There are many challenges to overcome when you need a study or interview for your research project in academic writing. One of these are the many different types of research bias that need to be considered because bias can highly influence the results of your study and thus deem it invalid. The following article will show you every type of research bias you could potentially encounter, give examples to explain it further, and how to avoid it.
Research Bias – In a Nutshell
Research bias is an umbrella term for many different types of bias that can occur before, during and after a study, influencing its results.
Definition: Research bias
Research bias refers to any negative factors and flaws that distort empirical studies. Excellent studies seek to remove this type of influence that might pose hurdles to findings and conclusions. Eliminating confounding factors, flaws, and faulty conclusions is essential to producing quality work that’s accurate and applicable.
- ✓ Free express delivery
- ✓ Individual embossing
- ✓ Selection of high-quality bindings
Pre-trial research bias
Many research biases can influence a study before it is even conducted. These happen due to the researcher’s influence in phrasing the questions, sampling participants, or missing out on contributing factors.
Researcher bias
Researcher bias, or design bias, happens when the beliefs and opinions of the researcher influence how they phrase the questions, what exactly they ask for or in what form they conduct the study. Oftentimes, the researcher unconsciously chooses a setting that will lead to his desired results and this way causes errors in study design.
A way to avoid researcher bias is to ask more open-ended questions that do not tempt the participant to give a specific answer, or to have a second, independent person check your questions.
Example
You will find this bias, especially in unintentionally leading questions, for example if you ask “Does owning a dog restrict your holiday plans?”
This question tempts the participant to answer “yes”, due to the Acquiescence bias, or at least makes them think about when owning a dog restricted them in their holiday plans.
Instead, you could ask something like, “How does having a dog affect your holiday plans?” This question also intents on the effects of owning a dog, but it is phrased more openly and does not imply a negative expectation.
Selection bias
Selection bias occurs during the stage of selecting participants for your study. Possible influences can be how or where you sample them. There are a few subtypes of selection bias, which will be explained in the following, including ideas on how to avoid these errors in study design.
- Sampling or ascertainment bias happens when your selected participants are not representative of the whole population. To avoid sampling bias, you should watch out that your group of participants does represent the entire population, at least according to the factors relevant in your study.
Example
For your survey, you sample participants in the city centre, wanting to examine their consumerist behaviour. You do, however, conduct your survey around noon in your mid-break, and are thus excluding all people who have to work at this time of the day. Thus, your sampled participants do not represent the whole population.
- Attrition bias is when only people with the same characteristics drop out of the study, thus leaving it less representative for the whole. Avoiding attrition bias can be complicated because you cannot force participants to stay part of the survey. You can either try to convince them into staying or take their dropouts into your result analysis.
Example
You are conducting a survey about exam anxiety among university students. Therefore, you sampled your participants randomly and have them write tests every week. It now happens that participants who do suffer from exam anxiety the most drop out of the study after a while, feeling overwhelmed with the pressure. This leads to warped results because the most important participants dropped out of your study and maybe even got replaced by ones without exam anxiety.
- Self-selection or volunteer bias refers to the circumstance that the people who volunteer to take part in the study all have the same or similar relevant characteristics. To avoid self-selection bias, simply sample randomly or check the important data in your participants’ vita to ensure representativeness.
Example
You want to conduct a survey about how extroverted and introverted people react to different scenarios. Therefore, you ask around your school for people who would like to participate in your study. However, as introverted people often also tend to be shy and thus avoid interactions like these, most of your volunteers are extroverts. This leads to a not representative result of the study.
- Survivorship bias is a type of research bias where the sampled participants only include people who have passed a certain trial beforehand. In some studies, this is done consciously. However, if the prior trial is not a requirement to become a participant, you should consider broadening your sampling.
Example
For your survey about the perception of getting one’s blood drawn, you sample participants around a hospital and ask them how stressful they experience this act. However, you did not consider that patients at a hospital most likely had to undergo far more stressful examinations than getting their blood drawn and will thus answer with a lower number on the scale.
- Nonresponse bias describes a condition when, in a survey, a certain group of people with similar characteristics does not answer the questions and thus makes the study less representative. This can be avoided by researching the reasons behind their nonresponse and remove those obstacles.
Example
Given you are conducting a study about how mobile elderly people are above a certain age, you visit an almshouse and ask the inhabitants to meet you in a secluded room for the interview. Some inhabitants, however, do not respond to your invitation. These are the ones that cannot get up from their beds anymore and are thus excluded from your study, while they would actually be quite relevant.
- Undercoverage bias refers to a situation where you sample participants from a subset of the population that is not representative for the whole, and you thus miss a certain piece of information that is characteristic for those left out. The only option to avoid this type of research bias is considering the possible factors that could determine the undercoverage.
Example
A common example for undercoverage bias is conducting an online survey about a general topic, for example, how much they are taking care of their mental health. The undercoverage in this case refers to the exclusion of people who do not have access to the internet or the special platform you are conducting the survey from.
Channeling bias
Channeling bias happens mostly in studies with non-randomized participants that are divided into groups. It is common practice in clinical trials to have an experimental group and a control group, which are sometimes selected by the staff. Therefore, it can happen that more participants with a higher probability of success end up in the experimental group, while participants with less likeliness to succeed end up in the control group.
To avoid this type of research bias, you may consider drawing the groups by lots or make sure the division is randomized in any other way.
Example
Channeling bias can happen, for example if you are studying the regeneration process after a surgery compared to the regeneration process of rehabilitation therapy for the same injury. The staff, however, subconsciously samples more young patients into the experimental group undergoing surgery, as their regeneration is generally better. This, of course, causes better results for the experimental group and tampers with the study.
Research bias during trial
The most common types of research bias happen during the trial of the study. As participants are only humans too, their answers can be dependent on external influences, individual experiences or social pressure. It is important to know which factors can influence your participants and reduce them to a minimum.
Information bias
Information bias, sometimes also called measurement bias, occurs when key study variables are inaccurately classified. This means that the information your participants give you might be influenced by different external or internal factors you might need to consider. There are a few subtypes of this kind of research bias, that will be explained in the following.
- Recall bias is the phenomenon that infrequent or special events are more likely to be remembered. However, as every person has different experiences and might also weight them differently, this leads to varying answers. If you want to avoid this, you should first consider the life of your participants and maybe ask them what led them to give a specific answer.
Example
You conduct a study about an illness and what might have contributed to it. For this case, you have a group of participants who are diagnosed with the illness and another group that is not diagnosed. If you ask them what they think could contribute to getting the illness, the people who are diagnosed may more likely remember certain things about their lives that could be a factor. This is because they have thought about it before, and the information is thus more present in their minds.
- Observer bias refers to the fact that people often see what they want or are expected to see. This warps their perception and thus the results of the study. To avoid this type of research bias, you should ask open-ended questions that cannot be answered with yes or no. Other than that, as the participant themselves does mostly not recognise they are biased, it is very difficult to do anything against it.
Example
You want to conduct a study about how dirty the environment of main railway stations is, and therefore inspect a couple of different stations in different cities. You expect those places to be rather dirty and unpleasant. In your perception, you therefore mostly recognise rubbish lying around somewhere or dusty pathways. Because you expect the place to be dirty, you only see what is dirty, and you will most likely not realise that you are picking on details while the whole station is actually rather clean.
- Performance bias results from unequal treatment between groups in research. The Hawthorne effect, or observer effect, refers to the change of behaviour participants show when they know they are being watched. The John Henry effect, on the other hand, explains that participants feel the need to compensate for the inequality between the groups. This can lead to a difference in motivation or general behaviour between the two groups.
Example
Performance bias can happen, for example, when you conduct a study about the connection between a healthy diet and the general wellbeing of the person. You divide your participants into the experimental group, which contains participants that have a healthy living style, and a control group, that either contains people with unhealthy or at least average living styles. However, the participants in the control group notice what you are inspecting and thus change their diet or habits to improve their results.
- Regression to the mean (RTM) happens mostly in studies that have more than one stage. The first stage tends to produce more extreme results due to possible nervousness of the participants or their need to do everything right. In the second stage, the participants are more relaxed and thus their answers might be more truthful, thus regressing to the mean of the population. To avoid this type of research bias, you could start the study with a first dummy stage that is not relevant to get the participants used to the setting and conduct follow-up surveys in the second stage for valid results.
Example
Given you go to a doctor’s appointment and they measure your blood pressure. Due to the nervousness you may feel in this environment, your numbers are probably a lot higher than average. This is why sometimes doctors repeat the measuring after a while, when the patient is settled down, to get more valid results.
Interviewer bias
Interviewer bias is a type of research bias where the reactions or behaviour of the interviewer influences the participant. This can be due to their tone of voice, phrasing of questions, but also their body language and facial expressions that have an impact on their answers during surveys.
A way to avoid interviewer bias is to watch our behaviour while conducting the interview and try to be as objective and neutral as possible.
Example
You conduct an interview about how popular different pets are and why. The following conversation could happen:
Participant: “I think cats and bunnies are my favourite pets.”
You: “Yes, cats are the cutest!”
You reacting positively could result in the participant elaborating more about cats than bunnies because they expect more encouraging answers.
Response bias
Response bias refers to the different things that can influence the answers of the participants. There are a few different subtypes of this research bias, which will be explained in the following.
- Acquiescence bias, also known as “yea-saying”, says that people are more likely to agree to a statement when asked a question than disagree. This is closely related to the social desirability bias because approving is considered polite and pleasing to the other person. A way to avoid this research bias is to phrase the questions in a way that the participants cannot just say yes or no, but have to elaborate their answer.
Example
This research bias also originates in the phrasing phase of your questions. For example, if you conduct a study about social orientation, you might be tempted to ask a question like “Would you say you are socially active?”, which means the answer would be either yes or no. To avoid having your participants answering only “yes”, you could rephrase the question in another way. “Do you prefer spending time on your own or in company?” would be a good example for a more open-ended question. This way, your participants are forced to at least think about the question a little more.
- Demand characteristics means that when participants know or guess the intent of questions in the study, they are more likely to give answers that will approve of the main question. This is why the actual theme of the study is mostly kept secret, or a fake topic is told to the participants.
Example
You conduct a study about how the diet of children changes in their teenage years, and thus ask them what they are usually eating once every year. Of course, the participants will notice that you are observing their diet and may lie about their nutrition to get better results.
- Social desirability bias is very common to happen because people feel the need to be socially accepted. This is why they will probably answer in a way that it follows norms of the society. Even if their actual opinion might differ from that, they will be hesitant to state an opinion that might not be socially accepted. To avoid this bias, you can conduct anonymous surveys online or via paper, so the participants do not feel pressured to answer in a certain way.
Example
You conduct a survey about how much cheating people think is allowed in exams. As most people know that cheating is generally not allowed, they may say that no cheating is socially acceptable because that is what society is teaching them to say. In reality, their opinion might be entirely different, depending on the subject, the importance of the exam or just generally.
- Question order bias happens when the order in which you ask the questions influences their answers. Occasionally, the previous question gives context to the next, leading to either giving away your intention or lacking content, and thus miscommunication might happen. A way to avoid this type of research bias might be to carefully select your order of questions or test them on your friends and family to gain another opinion on this factor.
Example
You are conducting a survey about the consumption of energy drinks, and therefore show your participants a selection of different brands. First, you ask them which of those they prefer to drink, while asking how many energy drinks they consume weekly. This can lead to the participants focusing on this one brand or at least on only the previously presented ones, while neglecting that they might consume even more from not mentioned ones.
- Extreme responding mostly happens when your participants are indifferent to a topic or question. They will choose what they feel like at that moment, not considering it carefully. Sometimes participants will also exaggerate on their opinion subconsciously or to please the interviewer. Avoiding this kind of research bias can be difficult, but you could try to double-check their opinion by repeating their answer and asking again if this is what they meant.
Example
When you conduct a study about how popular different pets are among randomly sampled participants, there may be a few among them that are not interested in pets at all. These people might answer in a black-and-white scheme, deciding on a whim if they like the questioned pet or not. This results into extreme answers that are based on “yes” or “no”.
- Courtesy bias happens due to the reluctance to give negative feedback. Especially in face to face or group interviews, participants might hold back on criticism to not hurt other people. This can be avoided by keeping your study anonymous or online, to have the participants give more truthful answers.
Example
Courtesy bias happens, especially in group or face-to-face interviews. If you ask your participants, for example, what they did not like about the company they are working for, aiming at constructive criticism and room for improvement, most people might say that there is nothing to improve. Actually, they might have a list of things they do not like, but they are afraid that it will autumn back on them negatively. And even if this is not the case, giving negative feedback always seems rude and is thus avoided.
Cognitive bias
Cognitive bias stems from our inability to judge information completely objective. This causes us to weight different input in different ways because of our individual experiences or mindsets. There are a few subtypes of this kind of research bias, that will be explained in the following.
- Anchoring bias happens when our mind sticks to one piece of information, mostly given at the beginning of a conversation. It functions as an anchor and as the basis for the whole following conversation, influencing our answers. To avoid this bias, we have to double-check our thought process or, as an interviewer, mix up the order of questions. When the follow-up questions are less related to the first one, the interviewee might let go of their anchor.
Example
An example for this bias in everyday life can be found at a marketplace. There, the seller places a base price for an item. The mind of the buyer remembers that price and will deem everything lower than that as a good deal, while the actual worth of the item might be even lower.
- The framing effect appears when the question is not presented neutrally and the participant gets the impression that one option is positive and the other one negative. This is also related to the interviewer bias, as the way we phrase or emphasize something might indicate our own beliefs and thus lead the participant to a certain answer. Avoid this research bias by staying impartial and neutral in your phrasing and body language.
Example
This type of research bias is highly influenced by our tone of voice. For example, if you ask a question like “Do you prefer sweet or salty snacks?” The word you emphasize in this sentence indicates which one you may prefer, and thus leads the interviewee to also give this as an answer to please you.
- The Actor-observer bias is a phenomenon, where when a person acts themselves, they tend to believe their actions are based on external influence. When they observe the act of other people, however, they link their actions more likely to their personal abilities. This is why people tend to evaluate themselves less positive than others or downplay their achievements. To avoid this kind of research bias, try to ask the participants what lead them to their beliefs or actions, or ask follow-up questions to find out about the causes yourself.
Example
If you can complete a difficult task, you may say that this was due to coincidence or luck, while if someone else does the same task, you might think they are especially intelligent or skilful in this area.
- Availability bias or availability heuristic describes the fact that we remember the most recent or vivid events best. This, however, also means that these events will have the biggest influence on the answers of your participants, clouding their judgment. To avoid this research bias, try asking them how they came to this conclusion to find out why they answer in a certain way. This reflection may help you and them recognise where their opinion is coming from and if it is objective and valid.
Example
For some years, you have been planning a cruise on a big ship for holiday and were looking forward to it. A few weeks before the start of the trip, however, there was another ship in the news that sunk tragically. Although this does not happen frequently, you now hesitate to go on your cruise because you are afraid the ship might sink too.
- Confirmation bias describes the phenomenon that humans always seek information that confirms their beliefs and neglect facts that contradict them. This is also the case for our memory, meaning that we remember facts supporting our view better than those contradicting them. A way to avoid this research bias is to confront the participants with information that is against their beliefs and see how they react to that or if it changes their opinion.
Example
Given you think that people wearing bright colors for clothing are mostly extroverted. Every time you meet a person with colorful clothes and find out they are extroverted, this seems to confirm your beliefs even more. At the same time, you tend to forget about every person you met that was introverted but still wearing colorful clothing.
- The Halo Effect happens when we build our opinion of something based on a single trait. This can refer to people but also brands, items or specific actions. Avoiding this research bias is difficult because oftentimes we do not know what influenced our opinion, but being confronted with a contradicting opinion might help to clear this up.
Example
When you meet someone for the first time, you usually get a first impression of them based on their momentary behaviour and manner of speech. People tend to stick to that first impression for a very long time and may not realise that the person is actually quite different from what we think they are. However, we do not notice this because our mind sticks to this first impression or single trait, we noticed at the first meeting.
- The Baader-Meinhof phenomenon, also known as “frequency illusion”, is when we acquire new information and suddenly seem to be confronted with it more often. This can be the case with a certain word or term, but also people, items, or brands. Thus, we may judge the frequency of something wrong because our focus on it shifted. Avoiding this bias is almost impossible because we cannot influence our perception. The only possibility is to note down our experiences to gain exact numbers that are valid.
Example
You recently got your divers license and are now free to drive around the city. Suddenly, you begin to notice how many tuition cars are on the streets and wonder if it has always been that many. Of course, the average of tuition cars on the streets does not change much over time, you just start to pay attention to this detail because it relates to you.
Procedural bias
Procedural bias happens when the procedure of the study influences the results. This can, for example, happen when participants do not have enough time to think about survey questions properly, or when the sampled participants are not interested in the survey and thus answer more quickly to get it done. To avoid this type of research bias, you can consider having non-timed surveys or sampling your participants more fittingly.
Example
You want to conduct a survey about the working environment around different companies, and therefore interview employees in their lunch break. This may cause the participants to answer more quickly or do not think enough about the questions because they only have a set amount of time for their break. They would like to enjoy this time or use it to eat, not answer questions from you.
Chronology bias
Chronology bias appears in long-term studies. It is caused by not considering the external change that influences the study when you watch development in people. Conducting long-term studies is thus very difficult because the change of environment or technical progress highly affects personal or skill development. This makes it harder to conclude whether the development is solely in the person themselves, or if the advancement of external factors contributed to the change in results.
Example
Given your research is about how working speed changes along with experience over a couple of years, you pick your participants and let them monitor their speed for certain actions. Their speed does indeed increase. However, this is also due to the fact that new technology is used to simplify some steps, which contributes highly to their speed in working.
Post-trial research bias
Even after you finished conducting your study, research bias can still affect your results. Whether the answers do not satisfy you or you realised that some information was not considered before your survey, even your opinion can influence the results. This is why during the step of analysing the results you need to be wary of the following types of research bias.
Confounding
Confounding can happen when the researcher does not consider every factor that can possibly affect the study. Occasionally, you want to prove the connection of two circumstances, while the actual influence comes from a third external influence you did not consider.
To avoid confounding, you have to take in every factor that could possibly impact your study. However, this can be an endless search, which is often not even possible. Thus, it can help to note details about your participants’ life and later compare them to find possible influences afterward.
Example
Your study is investigating the connection between an unhealthy diet and coronaries, wherefore you look at patients who already had a heart attack or are diagnosed as high risk. However, most of the patients you interviewed are also smokers, which is a factor that you did not include into your study but may highly affect the risk of coronaries.
Analysis bias
Analysis bias happens when you tend to focus only on the data that confirms your hypothesis or perspective, while leaving negative results aside. This can happen either consciously or subconsciously and will warp the results of your study negatively.
Avoiding this bias is highly necessary and can happen through staying objective. You need to focus on the facts given in the answers by the participants and double check if you did not exclude any important information.
Example
If you conduct a study and the results do not completely prove your hypothesis, some researchers then neglect those negative results and present only the positive ones to the public. This can also happen subconsciously, when you deem the participants with the negative results as less valid for some reason.
Publication bias
Publication bias refers to the fact that some studies will not get published if their results are not satisfying to the researcher. It can be caused by the fear of negative feedback or reduced sponsoring. When the sponsors hold back the study because the results do not fit them, it is called citation bias.
In the case of personal issues, it is important to be brave enough to state even unpleasant or unsatisfying results. When sponsors hold back a study, there may not be much you can do, except trying to persuade or convince them of the importance of the study.
Example
You are conducting a study about a possible medication for an illness. However, as you review the results, you realise that the patients did not experience a significant improvement in their wellbeing. Thus, you might be hesitant to publish your study because, on the one hand, you did not achieve the intended results and additionally, the company that produces the medication might not be amused with you publishing it.
How to avoid research bias
Knowing how to avoid any research bias is essential to keep your study valid and representative. While every bias may demand a different method to avoid it, there are a few general ones that apply to most types of research bias.
- The most important thing is to sample your participants randomly. Even when you need to seclude a certain group (age, social status, financial status, …), make sure that the rest is as diverse as possible. For example, if you conduct a study among people of a certain age, other factors like family background or wealth should be different in each participant.
- Try to keep a neutrality to your whole interview. This refers to the phrasing of your questions as well as your facial expression, body language and tone of voice while speaking. The environment of your interview might also influence the answers of the participant, for example if the surroundings distract them too much or if the time is convenient for them.
- When phrasing questions, make sure they are open-ended and not leading to a certain answer. This way, you can reduce many research biases stemming from personal interaction with the participant, such as social desirability bias.
- Include, if possible, personal data of the participants into your analysis. This way you can search for patterns of answers from interviewees with similar backgrounds or experiences.
- Use different types of sources and studies. This ensures that even if one study might be corrupted by bias, you have other ones to compare it to.
Especially in academic writing, you do not only need to be wary of research bias when conducting your study, but also when using existing studies as knowledge sources. Publication bias may be a big topic you might come across, leading you to wrong conclusions if contrary studies have never been published. However, knowing how all those different types of research bias work may also help you identify them or at least question every study a little before using it as a source.
Overview of all research bias articles
FAQs
Research bias is an umbrella term for many subtypes of influences that can warp the results of studies. As these flaws reduce the validity of results, researchers have to try their best to avoid bias in any study.
Research bias can influence a study before, during, and even after the trial. There are many factors contributing to how participants may answer questions and how you yourself view these results. Thus, it is important to know different types of research bias, to avoid them when conducting a study.
Although there are various countermeasures against the different types of research bias, you will most likely not be able to avoid it completely. This is because not every bias is consciously applied, and thus we may not even notice it happening.
The Hawthorne effect describes the phenomenon when participants of a study change their behaviour after realising they are being watched. This is why most studies are conducted by giving participants a pseudo-topic that is different from the actual main question, so they will not look into what the study is about.
The Baader-Meinhof phenomenon is a very common experience, where a piece of knowledge we recently acquired, seems to suddenly appear everywhere in our surroundings. This can be a word or an object or even a person. For example, after getting your driving license, you may see tuition cars more often on the streets than before.
