Persistency has always been the area of grave concern over the decades in insurance sector. To model the same one would initially make utilization of regression approach but that could not serve the purpose of identifying the risk factors and time trends which affect persistency. For this purpose, one should make use of survival models which account for censored data. But there are still certain concerns utilizing conventional survival models. One of which is left truncation of insurance data. Another concern is the large portfolios and absolute numbers involved in insurance sector. Thus, conventional survival models could not be adjusted for convergence when applied to such large portfolios and absolute numbers like amount of Sum Assured (SA), therefore, justifying the application of actuarial laws to model the problem. Objective of this paper is to make a comparative study between survival models and actuarial models to model persistency, so that the validity and robustness of the actuarial models may be established in case of analyzing the insurance phenomenon. Models which seem to be the best fitted models have been checked diagnostically for validity also. For this purpose, the plots of standardized residuals have been studied for randomness. To deal with dynamic structure of insurance data, stratifications have been used as per different criteria given by Insurance Regulatory and Development Authority of India (IRDAI) also. AIC values under each stratum have been weighted and then averaged to arrive at single value underneath each specification. It is found that actuarial models are robust in all cases and valid too. In fact, the choice of best fitted actuarial model has not been changed from case to case. Throughout our study only Gompertz curve fitted well to the data and is also found to be valid. The nature of relationships of Age and SA with persistency is found to be positive.

数十年来，持久性一直是保险业最为关注的领域。要对同一模型进行建模，最初将利用回归方法，但不能用于确定影响持久性的风险因素和时间趋势。为此，应该利用生存模型来解释审查数据。但是，利用传统的生存模型仍然存在某些担忧。其中之一是截断保险数据。另一个问题是涉及保险部门的庞大投资组合和绝对数量。因此，当应用于如此大的投资组合和诸如保额（SA）的绝对数字时，传统的生存模型无法针对收敛性进行调整，因此，有理由应用精算法对问题进行建模。本文的目的是对生存模型和精算模型之间的持久性进行比较研究，以便在分析保险现象时建立精算模型的有效性和鲁棒性。似乎是最合适模型的模型也已通过诊断检查了有效性。为此，已经研究了标准化残差图的随机性。为了处理保险数据的动态结构，还按照印度保险监管与发展局（IRDAI）给出的不同标准使用了分层。已加权每个层次下的AIC值，然后取平均值以得出每个规范下的单个值。发现精算模型在所有情况下都是鲁棒的，并且也是有效的。事实上，最佳选择的精算模型的选择因案例而异。在我们的整个研究中，只有Gompertz曲线很好地拟合了数据，也被认为是有效的。发现年龄和SA与持久性之间的关系是积极的。