Agent-based models of financial markets Development Framework. October 5, 2015

Transcription

1 Agent-based models of financial markets Development Framework October 5, 2015 Content A model of the Swiss real estate market Global Macro Model S&P 500 Model (Development Framework) Structure Sheets Macros Case study

2 Model of the Swiss real estate market Real estate prices Investment Real estate prices Private Trend followers Market BUY! Rents Interest rates Private residents Stocks SELL! Population Investors ABM of the Swiss real estate market (1) Real Estate Price Index & Modeled Trading Positions The chart shows the development of the SI Investment PR index and the BUY-/SELL-signals generated by the model Hit Rate: h = 0.77 Model Efficiency: r = 0.56 (June 30, 2015) 80 '87 '89 '91 '93 '95 '97 '99 '01 '03 '05 '07 '09 '11 '13 '15 Buy-/Sell-Signal SI Investment PR

3 ABM of the Swiss real estate market (2) Real Estate Price Index & Modeled Trading Positions The chart shows the development of the SI Private PR index and the BUY-/SELL-signals generated by the model Hit Rate: h = 0.77 Model Efficiency: r = 0.56 (June 30, 2015) '87 '89 '91 '93 '95 '97 '99 '01 '03 '05 '07 '09 '11 '13 '15 Buy-/Sell-Signal SI Private PR How can we avoid a real estate crisis? What are your actions if you are the regulator, government and central bank? What are your actions if you are a market participant? Please take five minutes to think about this question. Brieflypreparefora discussionin class-discuss itwithyour neighbours.

4 Scenario rising interest rates Real Estate Price Indices and Forecast (assuming Rising Interest Rates) 80 '87 '89 '91 '93 '95 '97 '99 '01 '03 '05 '07 '09 '11 '13 '15 '17 '19 SI Investment PR SI Private PR Forecast Swiss Yields Szenario Aside from producing trend predictions for the next quarter, the model can be used for long term scenario analysis and stress testing. In the following example, the effects of a longterm rise in interest rates on the SI Investment PR and SI Private PR are analyzed. The simulation is run up to the last quarter s end using historical data, then the interest rates are continuously increased. From this point on, the output time series (SI Investment PR and SI Private PR) are generated by the model. The chart shows the outcome of the simulation for both target indices and also the yield of the Swiss Index as the input variable. For both indices, a long-term increase of interest rates leads to a clearly observable and significant decrease. «Nil-scenario» 180 SI Investment PR 200 SI Private PR In a situation lacking both, positive and negative market forces, the simulated market nevertheless can correct negatively. The reason is that the majority of agents/market participants is already invested in real estate, and their potential to add further assetstotheir investment portfolio is limited due to a budget constraint. If however negative market impulses prevail, a significant negative correction is to be expected according to the simulation. Therefore, according to the model, a further long-term rise in the Swiss real estate market is to be expected only in a regime of prolonged, strong, and positive market forces.

5 Key points! The AVACO model of the Swiss real estate market fits the development of the market since 1986 well. Hit Rate and Model Efficiency are high, however the target time series has a low volatility. the target time series is in some parts interpolated. The model is simple, yet can effectively be used by investors to test investment strategies and by risk managers. Useful for forecasting (investment strategy) and simulation (scenario analysis, e.g. herding and crashes). Model relies on both fundamental and technical factors. The model combines implicit and explicit knowledge. In the absence of positive market forces, the market tends towards a negative correction, which becomes more pronounced in the presence of negative market forces such as rising interest rates. In the simulated scenarios, increasing interest rates can lead to a sharp correction of the real estate market. A soft landing of the Swiss real estate market is unlikely. Once the upward trend will stop, a long-term negative correction is most likely. Simplifying complexity

6 Global Macro Model Nikkei Gold JPY DAX Oil USD EUR S&P CHF SPI Global Macro Model Monthly data since 1982 Agent types Fundamental Technical Hit rate between 53.2% and 62.2% Model efficiency between 12.5% and 33.4%

7 Market Environment Assets EURUSD, YENUSD, EURCHF USD, EUR, YEN, CHF S&P 500, DAX, Nikkei 225, SPI Gold, Oil Market Maker (round based) The price calculation is based on demand and supply: = + () +() p is the time series and y is the predicted time series; D is demand and S is supply of all agents Example Model S&P 500 Nikkei Gold JPY DAX Oil USD EUR S&P CHF SPI

8 Example Model S&P 500 The Excel file of the Development framework is available Monthly data since 1982 S&P 500 Yield US Government s Agent types Investor (fundamental) Trend follower (technical) Lag 1 Lag 12 Market Maker (each round) Hit rate 56.9% Model efficiency 18.4% Structure

9 Agents Prices

10 Ticker Macros

11 Macros simulation

12 trend investor

13 clearing Walkthrough Please walk through the first trading circle. (Take 10 minutes) Briefly prepare for questions and a discussion in class- discuss it with your neighbours.

14 Case Study The case study consists of an agent-based model. It is to be performed in MS Excel with a short documentation. The case study should motivate and formulate a research question present a model of any subset of a financial market and its participants (agents) outline and discuss the results Write a one-page proposal and submit it to ta@avaco.ch. Deadline for the proposal: 16 October You have to present your model and its results as an elevator pitch in the last weeks of classes (see Seminar Overview). Deadline for the model, documentation and presentation slides is Friday before the presentation date. The documentation is to be 3 to 5 pages The presentation should consist of 3 to 5 slides; you have 5 minutes to present The grade depends on Originality Economic and academic foundation Quality and complexity of software Quality of documentation Quality of presentation Results