Abstract:Estimation of high dimensional covariance matrices is an interesting and important research topic. In this paper, we propose a dynamic structure and develop an estimation procedure for high dimensional covariance matrices. Asymptotic properties are derived to justify the estimation procedure and simulation studies are conducted to demonstrate its performance when the sample size is finite. By exploring a financial application, an empirical study shows that portfolio allocation based on dynamic high dimensional covariance matrices can significantly outperform the market from 1995 to 2014. Our proposed method also outperforms portfolio allocation based on the sample covariance matrix and the portfolio allocation proposed in Fan, Fan and Lv (2008).
What problem does this paper attempt to address?
This paper attempts to solve the problem of high - dimensional covariance matrix estimation, especially its application in financial portfolio allocation. Specifically, the paper proposes a dynamic structure and the corresponding estimation method to deal with high - dimensional covariance matrices, and through empirical research shows that portfolio allocation based on this dynamic high - dimensional covariance matrix can significantly outperform the market performance, especially during the period from 1995 to 2014.
### Background and Problems of the Paper
In statistics and econometrics, the estimation of covariance matrices is an important research topic, widely used in fields such as economics, finance, and psychology. Traditional covariance matrix estimation methods are based on sample covariance matrices, but in high - dimensional situations, especially when the inverse matrix needs to be calculated, the performance of sample covariance matrices is often not good. This is because when using the inverse of the sample covariance matrix to estimate the inverse of the true covariance matrix, the estimation error will accumulate, resulting in unacceptable results when constructing financial portfolios.
In recent years, many studies have attempted to solve the problem of high - dimensional covariance matrix estimation, usually by introducing sparsity conditions to balance variance and bias. For example, the work of Wu and Pourahmadi (2003), El Karoui (2008), Bickel and Levina (2008a, 2008b), Lam and Fan (2009), Fan, Liao, and Mincheva (2011), etc. Fan, Fan and Lv (2008) estimate the covariance matrix by introducing a factor model.
However, most of the existing literature assumes that the covariance matrix is constant over time, while in practical applications, the covariance matrix is often dynamically changing. For example, today's optimal portfolio allocation may no longer be optimal tomorrow or next month. Therefore, when applying Markowitz's optimal portfolio allocation formula, the covariance matrix used should be dynamic and allowed to change over time.
### Proposal of Dynamic Structure
In order to introduce the dynamic structure of the covariance matrix, the paper proposes a method based on the adaptive varying - coefficient model. Specifically, assume that \((X_t^T, Y_t^T)\) is a time series, where \(Y_t\) is a \(p_n\) - dimensional vector and \(X_t\) is a \(q\) - dimensional factor. Assume that \(p_n\) increases as \(n\) increases, while \(q\) is fixed. In addition, assume that \(X_t\) is a stationary Markov process.
The paper assumes that:
\[ Y_t = g(X_{t - 1}^T\beta)+\Phi(X_{t - 1}^T\beta)X_t+\epsilon_t, \]
where \(\beta = (\beta_1,\dots,\beta_q)^T\), \(\Phi(X_{t - 1}^T\beta)\) is a factor loading matrix that changes with \(X_{t - 1}^T\beta\), and \(\{\epsilon_t, t = 1,\dots,n\}\) are random error terms independent of \(\{X_t, t = 1,\dots,n\}\).
### Main Contributions
1. **Dynamic Structure**: The paper proposes a new dynamic structure for estimating high - dimensional covariance matrices.
2. **Estimation Method**: Develops an estimation method and verifies its performance through asymptotic properties and simulation studies.
3. **Empirical Research**: Through empirical research, it shows that portfolio allocation based on the dynamic high - dimensional covariance matrix can significantly outperform the market performance, especially during the period from 1995 to 2014.
### Experimental Results
The paper verifies the effectiveness of the proposed method through simulation studies and actual data sets. In the simulation study, the paper generated 1,000 data sets and compared the performance of different methods. The results show that the proposed dynamic structure method performs well in estimating the covariance matrix and its inverse matrix.
In the actual data analysis, the paper uses the daily return data of 49 industry portfolios and compares it with the sample covariance matrix - based and the estimation method proposed by Fan, Fan and Lv (2008). The results show that the proposed dynamic structure method is significantly better than other methods in terms of the annual Sharpe ratio and the final balance.