{
  "id": 477470,
  "title": "An example of pattern change caused by COVID: Impact of the weekday",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/477470",
  "author_name": "",
  "post_date": "2024-02-16T09:06:26.547294Z",
  "votes": 22,
  "comment_count": 4,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F494641%2F51971995ea7c7b1438e4bb486023a212%2Fnewplot%20(18).png?generation=1708074470176405&amp;alt=media\"></p>\n<ul>\n<li>Before covid, applications on Sat and Sun (weekday 6 &amp; 7) were 10-20% less likely to default</li>\n<li>After covid outbreak (April 2020), this was no longer the case</li>\n<li>After covid outbreak (April 2020), also the overall default rates have dropped significantly</li>\n</ul>\n<p>Research questions:</p>\n<ul>\n<li>Will the weekly pattern return? If so, when? (Impossible to answer because we only have data until Oct 2020). </li>\n<li>How to build our models to incorporate it?</li>\n</ul>",
  "messages": [
    {
      "id": "2654595",
      "postDate": "02/16/2024 09:06:26",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F494641%2F51971995ea7c7b1438e4bb486023a212%2Fnewplot%20(18).png?generation=1708074470176405&amp;alt=media\"></p>\n<ul>\n<li>Before covid, applications on Sat and Sun (weekday 6 &amp; 7) were 10-20% less likely to default</li>\n<li>After covid outbreak (April 2020), this was no longer the case</li>\n<li>After covid outbreak (April 2020), also the overall default rates have dropped significantly</li>\n</ul>\n<p>Research questions:</p>\n<ul>\n<li>Will the weekly pattern return? If so, when? (Impossible to answer because we only have data until Oct 2020). </li>\n<li>How to build our models to incorporate it?</li>\n</ul>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F494641%2F51971995ea7c7b1438e4bb486023a212%2Fnewplot%20(18).png?generation=1708074470176405&alt=media)\n\n- Before covid, applications on Sat and Sun (weekday 6 & 7) were 10-20% less likely to default\n- After covid outbreak (April 2020), this was no longer the case\n- After covid outbreak (April 2020), also the overall default rates have dropped significantly\n\nResearch questions:\n- Will the weekly pattern return? If so, when? (Impossible to answer because we only have data until Oct 2020). \n- How to build our models to incorporate it?",
      "votes": null
    },
    {
      "id": "2655476",
      "postDate": "02/16/2024 22:50:12",
      "content": "<p>Thank you <a href=\"https://www.kaggle.com/narsil\" target=\"_blank\">@narsil</a> for sharing your findings. I have not had the time yet, for an in-detail look at the data. </p>\n<p>My gut feeling, when reading your post, is that your findings reflect a very broad and common fact about defaults and bankruptcies in general during the covid pandemic. </p>\n<p>During the pandemic, there was so much government support that bankruptcies, defaults etc. became less frequent as the governments infused a significant amount of liquidity etc. into the system. </p>\n<p>Years later following the reduction in government funding (risk distortion) the bankruptcies have in general increased to a bit above expected levels. Hence defaults were postponed/delayed as a result of government actions, which cannot, and have not continued.</p>",
      "rawMarkdown": "Thank you @narsil for sharing your findings. I have not had the time yet, for an in-detail look at the data. \n\nMy gut feeling, when reading your post, is that your findings reflect a very broad and common fact about defaults and bankruptcies in general during the covid pandemic. \n\nDuring the pandemic, there was so much government support that bankruptcies, defaults etc. became less frequent as the governments infused a significant amount of liquidity etc. into the system. \n\nYears later following the reduction in government funding (risk distortion) the bankruptcies have in general increased to a bit above expected levels. Hence defaults were postponed/delayed as a result of government actions, which cannot, and have not continued.",
      "votes": null
    },
    {
      "id": "2655746",
      "postDate": "02/17/2024 07:25:20",
      "content": "<p>Indeed. Now let's think how to incorporate that into our models</p>",
      "rawMarkdown": "Indeed. Now let's think how to incorporate that into our models",
      "votes": null
    },
    {
      "id": "2657934",
      "postDate": "02/18/2024 20:40:45",
      "content": "<p>Interesting finding. My intuitive thought is that COVID forced people to WFH and thus the demand for \"home\" increased during this period. Which column does the x-axis come from? </p>",
      "rawMarkdown": "Interesting finding. My intuitive thought is that COVID forced people to WFH and thus the demand for \"home\" increased during this period. Which column does the x-axis come from?",
      "votes": null
    },
    {
      "id": "2658470",
      "postDate": "02/19/2024 07:57:28",
      "content": "<p><code>decision_date</code></p>",
      "rawMarkdown": "`decision_date`",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2655476,
      "author_name": "signator",
      "author_url": "",
      "post_date": "02/16/2024 22:50:12",
      "content": "<p>Thank you <a href=\"https://www.kaggle.com/narsil\" target=\"_blank\">@narsil</a> for sharing your findings. I have not had the time yet, for an in-detail look at the data. </p>\n<p>My gut feeling, when reading your post, is that your findings reflect a very broad and common fact about defaults and bankruptcies in general during the covid pandemic. </p>\n<p>During the pandemic, there was so much government support that bankruptcies, defaults etc. became less frequent as the governments infused a significant amount of liquidity etc. into the system. </p>\n<p>Years later following the reduction in government funding (risk distortion) the bankruptcies have in general increased to a bit above expected levels. Hence defaults were postponed/delayed as a result of government actions, which cannot, and have not continued.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2655746,
          "author_name": "narsil",
          "author_url": "",
          "post_date": "02/17/2024 07:25:20",
          "content": "<p>Indeed. Now let's think how to incorporate that into our models</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2657934,
      "author_name": "makeli",
      "author_url": "",
      "post_date": "02/18/2024 20:40:45",
      "content": "<p>Interesting finding. My intuitive thought is that COVID forced people to WFH and thus the demand for \"home\" increased during this period. Which column does the x-axis come from? </p>",
      "votes": null,
      "replies": [
        {
          "id": 2658470,
          "author_name": "narsil",
          "author_url": "",
          "post_date": "02/19/2024 07:57:28",
          "content": "<p><code>decision_date</code></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2654595": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F494641%2F51971995ea7c7b1438e4bb486023a212%2Fnewplot%20(18).png?generation=1708074470176405&alt=media)\n\n- Before covid, applications on Sat and Sun (weekday 6 & 7) were 10-20% less likely to default\n- After covid outbreak (April 2020), this was no longer the case\n- After covid outbreak (April 2020), also the overall default rates have dropped significantly\n\nResearch questions:\n- Will the weekly pattern return? If so, when? (Impossible to answer because we only have data until Oct 2020). \n- How to build our models to incorporate it?",
    "2655476": "Thank you @narsil for sharing your findings. I have not had the time yet, for an in-detail look at the data. \n\nMy gut feeling, when reading your post, is that your findings reflect a very broad and common fact about defaults and bankruptcies in general during the covid pandemic. \n\nDuring the pandemic, there was so much government support that bankruptcies, defaults etc. became less frequent as the governments infused a significant amount of liquidity etc. into the system. \n\nYears later following the reduction in government funding (risk distortion) the bankruptcies have in general increased to a bit above expected levels. Hence defaults were postponed/delayed as a result of government actions, which cannot, and have not continued.",
    "2655746": "Indeed. Now let's think how to incorporate that into our models",
    "2657934": "Interesting finding. My intuitive thought is that COVID forced people to WFH and thus the demand for \"home\" increased during this period. Which column does the x-axis come from?",
    "2658470": "`decision_date`"
  },
  "source": "meta"
}