-layered distance computation: Fine-grained anomaly indices indicate suspicious behavior of individual components of traffic features whereas collective anomaly scores for each feature enable effective detection of anomalies that affect multiple components. We show that the two detection layers reliably expose different types of anomalies. Compared with existing detection methods, the resulting alerts provide important additional information that enables administrators to draw early conclusions as to the anomaly causes. In the second part, we address the challenge of processing high-cardinality traffic information in behavior-based anomaly detection. We introduce an adaptive, locality preserving pre-processing method of measurement data into histogram representations with a manageable number of variable-sized bins. Our technique iteratively adapts, with limited, tunable memory requirements, to the empirical distribution of observations in a data stream in order to evenly balance the observations over the histogram bins. As an important result, we show that our method approximates input distributions well and improves the level of detail in histograms compared to traditional methods. Applied to behavior-based anomaly detection, higher detection sensitivity is achieved while, thanks to preserving the locality of observations, the meaning of bins and the interpretability of detection results is retained. In the third part, we focus on the problem that many low-volume anomalies, emerging from individual hosts, are likely to evade from detection because they are not reflected as significant deviations in the variability of aggregate behavior patterns of hosts on the network level. To address this problem, we examine the behavior of individual hosts from their network connection-level activities using an unobtrusive, passive monitoring approach. We develop an unsupervised method to track properties in the activities that recur over time and establish detailed behavior profiles. We propose three anomaly detectors that compare observed activities to the profiles in order to recognize suspicious changes in a host's activities, giving evidence of abnormal behavior. We demonstrate their effectiveness in revealing different types of anomalies, which are not detectable in aggregate network statistics, while providing meaningful alert information to administrators. In addition, we show that the profiles of individual hosts are stable over time and representative of their activities, and may even be used to identify hosts solely from their traffic behavior. In summary, the methods and algorithms presented in this thesis enable practical and interpretable detection of traffic anomalies on the network ow level with behavior-based methods.

EPFL2011